Datasets:
J0nasW
/
paperswithcode

Dataset card Data Studio Files Community
1
paperID
stringlengths
36
36
pwc_id
stringlengths
8
47
arxiv_id
stringlengths
6
16
⌀
nips_id
float64
url_abs
stringlengths
18
329
url_pdf
stringlengths
18
742
title
stringlengths
8
325
abstract
stringlengths
1
7.27k
⌀
authors
stringlengths
2
7.06k
published
stringlengths
10
10
⌀
conference
stringlengths
12
47
⌀
conference_url_abs
stringlengths
16
198
⌀
conference_url_pdf
stringlengths
27
199
⌀
proceeding
stringlengths
6
47
⌀
taskID
stringlengths
7
1.44k
areaID
stringclasses
688 values
embedding
stringlengths
9.26k
12.5k
umap_embedding
stringlengths
29
44
7b50646a-4740-4afe-adfa-58125079cb0b
latent-visual-cues-for-neural-machine
1811.00357
null
https://arxiv.org/abs/1811.00357v2
https://arxiv.org/pdf/1811.00357v2.pdf
Latent Variable Model for Multi-modal Translation
In this work, we propose to model the interaction between visual and textual features for multi-modal neural machine translation (MMT) through a latent variable model. This latent variable can be seen as a multi-modal stochastic embedding of an image and its description in a foreign language. It is used in a target-language decoder and also to predict image features. Importantly, our model formulation utilises visual and textual inputs during training but does not require that images be available at test time. We show that our latent variable MMT formulation improves considerably over strong baselines, including a multi-task learning approach (Elliott and K\'ad\'ar, 2017) and a conditional variational auto-encoder approach (Toyama et al., 2016). Finally, we show improvements due to (i) predicting image features in addition to only conditioning on them, (ii) imposing a constraint on the minimum amount of information encoded in the latent variable, and (iii) by training on additional target-language image descriptions (i.e. synthetic data).
['Iacer Calixto', 'Miguel Rios', 'Wilker Aziz']
2018-11-01
latent-variable-model-for-multi-modal
https://aclanthology.org/P19-1642
https://aclanthology.org/P19-1642.pdf
acl-2019-7
['multimodal-machine-translation']
['natural-language-processing']
[ 4.77817625e-01 4.30936366e-01 -3.27052385e-01 -1.91550151e-01 -1.05639219e+00 -6.19912386e-01 1.36749184e+00 -1.74261719e-01 -4.46479827e-01 6.21338964e-01 3.38901579e-01 -2.34771192e-01 4.63137627e-01 -5.23189902e-01 -1.29391181e+00 -6.36003911e-01 3.59083027e-01 5.98128140e-01 -1.55773923e-01 1.13221921e-01 -1.01962991e-01 2.19845884e-02 -1.29951513e+00 6.27047062e-01 4.30552840e-01 9.28228676e-01 5.86951911e-01 6.80659950e-01 1.28883615e-01 8.32468688e-01 -9.09324139e-02 -6.31577611e-01 4.17675786e-02 -5.73252380e-01 -8.18907499e-01 5.33934355e-01 4.69215572e-01 -2.62128174e-01 -2.64159530e-01 8.14268649e-01 1.02608919e-01 -1.17371351e-01 1.16323841e+00 -1.24711478e+00 -1.06631947e+00 3.60366613e-01 -4.92132157e-01 -1.65612265e-01 1.47839710e-01 3.51079941e-01 1.26235986e+00 -1.18943763e+00 1.03975129e+00 1.28074694e+00 2.88295627e-01 6.34767830e-01 -1.77558422e+00 -2.96946049e-01 2.15546012e-01 3.07880074e-01 -1.09135568e+00 -7.58452356e-01 6.76714778e-01 -7.62322545e-01 9.34901476e-01 4.53890152e-02 4.08694625e-01 1.63360834e+00 1.18649766e-01 1.15313363e+00 1.07118630e+00 -7.41202533e-01 -1.16844308e-02 3.97551417e-01 -5.50282598e-01 8.07802141e-01 -4.94828969e-01 1.16403840e-01 -5.60027719e-01 -8.73944834e-02 7.29010880e-01 -1.57329857e-01 -8.95422846e-02 -7.17644393e-01 -1.45679402e+00 1.22313130e+00 1.95893466e-01 1.40497446e-01 -4.77405488e-01 3.91801387e-01 2.63030976e-01 2.88954556e-01 6.89972937e-01 2.82500032e-02 -4.49544638e-01 -4.96784598e-02 -1.02470446e+00 1.87524445e-02 3.46790761e-01 1.02594912e+00 8.58967900e-01 1.35724545e-01 -3.84500146e-01 8.07453692e-01 4.96338218e-01 7.74473190e-01 5.23579717e-01 -9.39766824e-01 6.92944467e-01 7.63118491e-02 1.21467322e-01 -4.38868701e-01 1.75229028e-01 -1.01921126e-01 -5.21989703e-01 1.41183227e-01 2.42183879e-01 3.93213928e-02 -1.17348206e+00 2.25792146e+00 -1.20527044e-01 1.22231409e-01 1.96255878e-01 7.24180162e-01 4.36930925e-01 8.74489605e-01 9.63619649e-02 -2.89548010e-01 1.11247599e+00 -1.13772190e+00 -5.72313726e-01 -2.97773719e-01 5.89359224e-01 -7.45049059e-01 1.22000444e+00 1.78319708e-01 -1.33682406e+00 -4.21360821e-01 -8.19633901e-01 -4.25943762e-01 -2.43977621e-01 4.32465494e-01 4.32854265e-01 2.50552237e-01 -1.28019810e+00 3.31646085e-01 -1.06354594e+00 -2.41925091e-01 4.30882692e-01 1.48997217e-01 -5.73259771e-01 -2.54655629e-01 -1.00342274e+00 1.05053365e+00 -1.23612501e-01 -1.84676543e-01 -1.38280284e+00 -3.98847491e-01 -1.25085568e+00 -1.21606566e-01 1.44515038e-01 -9.31224585e-01 1.15095234e+00 -1.52220488e+00 -1.59367168e+00 1.04648721e+00 -5.11078000e-01 -4.10113603e-01 6.93152428e-01 -6.21443577e-02 4.47017066e-02 3.84430170e-01 2.81477779e-01 1.13455057e+00 1.14605427e+00 -1.41205549e+00 -2.64934570e-01 -1.28648952e-01 1.66524276e-01 2.28126317e-01 -1.12324394e-01 -1.19162202e-02 -5.03149569e-01 -6.24934793e-01 -1.68098643e-01 -1.21004605e+00 -5.21144569e-02 3.23441118e-01 -4.09207374e-01 4.12434526e-02 6.51103377e-01 -8.37114751e-01 2.91081995e-01 -2.20018744e+00 8.82978261e-01 -1.11807533e-01 2.11167201e-01 -1.56981006e-01 -4.42122042e-01 3.93296987e-01 -1.53726473e-01 2.00470895e-01 -3.46882135e-01 -9.40645158e-01 2.41674349e-01 3.74938905e-01 -5.36451042e-01 5.80040991e-01 5.37463367e-01 1.35631156e+00 -6.27330542e-01 -5.19261420e-01 4.20560211e-01 8.16058159e-01 -5.83339870e-01 5.37022911e-02 -6.55434847e-01 7.24905610e-01 -1.21245965e-01 1.91542402e-01 1.81075498e-01 -4.87130374e-01 1.37266694e-02 -1.53855428e-01 -1.70376360e-01 3.95895749e-01 -7.44416118e-01 1.89434993e+00 -9.58946645e-01 8.64233255e-01 -1.22499421e-01 -9.86262560e-01 3.14550310e-01 7.74055004e-01 3.13163012e-01 -7.06323802e-01 1.08410679e-01 2.37238370e-02 -3.42165112e-01 -3.85693192e-01 1.14209160e-01 -3.10701162e-01 -4.42784615e-02 6.99181020e-01 3.95756245e-01 -8.30749795e-03 1.07676908e-01 4.04063582e-01 6.46953821e-01 4.80895549e-01 -7.60824680e-02 -3.93399782e-02 2.98797607e-01 -2.52156943e-01 2.48369932e-01 5.04467368e-01 2.23842636e-01 6.36788666e-01 6.74259186e-01 -5.05669490e-02 -1.49840510e+00 -1.26808679e+00 -6.19818494e-02 1.04041326e+00 -3.81577879e-01 -1.93468854e-01 -5.04701078e-01 -5.02505720e-01 -1.96246296e-01 9.08152103e-01 -9.97698426e-01 -1.00208797e-01 -2.86597431e-01 -2.62886226e-01 3.35337162e-01 2.94471622e-01 3.99747491e-02 -8.88580739e-01 -4.81182724e-01 -6.82738423e-02 -4.55893278e-01 -1.25879776e+00 -6.03828609e-01 2.38743916e-01 -7.08209813e-01 -4.34320837e-01 -9.81183589e-01 -7.65340686e-01 7.38812625e-01 2.09310148e-02 1.02924836e+00 -3.01944405e-01 -3.28111649e-02 5.85440576e-01 -2.54395097e-01 -1.46806121e-01 -5.70030630e-01 -3.24872024e-02 -3.45087424e-02 2.55952567e-01 -1.19833268e-01 -6.27617955e-01 -3.60459238e-01 -6.43095048e-03 -9.31989968e-01 7.26916015e-01 7.88650572e-01 9.69780803e-01 6.53318763e-01 -6.71939015e-01 1.38267756e-01 -8.86866450e-01 4.68819104e-02 -4.61767942e-01 -4.87835199e-01 3.40403289e-01 -4.75681365e-01 4.15563762e-01 4.35712159e-01 -7.21997440e-01 -7.73300052e-01 9.02948454e-02 7.64730498e-02 -8.59591722e-01 5.67491055e-02 4.81304646e-01 -2.36293390e-01 2.13770211e-01 3.05286318e-01 5.31301975e-01 3.56152393e-02 -2.85463512e-01 7.64422596e-01 4.02283221e-01 2.36282915e-01 -4.85939920e-01 1.03476167e+00 5.31411409e-01 -1.90834738e-02 -5.66043973e-01 -6.19168282e-01 -1.95698384e-02 -1.00329602e+00 -5.94188608e-02 1.22937953e+00 -1.18243361e+00 -3.69630694e-01 1.08521707e-01 -1.21809590e+00 -6.28184915e-01 -1.52538389e-01 6.23334885e-01 -9.45736885e-01 1.69461384e-01 -6.99413121e-01 -7.11864591e-01 1.92192823e-01 -1.34232688e+00 1.35677266e+00 -5.33389330e-01 -1.57675743e-01 -1.19823599e+00 -1.37774155e-01 4.25761640e-01 3.39416802e-01 2.09042341e-01 1.09554803e+00 -1.51609614e-01 -9.68368173e-01 3.68898176e-02 -2.93267995e-01 2.63997704e-01 -1.06169380e-01 -1.04608566e-01 -1.00061774e+00 -4.29785073e-01 -1.99209880e-02 -7.95944154e-01 1.09122992e+00 3.98730963e-01 7.56972075e-01 -5.02852917e-01 -1.96281835e-01 6.33128345e-01 1.48876595e+00 -2.44371995e-01 4.61885065e-01 1.73572630e-01 8.58270705e-01 6.46262348e-01 1.61626175e-01 1.08010881e-01 6.97545648e-01 9.46899474e-01 2.23527461e-01 -2.59960473e-01 -1.83816373e-01 -4.07932997e-01 6.82372391e-01 8.36681128e-01 1.02920212e-01 -3.25890750e-01 -7.60193586e-01 7.98856080e-01 -1.84570622e+00 -9.30617511e-01 -1.01553664e-01 1.90012717e+00 9.79490578e-01 -5.04808910e-02 -8.41562357e-03 -3.62152606e-01 4.58878905e-01 2.33352348e-01 -5.44681609e-01 -4.10835862e-01 -3.11254054e-01 1.91205189e-01 5.40715575e-01 7.77364612e-01 -1.00524700e+00 1.09661448e+00 5.97415638e+00 5.30209661e-01 -1.26085949e+00 5.55965841e-01 5.57520628e-01 -2.50089169e-01 -7.90629029e-01 1.17230847e-01 -3.53238374e-01 3.04578036e-01 1.10286236e+00 8.07480738e-02 5.97199321e-01 3.96833062e-01 4.21581902e-02 7.96879455e-02 -1.29750562e+00 8.16057086e-01 4.37136531e-01 -1.24473536e+00 2.96483308e-01 3.63195300e-01 7.12712765e-01 2.06744239e-01 5.91797590e-01 2.71039575e-01 4.10078466e-01 -1.20277011e+00 1.17697740e+00 4.77397293e-01 1.14250052e+00 -4.16814446e-01 2.45882303e-01 4.27666306e-01 -7.76334584e-01 1.96115106e-01 -8.81641731e-02 7.45822489e-02 2.16652766e-01 7.56521076e-02 -4.76314753e-01 3.58373672e-01 2.45967373e-01 7.08965480e-01 -5.49141288e-01 2.68863827e-01 -5.07063389e-01 7.00977683e-01 -6.36901781e-02 2.94985592e-01 4.38419223e-01 -8.69166180e-02 4.69950020e-01 1.06287324e+00 1.39475957e-01 -3.09341013e-01 1.56305179e-01 1.21191978e+00 -7.75655136e-02 5.80305010e-02 -6.44473791e-01 -2.24551320e-01 -1.11117676e-01 9.05399919e-01 -4.42230225e-01 -2.33320549e-01 -7.39265501e-01 1.53167963e+00 4.43521321e-01 6.29818320e-01 -8.25325787e-01 2.81927317e-01 2.69467562e-01 2.71028765e-02 7.53964305e-01 -3.80000830e-01 -2.20974118e-01 -1.39623296e+00 3.77268414e-03 -4.82209444e-01 -1.27426922e-01 -1.09085464e+00 -1.02865458e+00 5.55909991e-01 -1.79040916e-02 -1.09288192e+00 -6.94377363e-01 -6.23147666e-01 -1.81413978e-01 1.26713681e+00 -1.52608895e+00 -1.90001416e+00 4.46950287e-01 6.56808019e-01 5.69849253e-01 -1.95242628e-01 1.03067493e+00 1.85917094e-01 -3.11084837e-01 7.08836615e-01 2.53491879e-01 4.99609746e-02 6.25805497e-01 -1.30988574e+00 4.82576162e-01 7.57659197e-01 5.74789226e-01 4.44328547e-01 8.15822184e-01 -4.22468752e-01 -1.41853416e+00 -1.12288666e+00 1.15156686e+00 -7.97590733e-01 7.45291889e-01 -8.36772025e-01 -6.28542185e-01 1.34989440e+00 5.43008268e-01 -7.56592825e-02 6.59566760e-01 1.67740174e-02 -5.56888998e-01 4.80437666e-01 -8.35710824e-01 6.37699068e-01 6.24073803e-01 -1.02807629e+00 -3.25036913e-01 4.63008165e-01 8.90037358e-01 -2.85647869e-01 -6.30138874e-01 8.83063897e-02 5.02283752e-01 -5.11516035e-01 8.74333024e-01 -6.19431138e-01 9.80456829e-01 7.85450786e-02 -4.33856130e-01 -1.24615824e+00 -2.94419229e-01 -3.68717134e-01 -1.18881546e-01 1.06716466e+00 8.10793400e-01 -2.22634569e-01 4.67619061e-01 5.24535239e-01 -4.96935062e-02 -7.71828175e-01 -1.10645151e+00 -5.28127909e-01 4.26384568e-01 -4.32200998e-01 -4.37406227e-02 9.91718173e-01 -2.95546621e-01 6.36513948e-01 -7.60388315e-01 6.13192543e-02 5.74558496e-01 3.31082451e-03 7.07636297e-01 -7.37396002e-01 -6.45111084e-01 -2.85332799e-01 -2.92598277e-01 -1.12784433e+00 4.86069858e-01 -1.25752461e+00 -3.61759961e-02 -1.58754575e+00 6.32875383e-01 8.16797465e-02 -1.48997679e-01 5.11278987e-01 6.04057983e-02 5.19680858e-01 2.56154537e-01 4.76541132e-01 -6.13157213e-01 8.53424191e-01 1.17416358e+00 -2.14809164e-01 2.64138520e-01 -2.58554965e-01 -4.48961347e-01 4.53882843e-01 3.39205861e-01 -4.56092119e-01 -3.77780229e-01 -8.05578411e-01 2.16886446e-01 3.92753839e-01 8.23370636e-01 -2.03579962e-01 -3.13480124e-02 -2.49328300e-01 5.02474785e-01 -2.54673421e-01 7.87958562e-01 -5.91309607e-01 -2.43924856e-02 2.84585506e-01 -6.96256757e-01 1.66707292e-01 9.80453640e-02 6.36252463e-01 -1.47147998e-01 -8.64192769e-02 7.89419293e-01 -1.39168963e-01 -3.61506134e-01 3.25969905e-01 -5.36695182e-01 -1.70791835e-01 8.54912341e-01 -1.15797527e-01 -3.98591161e-03 -5.60099661e-01 -1.01141942e+00 1.77560493e-01 7.55035996e-01 4.63404655e-01 5.59472501e-01 -1.63873887e+00 -8.08351040e-01 2.19820708e-01 3.59075785e-01 -3.98596913e-01 1.04463398e-01 9.16811049e-01 -2.72340756e-02 4.94958609e-01 -1.26586393e-01 -7.62300014e-01 -1.10674083e+00 5.88412762e-01 5.47769181e-02 -1.79013863e-01 -5.68825543e-01 8.03136766e-01 5.10277212e-01 -3.80102038e-01 9.45397839e-02 8.49467888e-03 6.63545802e-02 -7.59886280e-02 2.01372504e-01 -3.98032278e-01 -3.91520232e-01 -1.17874539e+00 -1.72901824e-01 3.98095727e-01 -1.95679948e-01 -1.05587757e+00 1.36595321e+00 -3.15088242e-01 9.97431874e-02 1.07366824e+00 1.59869611e+00 -2.37643756e-02 -1.39896953e+00 -5.70804358e-01 -1.55991554e-01 -2.87386835e-01 2.09032699e-01 -8.08601856e-01 -8.05945396e-01 1.13908565e+00 6.37709498e-01 -3.38617355e-01 7.43296146e-01 4.20785367e-01 6.83411360e-01 7.17588291e-02 2.32892737e-01 -8.40597808e-01 2.93058783e-01 4.10209328e-01 9.47873592e-01 -1.34577727e+00 -2.06770957e-01 2.65149511e-02 -9.71941710e-01 8.09809566e-01 1.21601187e-01 -2.02020295e-02 5.00513732e-01 -9.76416618e-02 5.33498228e-02 -8.13665912e-02 -1.23561859e+00 -2.18637407e-01 5.27777255e-01 5.39456904e-01 4.63506758e-01 1.07520983e-01 4.75870222e-02 2.21267551e-01 -1.28852114e-01 -1.83627710e-01 2.89443254e-01 6.36040211e-01 -1.06833756e-01 -1.30941308e+00 -4.84749563e-02 3.70920449e-01 -4.79820430e-01 -4.34825689e-01 -3.51924956e-01 6.17383718e-01 1.67508513e-01 7.11075187e-01 9.13048759e-02 -2.46256754e-01 -4.23507839e-02 2.45891646e-01 8.01455617e-01 -8.72038722e-01 -6.80041462e-02 2.27368876e-01 -1.55477360e-01 -3.25976610e-01 -4.11063910e-01 -1.14884067e+00 -7.21310198e-01 6.50256202e-02 -1.69747919e-01 -3.18102352e-02 8.90945375e-01 1.25737536e+00 1.52041584e-01 4.22309577e-01 5.90413392e-01 -8.95383775e-01 -3.95821154e-01 -1.16663730e+00 -2.40732983e-01 5.25795519e-01 5.74928522e-01 -7.23863602e-01 -3.33747625e-01 5.85516274e-01]
[11.283273696899414, 1.4756442308425903]
931d8108-a781-4115-91ae-b0bd09e0de98
learning-features-of-music-from-scratch
1611.09827
null
http://arxiv.org/abs/1611.09827v2
http://arxiv.org/pdf/1611.09827v2.pdf
Learning Features of Music from Scratch
This paper introduces a new large-scale music dataset, MusicNet, to serve as a source of supervision and evaluation of machine learning methods for music research. MusicNet consists of hundreds of freely-licensed classical music recordings by 10 composers, written for 11 instruments, together with instrument/note annotations resulting in over 1 million temporal labels on 34 hours of chamber music performances under various studio and microphone conditions. The paper defines a multi-label classification task to predict notes in musical recordings, along with an evaluation protocol, and benchmarks several machine learning architectures for this task: i) learning from spectrogram features; ii) end-to-end learning with a neural net; iii) end-to-end learning with a convolutional neural net. These experiments show that end-to-end models trained for note prediction learn frequency selective filters as a low-level representation of audio.
['Sham Kakade', 'Zaid Harchaoui', 'John Thickstun']
2016-11-29
null
null
null
null
['music-transcription']
['music']
[ 3.87685210e-01 -4.01189029e-01 -7.98897222e-02 -1.96406797e-01 -1.33060455e+00 -1.04702890e+00 2.41195962e-01 -3.73621017e-01 -1.84171855e-01 4.24317092e-01 5.21815717e-01 4.21871185e-01 -3.33424538e-01 -2.92745680e-01 -4.01262462e-01 -4.20776010e-01 -5.82180262e-01 4.91761595e-01 -1.65961847e-01 6.36668727e-02 3.82033080e-01 1.93822250e-01 -1.44424391e+00 1.03515315e+00 -1.04283191e-01 1.45877838e+00 -7.21633509e-02 1.16315091e+00 4.62602913e-01 1.19600999e+00 -7.23832905e-01 -1.50183722e-01 3.35885137e-01 -6.55028105e-01 -1.05719888e+00 -2.52041221e-01 8.68625820e-01 7.80445710e-02 -1.67097479e-01 7.75756299e-01 9.86272275e-01 5.14815092e-01 5.02081871e-01 -8.66228104e-01 -3.57466131e-01 1.40955496e+00 1.82501134e-02 2.08696112e-01 3.88189584e-01 -1.06074288e-01 1.69477665e+00 -8.71488929e-01 3.61017704e-01 8.57777536e-01 1.20561934e+00 4.50397134e-01 -1.25199389e+00 -9.27188158e-01 -4.09926057e-01 3.39025140e-01 -1.23803449e+00 -8.04551899e-01 1.09628594e+00 -5.42214036e-01 9.07936811e-01 4.51866120e-01 8.67839217e-01 1.44132233e+00 -1.93879589e-01 7.99963534e-01 8.54135692e-01 -3.49222958e-01 6.62529618e-02 -4.50130343e-01 -2.13958353e-01 3.68021190e-01 -9.83895361e-01 4.11877602e-01 -1.34536469e+00 -4.47603583e-01 8.14440370e-01 -4.69783664e-01 -2.96731055e-01 3.83501738e-01 -1.64189982e+00 4.56148416e-01 3.27794254e-01 3.84148866e-01 -1.77528590e-01 5.68054795e-01 9.58622873e-01 4.66710687e-01 3.24156314e-01 9.56476629e-01 -8.00565898e-01 -6.06081903e-01 -1.30445814e+00 4.90023255e-01 7.70728052e-01 5.28946877e-01 2.05392107e-01 3.75728697e-01 -4.07959282e-01 1.35410261e+00 -7.52893016e-02 2.63740662e-02 8.19765806e-01 -1.26999784e+00 2.57208139e-01 -2.34738097e-01 -1.57776698e-01 -4.32699412e-01 -7.76159465e-01 -1.00900722e+00 -7.61678517e-01 1.80022374e-01 2.48802662e-01 4.80543263e-02 -2.92985380e-01 1.73371267e+00 -2.49575615e-01 9.07287061e-01 -2.23032042e-01 1.02552414e+00 9.28326547e-01 5.74134111e-01 -2.76645541e-01 -2.57688820e-01 1.17442179e+00 -1.59288776e+00 -6.09415472e-01 1.58632576e-01 2.34725565e-01 -1.17631924e+00 1.28750336e+00 1.12381637e+00 -1.22029459e+00 -1.17779994e+00 -9.07764256e-01 -9.95872468e-02 3.05547595e-01 6.07491553e-01 4.53685164e-01 2.12609425e-01 -7.37201929e-01 1.30864692e+00 -4.51867312e-01 2.14290366e-01 3.92621875e-01 2.83990860e-01 5.73979728e-02 6.15985632e-01 -8.68755162e-01 2.05419257e-01 3.84537011e-01 -5.91225177e-03 -1.50634313e+00 -1.16222346e+00 -2.65934020e-01 1.53897345e-01 4.30071205e-02 -5.56511700e-01 1.95664132e+00 -1.22929776e+00 -1.95135725e+00 1.18271911e+00 3.93327117e-01 -5.26643038e-01 1.90569177e-01 -6.79106772e-01 -7.38401830e-01 -9.66033861e-02 1.45070612e-01 3.23302388e-01 8.18621457e-01 -8.86875570e-01 -7.70505548e-01 -6.34109899e-02 -2.76944995e-01 1.21319741e-01 -2.55219907e-01 4.11106199e-01 -1.97978362e-01 -1.38532495e+00 -2.74037607e-02 -1.22140539e+00 1.38594046e-01 -5.55199862e-01 -8.24336469e-01 -4.04326260e-01 3.31844121e-01 -6.40902281e-01 1.31579518e+00 -2.43892002e+00 1.29865617e-01 -5.74999452e-02 -2.07825124e-01 -1.37378022e-01 -4.51606393e-01 3.44740570e-01 -2.94889569e-01 -3.53433013e-01 -1.77362144e-01 -5.42480886e-01 2.27919906e-01 -4.94089812e-01 -8.97039235e-01 3.15292835e-01 -4.82670575e-01 5.70682466e-01 -8.58454108e-01 -1.64797261e-01 2.86105331e-02 2.50113308e-01 -6.34551048e-01 3.50424588e-01 -2.81568736e-01 8.07035267e-01 1.47909438e-02 7.91485190e-01 -9.41633955e-02 4.70494144e-02 -9.38912183e-02 -4.17524070e-01 -7.60617256e-02 9.96144295e-01 -1.19556868e+00 2.65713739e+00 -4.04432654e-01 6.81280375e-01 6.11949107e-03 -4.88115996e-01 9.57280338e-01 9.11714375e-01 7.56245196e-01 -1.27060667e-01 2.78730858e-02 4.30758566e-01 -6.19349480e-02 -3.40064317e-01 3.23206037e-01 -4.93436426e-01 -4.37878489e-01 5.72061598e-01 4.89125639e-01 -2.68795639e-01 -8.08310136e-02 -4.57018286e-01 1.27304220e+00 4.49919045e-01 -8.91618878e-02 -4.23172936e-02 3.53012085e-01 -2.19679698e-01 8.17612708e-01 4.82381821e-01 -1.01361163e-01 9.65251505e-01 -1.04321644e-01 -8.05414736e-01 -7.09722579e-01 -1.13308132e+00 -5.08189835e-02 1.94889557e+00 -5.57219207e-01 -9.45856988e-01 -3.07973742e-01 -2.43536845e-01 -6.33597225e-02 2.95800567e-01 -4.05420691e-01 4.80962992e-02 -6.70098722e-01 -2.30527103e-01 1.38171911e+00 5.98147988e-01 1.82179436e-01 -1.82930911e+00 -2.63995618e-01 4.85580951e-01 -4.80356187e-01 -7.36786783e-01 -8.06090176e-01 6.13859236e-01 -8.70799422e-01 -1.09175456e+00 -4.46189106e-01 -1.11180937e+00 -6.24594986e-01 -4.96822625e-01 1.47097361e+00 -4.26754355e-01 -4.07314301e-01 1.43332213e-01 -2.53596216e-01 -4.74545300e-01 -2.98984796e-01 3.72330189e-01 4.30499434e-01 6.16255552e-02 1.22556575e-01 -1.26146126e+00 -5.40845454e-01 1.87291846e-01 -3.93292874e-01 -2.29091778e-01 3.20130229e-01 6.04835570e-01 1.01326478e+00 -1.48510888e-01 8.93223524e-01 -7.74519384e-01 5.85703373e-01 1.39933959e-01 -2.72609502e-01 -9.09265280e-02 -2.30758846e-01 -4.88131434e-01 7.46590853e-01 -5.59591413e-01 -6.25620306e-01 2.48061746e-01 -3.88252616e-01 -7.60454178e-01 -1.89005852e-01 3.05223078e-01 1.52897000e-01 3.43343824e-01 1.05510259e+00 -8.44670609e-02 -6.99844718e-01 -1.07617402e+00 4.52766716e-01 8.08391333e-01 1.33977318e+00 -7.61253417e-01 4.66620624e-01 2.16801032e-01 -9.02911499e-02 -4.27298427e-01 -1.58707500e+00 -7.24962413e-01 -8.35763872e-01 -4.67150241e-01 7.19233394e-01 -1.28938115e+00 -1.04703057e+00 3.64274979e-01 -7.91737378e-01 -6.17312491e-01 -6.82678938e-01 9.60159123e-01 -1.25738263e+00 -2.56265014e-01 -1.27909470e+00 -6.57093465e-01 -7.37906218e-01 -4.75131035e-01 1.17597592e+00 -1.40334129e-01 -7.93667257e-01 -6.07196569e-01 8.63032401e-01 4.81054038e-01 3.02536134e-02 2.63983160e-01 7.08378553e-01 -5.76564848e-01 -3.70488912e-01 1.97878584e-01 4.12132710e-01 5.30749142e-01 -4.03539874e-02 -1.43379450e-01 -1.89134562e+00 -3.68544698e-01 -1.62104160e-01 -1.06535006e+00 1.14001954e+00 6.12022638e-01 1.58598018e+00 -1.42915353e-01 1.02040507e-01 1.10751832e+00 1.20676184e+00 -6.42404258e-02 2.69999951e-01 2.83894330e-01 6.47677839e-01 1.85613200e-01 4.54128385e-01 7.43489087e-01 -3.31029147e-01 8.56099188e-01 3.06662470e-01 3.21177095e-01 -5.14624238e-01 -4.20079052e-01 5.71964502e-01 1.41714025e+00 -5.30038238e-01 9.19585004e-02 -4.92327332e-01 6.15138948e-01 -1.68346822e+00 -1.26012850e+00 5.96494526e-02 2.02292943e+00 1.11555302e+00 -7.18693882e-02 7.38863528e-01 7.39262998e-01 5.20447850e-01 3.80233854e-01 -6.00326896e-01 -1.69349521e-01 -2.22880885e-01 7.17716217e-01 -9.50819328e-02 -7.04264864e-02 -1.56001735e+00 8.08988333e-01 6.90073586e+00 9.45427537e-01 -1.10371447e+00 2.22483024e-01 -3.70782167e-02 -8.29843104e-01 2.56821692e-01 -1.61022052e-01 -3.14221025e-01 1.33604616e-01 1.22199631e+00 2.54914284e-01 9.12200034e-01 9.27920938e-01 -7.66768679e-02 8.86266887e-01 -1.38003266e+00 1.40506411e+00 5.84903955e-02 -1.63948643e+00 -3.47978085e-01 -2.26924241e-01 8.16407442e-01 5.50899804e-01 4.05920893e-02 6.06345475e-01 6.47324026e-02 -1.10473061e+00 1.27582788e+00 5.67726254e-01 1.25615954e+00 -7.90475786e-01 2.32628524e-01 9.53841731e-02 -1.61692882e+00 -3.44095200e-01 -1.39621481e-01 -3.52429062e-01 2.94317991e-01 4.17319864e-01 -6.19814813e-01 3.67453068e-01 8.81037295e-01 1.34778106e+00 -1.93193659e-01 1.31653678e+00 -2.79870272e-01 1.33676875e+00 -1.86946224e-02 5.06438315e-01 3.49175334e-02 1.12163827e-01 6.86395884e-01 1.45084262e+00 4.11012948e-01 -2.90551960e-01 3.88065070e-01 7.30774641e-01 -5.00731051e-01 2.69797921e-01 -6.35198876e-02 -2.30045617e-01 4.56650048e-01 1.42981410e+00 -6.24243796e-01 -1.28929857e-02 -1.28995720e-02 9.27762091e-01 2.76376635e-01 1.14901401e-01 -7.32457101e-01 -4.55306828e-01 6.74991429e-01 -1.25330105e-01 1.56352639e-01 2.05194339e-01 -3.18352401e-01 -9.48616385e-01 -2.93334842e-01 -1.03431821e+00 6.80191100e-01 -9.70592856e-01 -1.71286082e+00 7.62172699e-01 -9.48713481e-01 -1.52728271e+00 -2.92273253e-01 -5.13799071e-01 -8.30842018e-01 6.78122461e-01 -1.13324118e+00 -1.09783900e+00 5.65721951e-02 7.89398611e-01 7.18333960e-01 -7.37970948e-01 1.58197296e+00 6.04846358e-01 -6.97426572e-02 5.25547922e-01 4.69428077e-02 4.88845825e-01 1.34149969e+00 -1.52292585e+00 1.89034641e-01 -1.03999267e-03 1.36206615e+00 4.99379300e-02 3.31221372e-01 -7.17495680e-02 -7.68820703e-01 -1.26172793e+00 8.32181156e-01 -5.83226204e-01 5.30704319e-01 -3.06261182e-01 -5.13845026e-01 9.02606964e-01 9.39046443e-02 -2.70298799e-03 1.41909838e+00 9.38773572e-01 -5.60152769e-01 -3.26024204e-01 -3.86278749e-01 5.50045595e-02 1.31256104e+00 -1.29900634e+00 -6.56841993e-01 6.24883413e-01 5.69863677e-01 -4.04860586e-01 -1.18992507e+00 4.85009313e-01 1.06204736e+00 -8.78331423e-01 1.03360295e+00 -7.49316931e-01 5.88634312e-01 -3.46861660e-01 -3.76651853e-01 -1.28782868e+00 -6.84468865e-01 -1.10241508e+00 -2.50903875e-01 1.28346598e+00 5.35570502e-01 5.15711784e-01 6.74112260e-01 -6.04618311e-01 -7.70125628e-01 -5.65821171e-01 -9.90723848e-01 -8.46200228e-01 -1.53802767e-01 -9.98847902e-01 3.93872291e-01 1.19178426e+00 6.87440187e-02 8.08270216e-01 -7.35922217e-01 -1.39034733e-01 3.67960572e-01 7.56312430e-01 6.61183476e-01 -1.67197084e+00 -8.29331040e-01 -7.18291640e-01 -3.64872277e-01 -8.86384249e-01 3.34540665e-01 -1.40533555e+00 7.59804025e-02 -1.04990315e+00 1.49821028e-01 -4.40084457e-01 -1.00420415e+00 4.76509452e-01 3.87610912e-01 1.03108335e+00 3.05952519e-01 6.68693244e-01 -7.10972607e-01 4.08541858e-01 9.11041081e-01 -3.21212560e-01 -3.24967384e-01 5.83455503e-01 -2.28042662e-01 1.04113507e+00 5.02388537e-01 -6.32122040e-01 -6.42181039e-02 -3.45602483e-01 5.33103406e-01 1.93228334e-01 2.86361724e-01 -1.49059021e+00 5.74653186e-02 2.84535795e-01 3.82409602e-01 -8.05474699e-01 7.22652435e-01 -2.85151213e-01 1.95879519e-01 1.43572032e-01 -1.11680377e+00 -3.88124973e-01 1.52559146e-01 4.63873506e-01 -6.50614023e-01 -9.02653486e-02 7.36198246e-01 -1.79083318e-01 -4.41391975e-01 3.45919102e-01 4.38294560e-02 2.00489908e-01 3.61590117e-01 3.75521451e-01 1.79837212e-01 -2.14287937e-01 -1.45686710e+00 -4.30001915e-01 -2.02779919e-01 6.64748430e-01 1.06788523e-01 -1.87271130e+00 -1.06427908e+00 -2.80398726e-02 2.78167635e-01 -4.49394047e-01 3.33205879e-01 4.46914762e-01 -2.97533035e-01 2.66814560e-01 -1.91914186e-01 -6.28581464e-01 -1.37194788e+00 3.05601090e-01 5.12601852e-01 -3.45913142e-01 -7.77579904e-01 1.55919802e+00 -1.40694045e-02 -6.53580070e-01 7.03005373e-01 -5.53439677e-01 -1.57357082e-01 1.74918607e-01 6.51738286e-01 2.76268303e-01 1.08391874e-01 -5.21472692e-01 -2.24656001e-01 5.87185800e-01 5.16497672e-01 -4.19456840e-01 1.51962197e+00 3.04630488e-01 3.91147062e-02 1.49517965e+00 9.39701080e-01 3.93442810e-01 -1.12795949e+00 -3.68311882e-01 5.72508015e-03 -2.19569981e-01 5.65438010e-02 -1.27806759e+00 -8.69345903e-01 8.68826270e-01 5.86993694e-01 1.63780361e-01 1.29693794e+00 -6.83773682e-02 1.12161791e+00 5.60781658e-01 2.74148703e-01 -1.31678855e+00 3.42837453e-01 6.07617319e-01 1.15875196e+00 -7.25991368e-01 -3.50356311e-01 1.65559947e-01 -6.18076265e-01 1.28218722e+00 5.47090136e-02 -4.06653404e-01 9.89993691e-01 2.18542919e-01 4.58116800e-01 -1.34888068e-01 -8.96730840e-01 -1.20006368e-01 7.11816847e-01 2.87650853e-01 9.22375739e-01 2.79252321e-01 4.35182303e-01 1.37087178e+00 -1.04180264e+00 2.10304752e-01 -1.18047267e-03 3.09194803e-01 -2.67210573e-01 -9.59429979e-01 -2.14515507e-01 2.83570021e-01 -1.02707362e+00 -1.86906949e-01 -5.74819326e-01 7.72974417e-02 7.02615678e-01 1.02247000e+00 7.69668631e-03 -6.62109494e-01 4.13079768e-01 4.14103687e-01 6.96182311e-01 -8.07074308e-01 -1.41536260e+00 5.51163495e-01 2.64904439e-01 -5.76294720e-01 -6.99617922e-01 -4.49570864e-01 -1.14215243e+00 2.20464528e-01 -1.68706849e-01 3.90871733e-01 5.44364512e-01 6.71508193e-01 9.00314972e-02 1.17846155e+00 8.67443085e-01 -1.17436683e+00 -5.48578620e-01 -1.59000289e+00 -1.35958469e+00 5.78832090e-01 3.71296197e-01 -2.47643948e-01 -3.40242654e-01 6.06305897e-01]
[15.85065746307373, 5.28391170501709]
fa73a4c4-9251-44b5-944e-900157a03803
personalized-graph-signal-processing-for
2302.02113
null
https://arxiv.org/abs/2302.02113v1
https://arxiv.org/pdf/2302.02113v1.pdf
Personalized Graph Signal Processing for Collaborative Filtering
The collaborative filtering (CF) problem with only user-item interaction information can be solved by graph signal processing (GSP), which uses low-pass filters to smooth the observed interaction signals on the similarity graph to obtain the prediction signals. However, the interaction signal may not be sufficient to accurately characterize user interests and the low-pass filters may ignore the useful information contained in the high-frequency component of the observed signals, resulting in suboptimal accuracy. To this end, we propose a personalized graph signal processing (PGSP) method for collaborative filtering. Firstly, we design the personalized graph signal containing richer user information and construct an augmented similarity graph containing more graph topology information, to more effectively characterize user interests. Secondly, we devise a mixed-frequency graph filter to introduce useful information in the high-frequency components of the observed signals by combining an ideal low-pass filter that smooths signals globally and a linear low-pass filter that smooths signals locally. Finally, we combine the personalized graph signal, the augmented similarity graph and the mixed-frequency graph filter by proposing a pipeline consisting of three key steps: pre-processing, graph convolution and post-processing. Extensive experiments show that PGSP can achieve superior accuracy compared with state-of-the-art CF methods and, as a nonparametric method, PGSP has very high training efficiency.
['Ning Gu', 'Li Shang', 'Peng Zhang', 'Tun Lu', 'Hansu Gu', 'Dongsheng Li', 'Jiahao Liu']
2023-02-04
null
null
null
null
['collaborative-filtering']
['miscellaneous']
[ 1.35331526e-01 -6.03070967e-02 2.40035757e-01 -2.53064662e-01 -2.22047135e-01 -2.86923677e-01 1.39982283e-01 3.29382688e-01 2.13888645e-01 8.47615376e-02 5.11510074e-01 -1.15670532e-01 -4.86319065e-01 -1.14334536e+00 -3.77434373e-01 -3.37020934e-01 -5.20020366e-01 -1.88119411e-01 3.89748424e-01 -3.01100999e-01 -1.36728376e-01 1.85072869e-01 -1.41259801e+00 5.84918737e-01 1.31260014e+00 1.08973467e+00 1.66931778e-01 7.30399251e-01 -2.47381374e-01 5.47733486e-01 -4.55359638e-01 -1.77535295e-01 1.76480994e-01 -7.90141761e-01 -3.97189200e-01 1.60325035e-01 2.08684474e-01 8.89147073e-02 -6.05740428e-01 1.38671076e+00 4.62963253e-01 6.79547310e-01 1.80795699e-01 -1.08350062e+00 -6.91829801e-01 9.00135577e-01 -3.27386498e-01 1.36280969e-01 8.68314624e-01 -1.70480043e-01 1.11017025e+00 -7.26555049e-01 4.67247486e-01 1.34076846e+00 9.76315439e-01 -9.79012027e-02 -1.32399678e+00 -4.74204093e-01 5.71420014e-01 3.26100826e-01 -1.36501360e+00 -2.45972676e-03 1.21870470e+00 -2.64788538e-01 4.99912322e-01 5.04315794e-01 1.04704261e+00 7.49932587e-01 -1.68136545e-02 7.98335791e-01 7.12961316e-01 -1.81709483e-01 3.26012485e-02 -1.04266904e-01 3.78077418e-01 8.34180772e-01 -2.00204805e-01 8.43679067e-03 -6.46038592e-01 -4.26563263e-01 7.55440414e-01 3.81627560e-01 -7.95499504e-01 1.43695474e-01 -9.12621617e-01 6.60027742e-01 8.14846694e-01 6.82104468e-01 -6.46935046e-01 -2.28760153e-01 5.54101616e-02 6.62110090e-01 5.98429978e-01 2.51363605e-01 -2.50135422e-01 4.15209770e-01 -9.40646887e-01 1.50359109e-01 9.88683045e-01 9.54232275e-01 7.49246180e-01 1.02485090e-01 -6.00357950e-01 8.15101147e-01 5.48242033e-01 2.17766538e-01 2.55523413e-01 -6.58632874e-01 2.28656188e-01 7.21200466e-01 -1.13653794e-01 -1.60206318e+00 -6.17158353e-01 -9.58483696e-01 -1.10299098e+00 -5.04888654e-01 5.53127289e-01 -3.81441593e-01 -4.13296998e-01 1.35106075e+00 2.48064548e-01 8.99632633e-01 -3.07606816e-01 1.02887225e+00 1.05568027e+00 8.48206818e-01 -1.79136187e-01 -4.36557323e-01 1.31621373e+00 -8.19539905e-01 -7.93741584e-01 -5.43447882e-02 2.35284165e-01 -6.90638781e-01 1.22757888e+00 4.46068645e-01 -9.09930944e-01 -9.64879990e-01 -8.12441409e-01 3.75547647e-01 -4.51307818e-02 2.10864410e-01 6.80285394e-01 6.14017308e-01 -8.37273538e-01 1.00829959e+00 -4.09424573e-01 -1.89160764e-01 2.79607177e-01 1.24669343e-01 9.71985701e-03 1.93310417e-02 -1.42717767e+00 5.77321760e-02 6.98265561e-04 4.71037701e-02 -4.89154533e-02 -1.17962742e+00 -7.28574872e-01 5.40395141e-01 4.92414683e-01 -7.60294318e-01 6.82582974e-01 -9.35694873e-01 -1.60330641e+00 -3.15312408e-02 -1.06030017e-01 -3.85960877e-01 1.64472267e-01 -6.96113929e-02 -9.64933157e-01 1.47312135e-01 -3.26768637e-01 -2.99583703e-01 1.06761479e+00 -9.62936342e-01 -6.37740493e-01 -2.22852424e-01 -9.67077687e-02 7.39004686e-02 -4.71984476e-01 -1.83559150e-01 -7.03142166e-01 -1.13056612e+00 3.93791467e-01 -5.90840816e-01 -3.98235470e-01 -4.42464530e-01 -1.88230321e-01 -1.10471867e-01 7.20096648e-01 -9.20291662e-01 1.78878188e+00 -2.56891394e+00 -1.46713391e-01 8.61460865e-01 3.85074884e-01 1.25862107e-01 -2.72933275e-01 5.86543679e-01 8.03902224e-02 -3.55460286e-01 1.51209876e-01 -7.62865394e-02 -5.91956563e-02 -1.45056680e-01 -1.37338191e-01 2.98179775e-01 -1.78620681e-01 8.62141430e-01 -1.29280877e+00 -7.43463635e-02 2.31412858e-01 6.21055663e-01 -8.76219809e-01 1.10929124e-01 -6.14938289e-02 6.03429496e-01 -4.87973720e-01 1.86885029e-01 7.19010353e-01 -3.96063983e-01 3.24930161e-01 -7.58975744e-01 7.01382458e-02 1.44330472e-01 -1.57945049e+00 1.55599129e+00 -2.65540421e-01 4.08805251e-01 4.02529329e-01 -1.06991339e+00 1.12295449e+00 1.56876847e-01 7.92863071e-01 -4.11495775e-01 2.30561569e-01 -2.07291067e-01 -1.24222897e-01 -2.99489111e-01 1.73233360e-01 2.40590453e-01 2.45442495e-01 1.63728490e-01 3.13994616e-01 3.84525396e-02 2.04427913e-01 3.52957398e-01 1.18391919e+00 -3.58405948e-01 2.22283900e-02 -1.82181805e-01 8.00593555e-01 -4.97276276e-01 4.55086797e-01 8.56984019e-01 2.65974868e-02 4.51393485e-01 3.64623636e-01 5.87352663e-02 -1.31877318e-01 -1.08016491e+00 2.98282355e-01 1.24449837e+00 2.95097142e-01 -8.43968213e-01 -5.32681286e-01 -6.36389673e-01 1.55311152e-01 5.31340539e-01 -2.51385003e-01 -3.53113592e-01 -3.50738198e-01 -6.07531905e-01 1.38587235e-02 1.87426671e-01 6.23606861e-01 -7.17202842e-01 3.83217514e-01 5.19115210e-01 -1.48644611e-01 -9.99531746e-01 -1.23873544e+00 -3.02720875e-01 -6.86110616e-01 -9.73038435e-01 -3.70984524e-01 -7.38199055e-01 6.23951137e-01 6.05481148e-01 8.55726302e-01 1.96428582e-01 4.14898284e-02 3.86533797e-01 -6.48443699e-01 -1.37759194e-01 -3.77502739e-02 -4.63322699e-01 -1.99548677e-01 8.69685411e-01 9.13120285e-02 -9.01696265e-01 -5.66275120e-01 2.76198179e-01 -4.50469911e-01 1.33287190e-02 4.52399433e-01 7.05030799e-01 5.45028090e-01 7.93539643e-01 6.76591933e-01 -1.15062571e+00 1.13049364e+00 -4.72616613e-01 -4.33692396e-01 1.94168836e-01 -3.92397344e-01 -2.96308488e-01 1.21407604e+00 -7.97415316e-01 -1.06817567e+00 -3.05049447e-03 -1.85821787e-01 -5.21727264e-01 2.46946886e-01 9.55541432e-01 -2.22213835e-01 -3.72566253e-01 8.36581767e-01 2.26041108e-01 -2.37742141e-02 -7.14296281e-01 6.56333685e-01 4.03046906e-01 6.02454960e-01 -1.36563510e-01 8.54225397e-01 2.21327156e-01 -9.14270729e-02 -1.10483515e+00 -8.43329966e-01 -8.46064270e-01 -2.89290071e-01 -4.99176592e-01 4.22629356e-01 -7.93394566e-01 -8.29976439e-01 2.24844933e-01 -6.63356185e-01 -1.14417545e-01 -4.17684257e-01 8.98090422e-01 -1.38143703e-01 6.67338669e-01 -6.76508069e-01 -7.56137311e-01 -4.75857556e-01 -5.96493244e-01 5.82145095e-01 3.49646866e-01 -2.40188628e-01 -9.68305290e-01 -4.61001635e-01 1.10833040e-02 5.06900012e-01 8.94281641e-02 5.67677319e-01 -6.28865123e-01 -3.33407938e-01 -4.12909061e-01 -3.10047448e-01 1.51577398e-01 2.63425380e-01 -9.59871411e-02 -5.87342799e-01 -2.74044096e-01 5.21437638e-02 6.66275561e-01 6.20570719e-01 7.55107343e-01 1.16958773e+00 -3.49974006e-01 -2.60172307e-01 6.84569359e-01 1.01410520e+00 -1.94740128e-02 4.14250642e-01 -5.35526752e-01 7.97387302e-01 4.82688189e-01 4.04746860e-01 5.61365128e-01 2.69715130e-01 5.40960550e-01 -1.32343113e-01 -1.71440721e-01 -3.78215700e-01 -4.92716104e-01 2.55583972e-01 9.87229705e-01 -1.95709005e-01 -2.10041955e-01 -3.31216365e-01 1.60863712e-01 -2.10720205e+00 -1.01940382e+00 -6.91341639e-01 2.17688179e+00 4.37305421e-01 -3.49067152e-03 3.76921654e-01 3.72434556e-01 7.19681084e-01 1.26252875e-01 -1.47652656e-01 1.15354963e-01 1.19847931e-01 4.37876701e-01 4.09395307e-01 5.82245886e-01 -1.05916893e+00 5.81157088e-01 5.83705854e+00 9.33886290e-01 -9.25672531e-01 3.40589397e-02 -2.92197645e-01 1.39168605e-01 -4.13176686e-01 -6.52948469e-02 -2.76190817e-01 6.59291625e-01 8.00267875e-01 -4.85662043e-01 7.76712418e-01 7.30450034e-01 5.30562460e-01 3.02052647e-01 -7.30258942e-01 1.07645917e+00 -2.18974933e-01 -1.21579766e+00 -1.45372376e-01 -2.38048449e-01 4.91704524e-01 -2.05313087e-01 -3.25429887e-01 3.36399972e-01 3.99357826e-01 -4.50641394e-01 4.69611734e-01 8.29893231e-01 1.98957294e-01 -6.28464520e-01 4.10770416e-01 4.05821472e-01 -1.84757650e+00 -3.26562196e-01 -1.66252002e-01 -8.97058323e-02 2.13075280e-01 1.26465499e+00 -4.83201861e-01 1.13270772e+00 6.95878267e-01 9.94444907e-01 -1.51308849e-01 1.37157595e+00 -2.97517061e-01 9.12082553e-01 -4.35849637e-01 -7.83099309e-02 -7.26345703e-02 -6.11826897e-01 7.51615465e-01 1.26374495e+00 3.18312973e-01 5.36676824e-01 5.54194450e-01 7.45044947e-01 -7.00390190e-02 3.98852259e-01 -1.66087329e-01 -1.50700524e-01 3.26224297e-01 1.18199742e+00 -6.78070188e-01 -4.29261625e-01 -7.56722629e-01 9.41180944e-01 -1.17524944e-01 6.60051882e-01 -5.93410194e-01 -6.24966919e-01 4.10016268e-01 2.34943911e-01 1.89074725e-01 -2.49224558e-01 -9.22955200e-02 -1.38541400e+00 -1.02072515e-01 -8.26883614e-01 7.16914415e-01 -4.17666823e-01 -1.55937719e+00 3.50576371e-01 -3.27785313e-01 -1.28049076e+00 2.71212459e-02 -1.61754750e-02 -9.49090242e-01 1.07212210e+00 -1.00084162e+00 -1.04343152e+00 -7.74502039e-01 1.00977683e+00 3.30912054e-01 -1.73226744e-02 5.54049909e-01 6.94969416e-01 -3.01775604e-01 5.78643680e-01 -1.94358781e-01 -5.06967790e-02 4.23896313e-01 -1.04775429e+00 3.96968395e-01 9.79554594e-01 3.39988977e-01 7.23547459e-01 5.48565865e-01 -8.31593335e-01 -1.62362659e+00 -1.35035372e+00 7.61215806e-01 1.65657431e-01 7.70213902e-01 -3.95753711e-01 -1.33363986e+00 3.78358960e-01 -2.78298408e-01 2.91272223e-01 6.55707002e-01 3.46864790e-01 -8.67661163e-02 -1.34718835e-01 -1.05544424e+00 5.11734307e-01 1.36825430e+00 -5.45306683e-01 -4.43530381e-01 3.58748704e-01 6.40700221e-01 -2.79502422e-01 -1.12284398e+00 2.10082397e-01 5.04963577e-01 -7.91666090e-01 1.16572583e+00 -1.82164282e-01 -4.60895330e-01 -5.64964950e-01 7.35078752e-02 -1.59387136e+00 -1.05187380e+00 -1.03988755e+00 -1.94441438e-01 1.22634697e+00 2.88638294e-01 -8.42782557e-01 7.33829260e-01 1.98737994e-01 -3.19110006e-01 -4.89163905e-01 -1.77367315e-01 -6.00437164e-01 -9.89221692e-01 -5.67965567e-01 6.35701537e-01 1.08781016e+00 4.93225098e-01 3.40482891e-01 -3.54343832e-01 4.25755769e-01 7.35325754e-01 4.06322688e-01 6.82743430e-01 -1.49963880e+00 -7.84352660e-01 -4.11585778e-01 -2.83357292e-01 -1.35708225e+00 -1.95538148e-01 -9.63513792e-01 -2.56717414e-01 -1.67539036e+00 -4.48229641e-01 -2.38865167e-01 -2.02013075e-01 6.12287074e-02 -3.49705756e-01 9.46258195e-03 2.53662884e-01 2.21051332e-02 -2.87656009e-01 3.41653913e-01 1.44770169e+00 -1.17032789e-01 -9.66432333e-01 4.61065769e-01 -8.03778231e-01 6.68854237e-01 3.29749078e-01 2.35893819e-02 -6.55451834e-01 1.93354860e-01 6.17903192e-03 2.34050155e-01 5.36178090e-02 -9.58835721e-01 4.74854022e-01 -7.84583539e-02 4.03798193e-01 -5.64344943e-01 8.45505074e-02 -8.36998105e-01 6.29155934e-01 4.61349875e-01 -7.44637847e-02 -8.26650500e-01 -1.61066741e-01 7.81398594e-01 -3.41601968e-01 2.71289676e-01 6.83476508e-01 9.34114903e-02 -5.54384351e-01 5.10998130e-01 -2.85147458e-01 -3.81721526e-01 6.07304394e-01 -1.44708440e-01 -4.12382768e-04 -6.79683864e-01 -1.20815527e+00 3.75436932e-01 -5.13341203e-02 3.15509140e-01 5.22046387e-01 -1.25542319e+00 -7.01015294e-01 6.31285369e-01 -2.92903781e-01 -4.77079719e-01 6.33514225e-01 1.03166175e+00 6.32554572e-03 -2.10652292e-01 2.25404993e-01 -3.25868458e-01 -1.21795511e+00 6.11190140e-01 1.43245995e-01 -2.14449868e-01 -1.01302958e+00 9.60143685e-01 8.06237683e-02 -1.70458421e-01 1.24221593e-01 -5.70564151e-01 -6.05453849e-01 3.06418926e-01 6.00545764e-01 4.94085103e-01 1.80960029e-01 -4.23003078e-01 -2.50476599e-01 6.63776875e-01 4.19530451e-01 4.60408777e-01 1.27590024e+00 -3.01304042e-01 -1.82367731e-02 7.51405358e-02 1.17243981e+00 4.64577973e-01 -7.57235467e-01 -4.62263167e-01 -2.78717309e-01 -7.14079261e-01 4.73165125e-01 -2.72863537e-01 -1.20422649e+00 3.13906103e-01 3.41855615e-01 9.28020298e-01 1.59551847e+00 -1.65244982e-01 8.62946570e-01 -6.12123497e-02 3.59647453e-01 -9.06373084e-01 -4.92234826e-02 4.28413719e-01 9.26445186e-01 -6.34254456e-01 5.35369478e-02 -1.22758937e+00 -3.18427712e-01 1.02955627e+00 8.38123728e-03 -4.76141572e-01 1.25261998e+00 6.34556785e-02 -3.35159183e-01 -2.06076026e-01 -3.94596130e-01 -5.47203422e-01 9.66377735e-01 6.05349422e-01 3.32862884e-01 8.70579109e-02 -4.58031178e-01 1.29162526e+00 -3.72892708e-01 1.18085518e-02 1.22593857e-01 3.36934626e-01 -4.95030522e-01 -9.16045725e-01 -4.32884008e-01 9.88471985e-01 -2.57154435e-01 -1.12349592e-01 -2.86425918e-01 7.89580122e-02 1.85120907e-02 1.43028414e+00 -7.33659863e-02 -8.58256519e-01 9.54308450e-01 -2.63974667e-01 4.08380091e-01 -6.05583370e-01 -9.23344672e-01 5.77271044e-01 1.50753677e-01 -7.93467700e-01 -1.13065854e-01 -3.83777708e-01 -1.27702355e+00 -3.85160655e-01 -5.75271547e-01 3.72264236e-01 1.81466475e-01 6.09689057e-01 5.34443021e-01 8.97178233e-01 8.93509448e-01 -9.55303371e-01 -6.85683042e-02 -8.42253983e-01 -1.02941787e+00 6.44812822e-01 7.69407004e-02 -4.01062608e-01 -5.29296219e-01 7.34615102e-02]
[10.143997192382812, 5.607990264892578]
9b23dfca-f79f-4ce6-9a2f-81a75cfde3c9
tcts-a-task-consistent-two-stage-framework
null
null
http://openaccess.thecvf.com/content_CVPR_2020/html/Wang_TCTS_A_Task-Consistent_Two-Stage_Framework_for_Person_Search_CVPR_2020_paper.html
http://openaccess.thecvf.com/content_CVPR_2020/papers/Wang_TCTS_A_Task-Consistent_Two-Stage_Framework_for_Person_Search_CVPR_2020_paper.pdf
TCTS: A Task-Consistent Two-Stage Framework for Person Search
The state of the art person search methods separate person search into detection and re-ID stages, but ignore the consistency between these two stages. The general person detector has no special attention on the query target; The re-ID model is trained on hand-drawn bounding boxes which are not available in person search. To address the consistency problem, we introduce a Task-Consist Two-Stage (TCTS) person search framework, includes an identity-guided query (IDGQ) detector and a Detection Results Adapted (DRA) re-ID model. In the detection stage, the IDGQ detector learns an auxiliary identity branch to compute query similarity scores for proposals. With consideration of the query similarity scores and foreground score, IDGQ produces query-like bounding boxes for the re-ID stage. In the re-ID stage, we predict identity labels of detected bounding boxes, and use these examples to construct a more practical mixed train set for the DRA model. Training on the mixed train set improves the robustness of the re-ID stage to inaccurate detection. We evaluate our method on two benchmark datasets, CUHK-SYSU and PRW. Our framework achieves 93.9% of mAP and 95.1% of rank1 accuracy on CUHK-SYSU, outperforming the previous state of the art methods.
[' Xilin Chen', ' Shiguang Shan', ' Hong Chang', ' Bingpeng Ma', 'Cheng Wang']
2020-06-01
null
null
null
cvpr-2020-6
['person-search']
['computer-vision']
[-2.51480550e-01 -2.10321337e-01 -1.03429325e-01 -2.47006327e-01 -9.86312926e-01 -5.41568696e-01 7.04873979e-01 -1.86541677e-01 -6.34155631e-01 4.33408260e-01 1.03225864e-01 1.67641118e-01 2.94091046e-01 -8.56450438e-01 -3.29038590e-01 -5.04066527e-01 3.02754849e-01 1.21608877e+00 9.19184566e-01 -3.60285304e-02 -5.45207551e-03 2.48259619e-01 -1.41843307e+00 2.62340069e-01 7.12953866e-01 9.51053381e-01 1.25599816e-01 6.43644869e-01 7.40783736e-02 2.93750435e-01 -8.16800237e-01 -6.84178233e-01 5.23622930e-01 -3.97827238e-01 -8.09276819e-01 -1.14240892e-01 6.23733044e-01 -6.56079471e-01 -5.24646044e-01 9.55000997e-01 9.79373872e-01 2.60227650e-01 6.10833406e-01 -1.17461181e+00 -6.74107075e-01 1.32934764e-01 -7.21968532e-01 3.74830157e-01 6.36911392e-01 1.71244621e-01 1.06246221e+00 -1.39131021e+00 7.36284971e-01 1.45119417e+00 6.11263394e-01 8.58674705e-01 -1.25760090e+00 -8.30703378e-01 2.48359427e-01 2.00740635e-01 -1.99216843e+00 -3.12570125e-01 1.73391774e-01 -4.98831242e-01 7.53878474e-01 3.09903800e-01 5.36375940e-01 9.30145144e-01 -4.23586488e-01 1.10122037e+00 7.53565490e-01 -3.01616609e-01 2.91702095e-02 2.77009189e-01 5.09965122e-01 8.21854472e-01 1.80186018e-01 3.68220747e-01 -4.80516613e-01 -4.49012488e-01 8.19071472e-01 -3.00016869e-02 -1.70547962e-01 -1.99348509e-01 -1.05642033e+00 7.71424592e-01 6.38823986e-01 2.42198072e-02 -2.58746505e-01 -8.32788944e-02 2.81288266e-01 -2.87740920e-02 4.08779502e-01 1.45974666e-01 -4.59169457e-03 4.24157977e-01 -1.07760763e+00 6.18601501e-01 4.92240697e-01 1.23997772e+00 6.96559548e-01 -5.72531641e-01 -1.02669597e+00 1.15406179e+00 2.07562804e-01 6.83897257e-01 3.10152650e-01 -4.10210341e-01 4.42980945e-01 6.97777748e-01 4.28154409e-01 -6.07639015e-01 -2.21298471e-01 -7.46533096e-01 -4.56033915e-01 -8.54217783e-02 6.11553848e-01 1.94311440e-01 -1.24212658e+00 1.51807880e+00 6.38644874e-01 1.03638358e-01 -1.91938519e-01 1.37114418e+00 1.13556540e+00 4.75767732e-01 2.41394773e-01 2.92002559e-01 1.92238915e+00 -1.37371004e+00 -3.22248310e-01 -3.06448519e-01 5.95722854e-01 -6.45155668e-01 7.33836830e-01 -2.95443553e-02 -1.09178162e+00 -9.24074054e-01 -8.81203413e-01 -1.48857966e-01 -3.69262189e-01 7.11663604e-01 8.38901550e-02 6.17151201e-01 -1.24585032e+00 6.69013429e-03 -2.29983851e-01 -7.28743970e-01 3.47420841e-01 4.39447880e-01 -2.50702292e-01 -1.27638116e-01 -1.31743836e+00 8.20949256e-01 4.35378432e-01 -1.08189605e-01 -8.53128672e-01 -3.59560728e-01 -5.69507003e-01 5.45471460e-02 5.57948768e-01 -8.93929541e-01 1.31395650e+00 -3.98974270e-01 -8.03790689e-01 1.50107396e+00 -4.26988900e-01 -2.17063367e-01 9.14162993e-01 -1.42581135e-01 -4.00129586e-01 6.67380542e-02 6.60844982e-01 7.62568712e-01 8.21026087e-01 -1.20335865e+00 -1.21124947e+00 -4.86786753e-01 -2.40587026e-01 2.04462484e-01 1.33113205e-01 4.25433636e-01 -1.38693523e+00 -6.27855837e-01 1.78801581e-01 -9.99536932e-01 5.76695502e-02 7.91719034e-02 -5.01400590e-01 -7.57505894e-01 6.27414048e-01 -6.74901903e-01 1.32516539e+00 -1.99758971e+00 -1.61836624e-01 4.12601233e-01 2.83803314e-01 4.03042078e-01 -2.05115482e-01 1.45520553e-01 2.24261150e-01 -1.81460515e-01 3.06914598e-01 -6.70112669e-01 2.63302084e-02 -1.34714484e-01 -2.34629631e-01 3.44200134e-01 1.60902426e-01 1.07267606e+00 -1.02326894e+00 -8.30886424e-01 -1.47118270e-01 3.14427838e-02 -3.94022495e-01 3.59794915e-01 1.04213797e-01 2.12486863e-01 -7.26509035e-01 9.98076618e-01 6.54681265e-01 -4.26825553e-01 -9.84738320e-02 -5.95197305e-02 -3.86290886e-02 9.16749910e-02 -1.30467439e+00 1.22686136e+00 1.71999618e-01 3.15767080e-01 -1.18029989e-01 -4.32823896e-01 9.82964158e-01 7.93973356e-02 9.53212157e-02 -7.46154130e-01 -1.59323111e-01 2.06232935e-01 -1.00202858e-01 -1.47548765e-01 7.23500431e-01 4.09727395e-01 -1.95184439e-01 5.46102107e-01 1.08497687e-01 6.48487985e-01 2.07784891e-01 4.37698305e-01 9.65147436e-01 1.95310310e-01 3.79186273e-01 -2.21426979e-01 8.22632432e-01 1.38042822e-01 7.69101858e-01 1.32561505e+00 -5.66330612e-01 8.77355397e-01 -2.40563676e-02 -6.90023541e-01 -1.00320435e+00 -1.16426075e+00 -1.52323872e-01 1.61223435e+00 5.25576711e-01 -3.80755067e-01 -8.37736428e-01 -1.00456154e+00 2.77140617e-01 2.81244069e-01 -7.03598559e-01 5.27832024e-02 -7.07389534e-01 -7.11555898e-01 6.25069022e-01 6.43225074e-01 7.20391512e-01 -9.12438154e-01 -1.08646609e-01 2.03204617e-01 -2.86257654e-01 -1.04186523e+00 -1.09927928e+00 -3.72617424e-01 -1.89073160e-01 -1.09223044e+00 -1.20426118e+00 -9.65429723e-01 7.47041821e-01 3.78913879e-01 1.15959537e+00 4.33285534e-01 -4.60224032e-01 3.87747079e-01 -3.23283106e-01 -2.66294539e-01 -1.56360388e-01 1.41090840e-01 1.92862704e-01 1.54129744e-01 7.93297768e-01 1.79188892e-01 -8.82063866e-01 9.06431854e-01 -1.70521781e-01 -3.30491886e-02 5.30847132e-01 9.72472250e-01 6.99583590e-01 -4.26242173e-01 4.30976212e-01 -5.88540554e-01 3.92315447e-01 -2.02050611e-01 -6.09526813e-01 6.14734769e-01 -6.15841627e-01 -1.66958481e-01 4.13375646e-02 -6.85037553e-01 -9.74034071e-01 2.63147891e-01 -8.44949335e-02 -3.05194587e-01 4.40112650e-02 -1.80184886e-01 -2.22840369e-01 -6.99479505e-02 7.90512204e-01 4.89989370e-01 -4.19095129e-01 -5.19712508e-01 6.38710931e-02 7.94615924e-01 9.44970489e-01 -6.05457485e-01 1.01131475e+00 5.04129827e-01 -5.53206861e-01 -4.18789119e-01 -9.48783994e-01 -1.20779836e+00 -7.49291539e-01 -3.71201098e-01 9.29306388e-01 -1.18379891e+00 -5.42324364e-01 4.11872387e-01 -1.20884931e+00 -5.75757250e-02 -1.08434923e-01 1.62464350e-01 -6.08697496e-02 1.63309500e-01 -6.10072732e-01 -1.06669807e+00 -6.19285226e-01 -1.08056474e+00 1.59587801e+00 5.79459488e-01 -2.81769186e-01 -4.96879876e-01 3.10010873e-02 5.16584218e-01 3.76625136e-02 -2.12252229e-01 3.90958697e-01 -1.19502163e+00 -8.81715238e-01 -6.77457809e-01 -6.34419978e-01 -2.66761690e-01 -3.24832320e-01 -6.82495475e-01 -1.17760551e+00 -4.38706815e-01 -5.28415263e-01 -3.09513867e-01 1.22205102e+00 7.28285238e-02 7.36499846e-01 2.15891656e-02 -9.50554550e-01 5.23146808e-01 1.06249702e+00 1.07300907e-01 5.28236151e-01 3.76783490e-01 4.39675659e-01 3.87752444e-01 9.25091684e-01 2.79813409e-01 5.42557955e-01 1.22944701e+00 -1.32309973e-01 -3.51884246e-01 -4.34673667e-01 -5.43217480e-01 6.65096790e-02 -2.21406356e-01 -2.98950464e-01 -2.27101043e-01 -9.55303192e-01 5.86994886e-01 -2.19755077e+00 -1.14878249e+00 -9.48770344e-02 2.38037753e+00 6.30853832e-01 6.70563951e-02 6.28881454e-01 -3.59353781e-01 1.09213746e+00 -8.35179463e-02 -4.69164610e-01 3.71609539e-01 2.96042711e-02 -7.89811984e-02 2.87726909e-01 4.03903216e-01 -1.45856571e+00 1.25867581e+00 5.77172375e+00 9.48285878e-01 -5.04630029e-01 4.42071557e-01 5.32805920e-01 -4.63134013e-02 3.11860412e-01 -1.59856662e-01 -1.70502079e+00 5.11336207e-01 1.93212673e-01 4.80298139e-02 -2.00401898e-03 1.08829093e+00 -5.69089875e-02 -2.21503481e-01 -1.27712703e+00 1.23782861e+00 1.58814207e-01 -1.03201437e+00 -3.35129015e-02 1.75491959e-01 5.38851380e-01 -2.16694772e-01 -1.14791118e-01 7.87514150e-01 3.05944949e-01 -6.72939360e-01 7.53987253e-01 5.12314796e-01 8.22488785e-01 -5.36294937e-01 9.31934834e-01 4.89859194e-01 -1.61583281e+00 -7.40537941e-02 -5.15980899e-01 3.30793053e-01 2.02455953e-01 4.28720266e-02 -8.53677630e-01 3.60217452e-01 9.63559747e-01 8.57183188e-02 -8.29101384e-01 1.22678614e+00 -2.21145123e-01 2.08692595e-01 -3.32992166e-01 -4.00737301e-03 9.15742218e-02 -5.75868860e-02 5.89144349e-01 1.34355724e+00 7.60899857e-02 3.17441344e-01 6.05488420e-01 1.07012057e+00 -1.08838387e-01 5.03824614e-02 8.45087599e-03 4.76603866e-01 7.35689163e-01 1.24979031e+00 -7.68017411e-01 -7.18115568e-01 -4.36787188e-01 1.43235099e+00 4.66346681e-01 4.92970347e-01 -8.83813143e-01 1.64707080e-02 6.57246292e-01 3.40592235e-01 2.20999941e-01 2.15250343e-01 1.03748210e-01 -1.07394803e+00 1.07199915e-01 -5.86637378e-01 9.28676605e-01 -5.72903275e-01 -1.42344904e+00 5.12901247e-01 1.15194783e-01 -1.05805123e+00 -2.72399396e-01 -3.06931853e-01 -7.23599255e-01 1.39244640e+00 -1.24586558e+00 -1.58065999e+00 -4.87749130e-01 4.55663413e-01 5.74986219e-01 -4.11889493e-01 5.85497797e-01 4.26327407e-01 -8.83703947e-01 1.09175146e+00 -2.80917704e-01 6.90867305e-01 9.36203599e-01 -1.17271411e+00 8.64314735e-01 8.88172686e-01 -8.37679654e-02 8.74102235e-01 3.90978903e-01 -9.69814360e-01 -7.95739233e-01 -1.18708551e+00 1.19966686e+00 -8.96631300e-01 1.12370886e-01 -6.61018670e-01 -8.38711679e-01 4.66989547e-01 -5.07983565e-01 1.59016520e-01 5.17626226e-01 2.77585089e-01 -4.53096479e-01 7.93681964e-02 -1.14293599e+00 5.39874792e-01 1.36952889e+00 -5.58009803e-01 -6.00118518e-01 4.24205631e-01 3.62813622e-01 -4.74378794e-01 -3.14805001e-01 1.78417712e-01 6.73752248e-01 -7.78752506e-01 1.52129340e+00 -4.16164577e-01 -2.90237099e-01 -6.07256472e-01 1.69249743e-01 -6.21341527e-01 -7.76240349e-01 -2.74392366e-01 -1.07178107e-01 1.32592034e+00 2.66411364e-01 -5.13100266e-01 9.79787171e-01 7.35518992e-01 2.00147375e-01 -7.14555800e-01 -8.57924879e-01 -1.03303456e+00 -3.74849439e-01 -2.94208247e-02 5.81313550e-01 5.06408334e-01 -3.26449186e-01 4.21702355e-01 -3.52948099e-01 5.10527551e-01 8.45502317e-01 1.70783401e-01 1.07551849e+00 -1.32509470e+00 -4.48238432e-01 -4.15806711e-01 -2.29017928e-01 -1.46515870e+00 -2.16299966e-01 -8.59255850e-01 1.30039603e-01 -1.47998250e+00 8.74209404e-01 -4.94462579e-01 -8.99161920e-02 4.97167081e-01 -7.52559066e-01 4.57606137e-01 3.42345566e-01 8.15824389e-01 -9.03741002e-01 2.80174464e-01 9.31631505e-01 -2.19673187e-01 -4.33305681e-01 3.37379813e-01 -3.55349004e-01 4.97862756e-01 2.78921425e-01 -4.34167176e-01 4.33518961e-02 -4.38903049e-02 -2.65397012e-01 -3.01324010e-01 6.90102994e-01 -1.01876247e+00 6.61993742e-01 2.57061929e-01 8.65894556e-01 -1.05430794e+00 4.22229588e-01 -2.61417419e-01 -4.94121946e-02 5.11713982e-01 -1.39045864e-01 -9.12870020e-02 -2.94759631e-01 5.84286511e-01 1.03191994e-01 -2.68171608e-01 8.01404655e-01 -2.08398581e-01 -1.02050328e+00 6.09292805e-01 1.28496394e-01 -5.35087623e-02 9.44881678e-01 -6.08946145e-01 -3.56110066e-01 -2.60655105e-01 -7.80587316e-01 7.49729991e-01 4.41083461e-01 3.88921738e-01 5.95473051e-01 -1.32490122e+00 -8.16446245e-01 1.84473082e-01 4.93523747e-01 -1.14900142e-01 1.58684060e-01 5.85138738e-01 -2.62963206e-01 5.36627114e-01 3.75422269e-01 -6.93505645e-01 -1.51441300e+00 6.84187174e-01 5.12505233e-01 -4.87484217e-01 -5.60237706e-01 1.17585111e+00 7.65364170e-01 -3.87853116e-01 5.07558167e-01 3.95641685e-01 -1.63055778e-01 7.09394887e-02 8.78369093e-01 4.09447610e-01 -3.19100410e-01 -9.35594022e-01 -6.76533461e-01 4.97252613e-01 -3.85134220e-01 -2.47497365e-01 4.86792177e-01 -1.01909697e-01 3.07895422e-01 -2.45485500e-01 7.14067578e-01 -1.60256997e-01 -1.00673378e+00 -6.34092391e-01 3.25657606e-01 -5.28349817e-01 -3.03561002e-01 -8.37246716e-01 -5.85926056e-01 5.07840812e-01 7.62123644e-01 -1.36792943e-01 8.73939455e-01 5.18664718e-01 8.99504185e-01 2.52949536e-01 6.44969702e-01 -1.32070637e+00 1.70925334e-01 5.00444770e-01 9.69248474e-01 -1.32865143e+00 1.14069372e-01 -6.10091269e-01 -4.92861837e-01 6.69142187e-01 1.00304067e+00 4.20796406e-03 1.83319286e-01 -1.61216348e-01 -2.94614956e-02 -8.91316086e-02 -3.38020414e-01 -8.47031236e-01 7.03728080e-01 7.24061847e-01 -6.44760206e-02 -6.49981434e-03 -2.93396801e-01 8.94857407e-01 -2.64388043e-02 -2.65123516e-01 -3.52102488e-01 6.57155514e-01 -5.61134398e-01 -1.10682118e+00 -8.29640090e-01 3.77928048e-01 -1.74001321e-01 -6.81923926e-02 -6.91075981e-01 9.61295187e-01 4.61069256e-01 7.50535727e-01 7.31194094e-02 -5.16896546e-01 4.49179232e-01 2.76716381e-01 2.85110354e-01 -7.78407931e-01 -8.36719394e-01 2.48810753e-01 2.09209830e-01 -5.77147663e-01 8.50407258e-02 -6.80640042e-01 -1.08724618e+00 -1.24454543e-01 -5.54074287e-01 1.48261830e-01 1.59107208e-01 8.59769344e-01 2.75262117e-01 3.45568918e-02 2.23837554e-01 -7.04353631e-01 -6.70542419e-01 -9.89958584e-01 -4.00769025e-01 5.25088310e-01 1.12833165e-01 -7.24065542e-01 5.65048903e-02 -1.91923261e-01]
[14.823991775512695, 0.8215827345848083]
2d6ccdd5-8ef4-4763-b368-35fcd80d8aa1
towards-unifying-feature-attribution-and
2011.04917
null
https://arxiv.org/abs/2011.04917v3
https://arxiv.org/pdf/2011.04917v3.pdf
Towards Unifying Feature Attribution and Counterfactual Explanations: Different Means to the Same End
Feature attributions and counterfactual explanations are popular approaches to explain a ML model. The former assigns an importance score to each input feature, while the latter provides input examples with minimal changes to alter the model's predictions. To unify these approaches, we provide an interpretation based on the actual causality framework and present two key results in terms of their use. First, we present a method to generate feature attribution explanations from a set of counterfactual examples. These feature attributions convey how important a feature is to changing the classification outcome of a model, especially on whether a subset of features is necessary and/or sufficient for that change, which attribution-based methods are unable to provide. Second, we show how counterfactual examples can be used to evaluate the goodness of an attribution-based explanation in terms of its necessity and sufficiency. As a result, we highlight the complementarity of these two approaches. Our evaluation on three benchmark datasets - Adult-Income, LendingClub, and German-Credit - confirms the complementarity. Feature attribution methods like LIME and SHAP and counterfactual explanation methods like Wachter et al. and DiCE often do not agree on feature importance rankings. In addition, by restricting the features that can be modified for generating counterfactual examples, we find that the top-k features from LIME or SHAP are often neither necessary nor sufficient explanations of a model's prediction. Finally, we present a case study of different explanation methods on a real-world hospital triage problem
['Ramaravind Kommiya Mothilal', 'Amit Sharma', 'Chenhao Tan', 'Divyat Mahajan']
2020-11-10
null
null
null
null
['counterfactual-explanation']
['miscellaneous']
[ 3.19044679e-01 6.98252738e-01 -5.82735360e-01 -5.76252043e-01 -2.67681897e-01 -4.70784396e-01 8.65643740e-01 3.56271625e-01 -1.97777245e-02 1.23602891e+00 6.29300416e-01 -5.48182726e-01 -5.99987447e-01 -7.56837964e-01 -6.69815242e-01 -4.75423366e-01 1.21975243e-02 3.95917445e-01 -3.24685752e-01 -1.79491758e-01 5.08592010e-01 3.02941233e-01 -1.72249126e+00 4.40497398e-01 8.22049797e-01 5.07833958e-01 -3.09563369e-01 2.53631353e-01 -7.68117085e-02 6.18954301e-01 -5.36612391e-01 -7.99969912e-01 2.50459731e-01 -6.84160113e-01 -8.43107641e-01 -9.44633111e-02 1.95587471e-01 -2.44415060e-01 -1.02297291e-01 7.11029172e-01 3.19921196e-01 -1.17375091e-01 1.20992386e+00 -1.84460211e+00 -9.07876194e-01 1.08124971e+00 -1.46492675e-01 7.99714774e-02 6.58120334e-01 2.80975103e-01 1.23546863e+00 -6.86977923e-01 6.11889482e-01 1.34537256e+00 6.88250363e-01 5.56496918e-01 -1.39220345e+00 -7.49565601e-01 4.14702982e-01 2.76597440e-01 -7.81606436e-01 -1.95586815e-01 7.84544230e-01 -5.67248404e-01 7.47618318e-01 7.33713865e-01 6.11173213e-01 8.87893975e-01 3.84691477e-01 4.25752580e-01 1.34286046e+00 -4.72224563e-01 3.16715568e-01 2.36575454e-01 1.60934627e-01 5.01585007e-01 6.89481914e-01 6.61822081e-01 -7.52035320e-01 -7.08589435e-01 5.33295572e-01 2.45644286e-01 -4.41095531e-01 -2.66913712e-01 -1.48722875e+00 1.18576920e+00 5.52357614e-01 8.03946331e-02 -5.17205179e-01 2.73438305e-01 1.89822033e-01 4.08282459e-01 3.45561832e-01 9.99522507e-01 -6.76076591e-01 3.23434681e-01 -7.66489208e-01 5.73027313e-01 7.25768268e-01 4.95254308e-01 7.31764257e-01 -2.32254505e-01 -4.70133752e-01 2.84251958e-01 2.71227360e-01 2.13217214e-01 7.02903092e-01 -1.09160936e+00 1.04770750e-01 6.96280122e-01 4.05058056e-01 -9.44106758e-01 -4.37351316e-01 -3.39007258e-01 -6.89296365e-01 3.44830424e-01 4.84775513e-01 -4.93057072e-02 -4.90150809e-01 1.96350658e+00 2.57257611e-01 1.48563564e-01 1.15260504e-01 8.82136941e-01 5.90507746e-01 -2.52888463e-02 8.95023569e-02 -5.79205155e-01 1.26637733e+00 -6.64655685e-01 -7.74969101e-01 1.99375587e-04 7.27042496e-01 -7.60256648e-01 1.01158357e+00 7.43478015e-02 -8.75917137e-01 -3.53610039e-01 -8.91269565e-01 5.44710219e-01 -2.86652297e-01 -3.21455598e-01 1.07214630e+00 6.86594367e-01 -7.37897098e-01 1.11896729e+00 -3.83243948e-01 -6.72122091e-02 2.66667932e-01 3.76208901e-01 -4.15690690e-01 1.31254807e-01 -1.43805492e+00 8.84089649e-01 2.91054785e-01 -3.68962049e-01 -4.64771748e-01 -1.16443825e+00 -7.01959848e-01 4.40884918e-01 1.44645184e-01 -1.21111262e+00 1.07560050e+00 -1.27065706e+00 -9.95200038e-01 5.02142727e-01 -1.11407660e-01 -6.12572074e-01 8.90425146e-01 2.22350374e-01 -4.04060066e-01 -1.58911809e-01 4.78227407e-01 6.00588322e-01 6.29355311e-01 -1.26682973e+00 -6.81791484e-01 -1.92284077e-01 8.79334584e-02 1.84835121e-01 -5.54114804e-02 -1.00661919e-01 4.42980349e-01 -8.22888255e-01 7.43514895e-02 -8.16878200e-01 -4.48939949e-01 1.96335278e-02 -8.32750499e-01 -9.32902917e-02 3.25315922e-01 -8.89600664e-02 1.11052668e+00 -1.79022062e+00 -3.37492019e-01 2.39256516e-01 3.10647011e-01 -4.43837583e-01 1.28456444e-01 3.08196753e-01 -6.74992561e-01 6.49305105e-01 -4.34236914e-01 -7.59203583e-02 2.02962637e-01 1.32065967e-01 -5.24822533e-01 3.99553448e-01 2.46371001e-01 7.62671351e-01 -8.97507191e-01 -3.22242975e-01 2.34987095e-01 9.90361348e-02 -8.44749629e-01 1.39295757e-01 2.23470420e-01 2.68366903e-01 -1.75550759e-01 2.85061300e-01 4.53305721e-01 -1.82752401e-01 2.74360567e-01 6.20545670e-02 -2.98215244e-02 5.85060477e-01 -1.31516206e+00 7.61084616e-01 -3.30766022e-01 4.00637805e-01 -7.72110522e-01 -6.98941171e-01 7.84548700e-01 4.00611550e-01 2.81923443e-01 -1.33479983e-01 -1.13958642e-01 4.38861519e-01 2.07957268e-01 -2.84223258e-01 2.64100730e-01 -6.77491426e-01 -1.29830435e-01 9.21937346e-01 -3.42636466e-01 -6.54368177e-02 -1.06258042e-01 2.33481511e-01 9.89747703e-01 -1.30872086e-01 1.08324015e+00 -3.72640163e-01 3.58598918e-01 3.32528092e-02 7.54878938e-01 1.01271009e+00 -5.91319464e-02 7.91486681e-01 8.37934494e-01 -7.35338807e-01 -8.52267027e-01 -9.94761825e-01 -3.05224836e-01 6.94813371e-01 -1.53342888e-01 -1.14694521e-01 -3.95212948e-01 -1.18202019e+00 5.73430598e-01 1.39409971e+00 -1.41047037e+00 -4.28171992e-01 -1.47812739e-01 -8.99273157e-01 3.08006763e-01 4.93766040e-01 -5.76084405e-02 -1.12360990e+00 -1.01789057e+00 1.24946296e-01 -1.71656966e-01 -2.80882239e-01 -5.41907489e-01 1.93052620e-01 -8.65709186e-01 -1.49732912e+00 -3.93962115e-01 1.00510195e-01 9.08798397e-01 5.38959429e-02 1.31442177e+00 4.82092202e-01 9.15587321e-02 1.58231463e-02 -2.99511492e-01 -5.75504303e-01 -6.50300980e-01 -3.07239354e-01 3.85135621e-01 -1.62275448e-01 2.41307572e-01 -4.80891198e-01 -6.82242870e-01 3.82003993e-01 -7.33581543e-01 1.40211374e-01 4.86032635e-01 1.16921031e+00 4.74421471e-01 -2.55262226e-01 8.29461634e-01 -1.40396857e+00 7.49466836e-01 -7.44967818e-01 -6.50556386e-02 2.33986944e-01 -1.41263258e+00 3.66754651e-01 6.83987379e-01 -4.16311353e-01 -1.01922822e+00 -3.60854864e-02 3.16577733e-01 -2.78132409e-01 -2.12301940e-01 5.20559251e-01 -1.17928065e-01 4.84868258e-01 9.23333645e-01 -2.39372104e-02 -6.42175600e-02 -4.41572905e-01 4.54445541e-01 3.26644719e-01 4.29012030e-01 -4.29502189e-01 7.29513526e-01 5.29350042e-01 6.04997166e-02 2.29217604e-01 -8.56275260e-01 9.40127112e-03 -6.41741037e-01 9.36832204e-02 4.39976752e-01 -3.35047543e-01 -7.04482853e-01 -2.50702739e-01 -1.03351784e+00 -1.04398355e-01 -7.20071375e-01 6.25393033e-01 -7.64744282e-01 6.70473278e-02 9.63437781e-02 -6.08479440e-01 9.82130878e-03 -1.18358850e+00 6.27727389e-01 -2.42043240e-03 -7.16714025e-01 -1.10597456e+00 3.21848206e-02 2.64525842e-02 2.32459426e-01 5.63693523e-01 1.21464443e+00 -9.95554030e-01 -1.52247787e-01 -1.08595632e-01 -4.14801091e-02 -1.84430048e-01 5.40049255e-01 1.46590531e-01 -9.77927327e-01 -5.85306995e-03 -1.41967207e-01 1.66417956e-01 9.09434080e-01 5.94450712e-01 1.07326186e+00 -8.88345778e-01 -4.90904361e-01 3.41190457e-01 1.16318023e+00 -7.07911849e-02 3.35903138e-01 3.78008336e-01 2.37782344e-01 9.20675218e-01 8.22612822e-01 5.05938590e-01 3.81402314e-01 8.06431711e-01 6.23433053e-01 -1.86117232e-01 -4.68339473e-02 -5.07107615e-01 1.67731747e-01 -3.01399231e-02 -1.23944350e-01 2.16385379e-01 -6.57784879e-01 4.40407991e-01 -1.96695244e+00 -1.21019626e+00 -4.41148937e-01 2.38381386e+00 7.56237090e-01 1.06383383e-01 1.54902384e-01 3.03988099e-01 6.36414528e-01 -2.04886436e-01 -6.12054765e-01 -6.95725203e-01 -1.84838936e-01 -3.00086558e-01 3.72445345e-01 4.99418169e-01 -7.03739047e-01 4.03516829e-01 6.82992029e+00 3.01242054e-01 -7.56721020e-01 1.81418527e-02 9.48172152e-01 -8.89137387e-02 -9.80533004e-01 2.19908193e-01 -3.15823913e-01 5.40031970e-01 8.56132686e-01 -9.16956067e-01 4.42287028e-02 9.36014533e-01 4.49459761e-01 4.80914004e-02 -1.72245514e+00 5.24739206e-01 -2.35670358e-01 -1.67466509e+00 4.29139793e-01 1.80352822e-01 7.87951410e-01 -5.86201668e-01 1.37826428e-01 5.81601039e-02 5.81210136e-01 -1.26216936e+00 9.85844016e-01 7.05516160e-01 7.02617407e-01 -7.25573659e-01 1.15859938e+00 1.45130187e-01 -6.01930618e-01 -1.43247649e-01 -2.44627312e-01 -5.62825143e-01 7.55413696e-02 6.85063601e-01 -1.15132415e+00 5.90757370e-01 4.08499181e-01 3.84442776e-01 -5.02265751e-01 8.24196815e-01 -6.84920430e-01 6.77284181e-01 7.56592378e-02 1.50072753e-01 -5.32656647e-02 2.95530915e-01 5.01321375e-01 1.07250822e+00 2.69754767e-01 2.25273579e-01 -1.44619688e-01 1.19386840e+00 4.65069376e-02 1.02878161e-01 -8.60350370e-01 5.97551703e-01 5.82487881e-01 8.68887484e-01 -5.88200867e-01 -5.22675037e-01 -2.66310096e-01 7.32245922e-01 2.38344353e-02 1.25999033e-01 -8.11977327e-01 3.44765037e-02 8.99363399e-01 1.84746444e-01 -1.91423222e-01 8.10595274e-01 -7.01321542e-01 -1.03819025e+00 -2.39315808e-01 -9.27721381e-01 7.07421482e-01 -7.69865870e-01 -1.47703755e+00 4.04691994e-01 2.03394666e-01 -1.34900689e+00 -5.91804326e-01 -2.97789544e-01 -9.07238901e-01 9.74430859e-01 -1.52142727e+00 -8.28571916e-01 -6.52878806e-02 3.97874326e-01 3.78911674e-01 -1.90619137e-02 9.35592353e-01 -2.83179134e-01 -3.24133426e-01 7.16759324e-01 -1.27241462e-01 -2.49214217e-01 7.89105296e-01 -1.55684745e+00 2.29345664e-01 5.40385246e-01 -2.49787066e-02 8.49946260e-01 1.17331755e+00 -7.23127782e-01 -5.60770988e-01 -1.04409993e+00 1.27407742e+00 -6.50642276e-01 3.86294186e-01 2.46473402e-01 -8.71365786e-01 7.58081555e-01 -9.21712816e-02 -1.30638331e-01 1.00578463e+00 3.86836439e-01 -4.10085112e-01 2.37272874e-01 -1.45100236e+00 6.95610940e-01 9.64962721e-01 6.74110278e-02 -1.07010591e+00 4.14638370e-01 7.94082046e-01 6.25088438e-02 -7.18182027e-01 2.65069604e-01 7.36480772e-01 -1.42894077e+00 9.02195573e-01 -1.23462534e+00 8.95733178e-01 -2.28309184e-01 -2.61939466e-01 -1.68357897e+00 -5.89611530e-01 -3.73986661e-01 2.71490544e-01 1.07912731e+00 8.10508907e-01 -1.03718340e+00 5.06099403e-01 1.07962620e+00 7.12163299e-02 -8.48876774e-01 -9.48023498e-01 -6.53246343e-01 1.99236259e-01 -3.68427843e-01 1.42485380e+00 1.46826935e+00 3.29193026e-01 -1.08017894e-02 -2.08160371e-01 2.21084149e-04 5.99254191e-01 5.01225889e-01 6.86491549e-01 -1.35985470e+00 -3.47141802e-01 -5.37752628e-01 -2.84851134e-01 4.98852096e-02 4.06265706e-01 -9.95176017e-01 -4.43558335e-01 -1.48012555e+00 6.42353237e-01 -4.05496031e-01 -3.73441011e-01 8.29181492e-01 -6.24615014e-01 -8.02790076e-02 3.73643905e-01 5.57832479e-01 9.28623080e-02 3.77688646e-01 8.29010487e-01 -1.11874659e-02 -1.65251881e-01 1.16991825e-01 -1.16282165e+00 1.01970720e+00 7.89726019e-01 -9.31319475e-01 -2.71265358e-01 2.00009257e-01 1.88533232e-01 1.62767366e-01 6.99255526e-01 -1.80434912e-01 -2.20746487e-01 -6.02567255e-01 4.26503092e-01 1.65113751e-02 -1.83410406e-01 -7.23145962e-01 5.20324171e-01 1.07274687e+00 -7.91582465e-01 2.11364523e-01 -1.33940637e-01 5.68680823e-01 1.68887284e-02 -3.46013337e-01 5.79041600e-01 -1.56547725e-01 -3.17442656e-01 -1.62287503e-01 -1.20750651e-01 -1.49845421e-01 7.71962106e-01 -2.98232883e-01 -3.99482816e-01 -5.80298722e-01 -7.19444692e-01 9.29894075e-02 6.17971063e-01 3.23242933e-01 5.89778125e-01 -1.52865875e+00 -1.04183590e+00 2.17741560e-02 2.78038561e-01 -7.53933609e-01 2.36726459e-02 8.58802557e-01 2.96122972e-02 5.08661330e-01 -3.12267214e-01 -8.19735527e-02 -1.11344337e+00 7.32925415e-01 4.11575168e-01 -4.61126029e-01 -3.95562589e-01 4.85490203e-01 6.95822597e-01 -5.75626075e-01 -4.12111789e-01 -3.55196804e-01 -2.66915113e-01 5.00285402e-02 4.27310407e-01 4.89623725e-01 -1.15087099e-01 -3.55753034e-01 -6.27549708e-01 1.33589404e-02 1.90793499e-01 -1.67297557e-01 1.40385759e+00 -2.71945596e-02 4.81689312e-02 4.52092677e-01 6.91460431e-01 -3.78262848e-02 -1.22319996e+00 1.73274174e-01 -3.99588495e-02 -8.04486096e-01 -1.98560059e-01 -1.10857558e+00 -9.18280184e-01 4.89496768e-01 3.00031662e-01 3.73567343e-01 9.63812649e-01 7.24561093e-03 -1.99976355e-01 6.37393072e-02 1.68193161e-01 -7.20919669e-01 -2.33283520e-01 -2.47747630e-01 1.36936331e+00 -1.37701225e+00 1.93959475e-01 -4.34032828e-01 -8.16709936e-01 1.07221913e+00 3.40047896e-01 5.69902696e-02 4.48691875e-01 -7.98746124e-02 -4.49652262e-02 -2.20373362e-01 -1.07538557e+00 1.28195763e-01 3.90885234e-01 6.10445678e-01 7.16290355e-01 5.67946613e-01 -8.05824041e-01 1.08237278e+00 -8.04195046e-01 -1.47874251e-01 8.73458743e-01 2.37954482e-01 -4.95499447e-02 -8.64394248e-01 -6.77702487e-01 1.01437044e+00 -4.28349406e-01 -1.45878419e-01 -7.16134131e-01 1.19647431e+00 1.47212625e-01 9.50387776e-01 4.69164401e-02 -1.49247825e-01 4.68688518e-01 1.71551928e-01 8.85930061e-02 -6.68277323e-01 -7.97630548e-01 -2.61799693e-01 2.12887481e-01 -6.43789828e-01 -3.47668767e-01 -1.05278325e+00 -1.33478725e+00 -4.31590647e-01 -4.93911654e-01 4.04212207e-01 3.06511074e-01 9.94613767e-01 2.83950210e-01 5.38603365e-01 8.15153956e-01 -5.46912611e-01 -7.82031775e-01 -8.04451585e-01 -5.34916639e-01 8.31892610e-01 2.57483065e-01 -1.04168725e+00 -8.68388772e-01 -5.24997618e-03]
[8.722393989562988, 5.619814395904541]
fbfd7a01-9eda-4107-a27e-93a45d847737
beyond-imitation-zero-shot-task-transfer-on
1812.02788
null
http://arxiv.org/abs/1812.02788v1
http://arxiv.org/pdf/1812.02788v1.pdf
Beyond imitation: Zero-shot task transfer on robots by learning concepts as cognitive programs
Humans can infer concepts from image pairs and apply those in the physical world in a completely different setting, enabling tasks like IKEA assembly from diagrams. If robots could represent and infer high-level concepts, it would significantly improve their ability to understand our intent and to transfer tasks between different environments. To that end, we introduce a computational framework that replicates aspects of human concept learning. Concepts are represented as programs on a novel computer architecture consisting of a visual perception system, working memory, and action controller. The instruction set of this "cognitive computer" has commands for parsing a visual scene, directing gaze and attention, imagining new objects, manipulating the contents of a visual working memory, and controlling arm movement. Inferring a concept corresponds to inducing a program that can transform the input to the output. Some concepts require the use of imagination and recursion. Previously learned concepts simplify the learning of subsequent more elaborate concepts, and create a hierarchy of abstractions. We demonstrate how a robot can use these abstractions to interpret novel concepts presented to it as schematic images, and then apply those concepts in dramatically different situations. By bringing cognitive science ideas on mental imagery, perceptual symbols, embodied cognition, and deictic mechanisms into the realm of machine learning, our work brings us closer to the goal of building robots that have interpretable representations and commonsense.
['Miguel Lázaro-Gredilla', 'J. Swaroop Guntupalli', 'Dianhuan Lin', 'Dileep George']
2018-12-06
null
null
null
null
['novel-concepts']
['reasoning']
[ 5.67939579e-01 5.12121856e-01 1.98055610e-01 -2.36348987e-01 3.93958688e-01 -8.06994617e-01 8.48822594e-01 2.74809837e-01 -1.84080720e-01 4.20054108e-01 1.93822011e-01 -6.12970948e-01 -6.94106391e-04 -1.03105354e+00 -7.53444374e-01 -3.26463342e-01 -4.93128411e-02 4.47823495e-01 1.69011876e-01 -4.88415182e-01 6.53202713e-01 5.54027140e-01 -1.79344153e+00 4.43812937e-01 6.64146483e-01 3.82599413e-01 8.11131477e-01 7.10710287e-01 -2.27129743e-01 1.39174855e+00 -5.06179214e-01 1.71320572e-01 -2.03369811e-01 -5.35301745e-01 -1.24670339e+00 2.84864843e-01 4.26033395e-04 -2.61822671e-01 -1.43602997e-01 7.64914274e-01 -5.51073670e-01 4.94680196e-01 7.50885367e-01 -1.26158249e+00 -1.41307950e+00 6.05492830e-01 -1.05104513e-01 -1.24169827e-01 8.36814404e-01 5.16315520e-01 5.54174542e-01 -4.72041577e-01 4.86886859e-01 1.62591839e+00 3.18481714e-01 7.02978730e-01 -1.32639909e+00 -2.86041558e-01 3.17113489e-01 2.93397605e-01 -9.83144999e-01 -1.36204153e-01 3.73804420e-01 -7.46364713e-01 1.10622025e+00 8.20555612e-02 1.07464814e+00 6.88498318e-01 2.75826514e-01 4.62474406e-01 1.00108540e+00 -8.29558551e-01 2.77364463e-01 1.55818135e-01 2.07163855e-01 1.01052046e+00 1.78707495e-01 4.21319515e-01 -5.39372265e-01 2.62953103e-01 1.08880758e+00 1.92558855e-01 -2.80072361e-01 -4.52480286e-01 -1.40835321e+00 6.42288327e-01 7.68886387e-01 5.77870607e-01 -3.37174058e-01 4.74745959e-01 1.12147100e-01 4.28528517e-01 -4.51182753e-01 1.06197464e+00 -1.09905213e-01 1.95945591e-01 -3.26326030e-04 -5.01353405e-02 7.79446006e-01 8.07059467e-01 8.56539488e-01 1.87165327e-02 3.11384678e-01 2.45640054e-01 2.84880638e-01 4.58167553e-01 6.63658202e-01 -1.61491811e+00 -1.84704542e-01 6.14191175e-01 7.88936913e-02 -9.66174841e-01 -4.34793681e-01 4.01540473e-02 -1.74638331e-01 5.90950966e-01 1.14544787e-01 8.47313851e-02 -8.83110702e-01 1.64183295e+00 -1.32573582e-02 -1.64045095e-01 2.89021969e-01 9.12085533e-01 3.52898359e-01 8.73003781e-01 3.85794699e-01 1.29037052e-01 1.37400925e+00 -7.29169667e-01 -3.23163718e-01 -6.52581215e-01 9.48141694e-01 -8.87320265e-02 1.43536222e+00 4.69516993e-01 -9.59510207e-01 -9.04016018e-01 -1.40978062e+00 -4.32338804e-01 -7.88982034e-01 -2.82561213e-01 1.11931634e+00 2.37565577e-01 -1.24901426e+00 5.35057604e-01 -6.76229358e-01 -7.42851555e-01 3.44733566e-01 1.76446393e-01 -3.30165058e-01 -2.53202394e-02 -7.58302748e-01 1.39583457e+00 9.43781912e-01 -1.71267465e-01 -9.89103734e-01 -3.11436683e-01 -1.12865973e+00 2.38849580e-01 2.79400617e-01 -1.17228305e+00 1.37587762e+00 -1.43478167e+00 -1.32853758e+00 1.05085790e+00 9.50776041e-02 -4.21843290e-01 -4.71285522e-01 -1.55278414e-01 -1.02769814e-01 3.22993129e-01 1.38066873e-01 1.06129229e+00 7.30279505e-01 -1.61928010e+00 -4.87206817e-01 -4.69122946e-01 7.31385052e-01 3.04456055e-01 3.44156809e-02 -3.04714262e-01 1.52076915e-01 -2.46933952e-01 4.48715895e-01 -9.51999307e-01 -6.76216781e-02 1.57734439e-01 7.86335543e-02 3.64878438e-02 6.06376827e-01 -3.30291927e-01 3.98827344e-01 -2.28839946e+00 4.92786676e-01 3.45463157e-02 3.72549653e-01 -1.48055613e-01 -6.32910952e-02 4.64765221e-01 -2.99825698e-01 9.11516771e-02 -5.34174144e-02 2.82402426e-01 8.12553316e-02 3.87861818e-01 -4.66931105e-01 -1.25204831e-01 1.36557043e-01 9.53114688e-01 -1.11208808e+00 -1.00983985e-01 4.84144866e-01 2.22063482e-01 -6.81738853e-01 2.41613463e-01 -4.03115869e-01 5.59059441e-01 -3.05956453e-01 2.65838113e-02 -3.98095623e-02 -2.64995158e-01 5.87518930e-01 2.06072792e-01 -9.89139155e-02 3.64625633e-01 -7.42625296e-01 1.78689814e+00 -8.98616254e-01 8.90273988e-01 -1.21267647e-01 -1.12625968e+00 7.29100168e-01 1.29775882e-01 -4.12595093e-01 -7.01680243e-01 2.16463685e-01 -3.54672074e-01 3.24955046e-01 -1.00095022e+00 3.01717639e-01 -5.71254432e-01 -1.53412893e-01 7.56813824e-01 -1.78039446e-02 -9.44304109e-01 1.64204948e-02 3.61723274e-01 7.99548447e-01 4.49360460e-01 5.37123859e-01 -2.16218367e-01 2.88276911e-01 3.93702775e-01 -2.15016991e-01 7.73208499e-01 1.43075332e-01 -2.92336475e-02 2.63472527e-01 -7.57067680e-01 -7.48139143e-01 -1.59192204e+00 4.48165387e-01 1.64418769e+00 2.10967556e-01 -4.92181256e-02 -6.61815345e-01 1.10663004e-01 -2.16727793e-01 1.43634427e+00 -7.07948089e-01 -5.40107369e-01 -5.26003361e-01 -4.54677492e-02 2.00196341e-01 5.59073329e-01 3.66190404e-01 -1.50813258e+00 -1.44759977e+00 9.66698583e-03 -5.99048994e-02 -7.78487802e-01 1.89164028e-01 3.57790798e-01 -1.08768749e+00 -1.07580745e+00 1.97190285e-01 -1.26882243e+00 9.74729836e-01 5.42410254e-01 1.12880075e+00 4.30335730e-01 -3.78046662e-01 9.15688217e-01 -3.34755093e-01 -6.61239684e-01 -7.52337337e-01 -5.14512420e-01 -1.36410877e-01 -5.53648591e-01 1.75690785e-01 -8.33142042e-01 -7.98620805e-02 -2.58421719e-01 -1.15623212e+00 5.95969260e-01 5.75703561e-01 5.83456159e-01 -9.87849236e-02 -1.04829028e-01 1.83971584e-01 -6.01127982e-01 6.93615198e-01 -4.99615788e-01 -2.03106329e-01 2.53192097e-01 -2.82184761e-02 4.50202972e-01 4.77934450e-01 -5.03710270e-01 -1.38563383e+00 3.70591171e-02 5.43555915e-01 8.93939584e-02 -4.13616359e-01 5.24171889e-01 -7.11783469e-02 2.40959510e-01 9.74241793e-01 4.30941492e-01 1.57674223e-01 -1.21449530e-02 9.95063961e-01 4.06058788e-01 9.10806775e-01 -9.66718614e-01 6.21987224e-01 5.09980023e-01 -9.29167718e-02 -8.14790189e-01 -5.96181631e-01 6.83820620e-02 -9.18886006e-01 2.25312859e-02 1.00656664e+00 -6.84875011e-01 -1.05582762e+00 6.39473945e-02 -1.20071316e+00 -6.51671350e-01 -6.06834114e-01 3.17851633e-01 -1.02757943e+00 1.28723517e-01 -4.34016585e-01 -4.93092924e-01 3.25434357e-01 -1.00591087e+00 7.09064543e-01 1.24039754e-01 -7.87895977e-01 -9.13616896e-01 -1.35759041e-01 1.63560614e-01 2.18619257e-01 6.00390472e-02 1.53106141e+00 -2.30059177e-01 -6.22395098e-01 2.20372245e-01 -1.33814618e-01 9.86577123e-02 2.01889783e-01 -1.98006600e-01 -7.37904966e-01 9.30572301e-02 3.43417376e-01 -6.30217612e-01 5.48754811e-01 -1.29056260e-01 1.27513647e+00 -2.42116794e-01 -6.60272717e-01 1.89920396e-01 1.24718714e+00 9.08861220e-01 8.93080592e-01 2.98365474e-01 3.84248316e-01 9.47966278e-01 3.05406362e-01 -1.80122834e-02 6.16490126e-01 3.00816417e-01 1.13663003e-01 2.45790645e-01 3.52120474e-02 -1.93833530e-01 3.12352508e-01 4.70111072e-01 -1.41810954e-01 1.53319806e-01 -1.09068680e+00 3.93865585e-01 -1.63139451e+00 -1.19952440e+00 2.52499521e-01 1.74537492e+00 8.57453167e-01 1.26375020e-01 -3.21407050e-01 -1.23984087e-02 4.55053538e-01 -3.79874885e-01 -4.54277486e-01 -9.33971107e-01 3.99138063e-01 3.07187706e-01 -3.69903505e-01 5.94931841e-01 -6.05046868e-01 1.12048542e+00 6.83466148e+00 -8.98169875e-02 -9.49289799e-01 -2.32444033e-02 2.64705002e-01 2.76481867e-01 -1.20187446e-01 2.47868583e-01 8.76339376e-02 -9.78017524e-02 8.49778116e-01 -4.87755649e-02 8.86487186e-01 5.98562479e-01 -1.63951293e-02 -5.67093492e-01 -1.72172093e+00 9.20478940e-01 3.44578803e-01 -1.23412991e+00 4.86191243e-01 -2.42194042e-01 2.52127737e-01 -5.07490456e-01 4.00346257e-02 5.06606102e-01 7.00365543e-01 -1.44860363e+00 8.41165662e-01 8.20866346e-01 3.62760961e-01 -2.57625788e-01 -5.05900495e-02 5.45612156e-01 -7.37549305e-01 -4.76015121e-01 -2.36925066e-01 -1.01080346e+00 -1.51323378e-01 -2.47103319e-01 -8.03463578e-01 -1.25711232e-01 3.66044283e-01 3.67402494e-01 -4.08669323e-01 4.75923061e-01 -4.23456639e-01 -1.05658814e-01 2.14127954e-02 -1.93195283e-01 -6.40870407e-02 -9.34653878e-02 1.54898584e-01 8.51508439e-01 1.20010555e-01 7.59311080e-01 -3.79991974e-03 1.25183725e+00 2.99504161e-01 -3.70404929e-01 -9.41567540e-01 -4.31567654e-02 5.12488663e-01 7.91310489e-01 -8.19662392e-01 -7.02782989e-01 -3.71049166e-01 1.05648780e+00 3.50651443e-01 3.95596683e-01 -6.95396960e-01 -4.54926163e-01 5.18393159e-01 2.29920838e-02 9.31902006e-02 -6.40882432e-01 -3.33376318e-01 -9.31104362e-01 -4.37427521e-01 -8.50682616e-01 -8.65017027e-02 -1.71363199e+00 -6.74073875e-01 2.36354038e-01 2.74818718e-01 -7.36762702e-01 -3.91421199e-01 -1.05099237e+00 -6.61869109e-01 5.43754339e-01 -6.71073794e-01 -1.01356447e+00 -5.45573294e-01 3.97761703e-01 5.16082525e-01 6.19215854e-02 1.14489591e+00 -7.01305687e-01 2.56135821e-01 -2.56386936e-01 -4.61663872e-01 2.66414192e-02 1.22479543e-01 -9.84135687e-01 3.88430864e-01 3.16518575e-01 3.03435892e-01 1.03438497e+00 6.36240363e-01 -3.23885381e-01 -1.39083195e+00 -4.56306219e-01 3.35448772e-01 -8.70803416e-01 6.42025650e-01 -3.99288237e-01 -9.26850915e-01 1.28455603e+00 4.25269455e-01 -5.28279126e-01 7.19323397e-01 7.51052946e-02 -5.93743503e-01 2.01170161e-01 -9.18475270e-01 1.06402087e+00 1.22127450e+00 -8.49584818e-01 -1.62702894e+00 2.89179593e-01 9.24154520e-01 -3.05748638e-02 -4.10015881e-01 -1.96361259e-01 7.48976886e-01 -9.43094909e-01 1.09521329e+00 -8.54216397e-01 4.20979768e-01 -5.91547787e-01 -1.62256882e-01 -1.48013449e+00 -6.57106698e-01 -1.39294788e-01 2.77511775e-01 5.89843869e-01 2.39272118e-01 -7.56586075e-01 7.55341500e-02 6.43017650e-01 -3.56507629e-01 -7.14268163e-03 -2.43458360e-01 -4.84760135e-01 1.90379441e-01 -5.16455710e-01 5.64040244e-01 9.32058334e-01 8.85152936e-01 6.60276115e-01 5.40401399e-01 1.29169717e-01 1.98277518e-01 2.93465674e-01 7.15380073e-01 -1.28962696e+00 -5.73194921e-01 -3.38260978e-01 -4.99720395e-01 -1.01257885e+00 3.25450927e-01 -9.33346450e-01 3.04999113e-01 -1.59443080e+00 2.68880397e-01 -2.74007857e-01 1.90102592e-01 7.38773942e-01 1.99473083e-01 -7.80907348e-02 6.87244475e-01 4.04835045e-01 -4.50070322e-01 1.83336735e-01 1.24288750e+00 -2.70564675e-01 -2.85819650e-01 -7.97259033e-01 -9.98637974e-01 1.22519648e+00 8.76494467e-01 -2.28522733e-01 -7.27866054e-01 -7.21422493e-01 2.99371719e-01 5.52932033e-03 8.95457149e-01 -1.37853646e+00 2.64451504e-01 -4.80530888e-01 7.93317437e-01 7.39585906e-02 3.98660302e-01 -9.23450351e-01 1.30775973e-01 8.81111562e-01 -5.39032876e-01 1.14377573e-01 4.99186575e-01 3.99958402e-01 1.29802793e-01 -4.24945295e-01 5.37848890e-01 -8.26243818e-01 -1.45236576e+00 -8.64302218e-01 -1.07195246e+00 -9.75310728e-02 1.18834281e+00 -5.23935974e-01 -4.57709849e-01 -4.59948838e-01 -1.23642945e+00 2.81109344e-02 6.93039238e-01 5.26996732e-01 7.98513055e-01 -8.39802921e-01 -9.35597159e-03 2.93505698e-01 2.08921328e-01 -1.93776324e-01 -1.32060543e-01 1.55371189e-01 -8.01862538e-01 4.10899103e-01 -7.23970592e-01 -3.92093927e-01 -7.31538534e-01 9.33303237e-01 4.74376380e-01 5.37837565e-01 -4.61331248e-01 7.11400867e-01 1.02845931e+00 -5.48261821e-01 -1.85886815e-01 -7.83758163e-01 -2.30539173e-01 -3.71135861e-01 6.68938994e-01 -6.45321282e-03 -5.91082454e-01 -3.75307411e-01 -9.54401493e-02 6.07756317e-01 1.79770425e-01 -2.25049868e-01 9.36509252e-01 -3.18769306e-01 -4.86935675e-01 7.42381394e-01 7.20938027e-01 -4.51359510e-01 -1.01028192e+00 1.61133155e-01 -7.16099143e-02 -1.12290911e-01 -4.29653198e-01 -9.20024872e-01 -2.71967381e-01 1.16132903e+00 3.61263007e-01 1.85707912e-01 1.14837253e+00 4.43201452e-01 2.43001834e-01 1.09909105e+00 6.56997919e-01 -8.81636560e-01 8.95628273e-01 7.13922024e-01 1.14436793e+00 -8.78263295e-01 7.30204582e-02 -5.18320203e-01 -6.51382565e-01 1.35915804e+00 7.53814518e-01 -1.67765453e-01 2.46581003e-01 1.65594056e-01 -2.29272366e-01 -5.27305186e-01 -7.57200778e-01 -1.54412299e-01 -1.72992483e-01 1.15885258e+00 2.09727332e-01 1.20949350e-01 3.12112570e-01 2.63760328e-01 -4.31814700e-01 1.87499985e-01 8.61234188e-01 1.21597385e+00 -9.75697577e-01 -4.71662760e-01 -5.42139471e-01 7.96304345e-02 3.58148724e-01 -6.81026056e-02 -4.25668031e-01 9.85153794e-01 5.20252764e-01 8.72727513e-01 6.00871980e-01 -2.33442098e-01 1.41715199e-01 3.61304522e-01 1.00918305e+00 -1.16365802e+00 -1.95878908e-01 -7.97821820e-01 -1.53537691e-01 -5.54067254e-01 -6.05687022e-01 -4.55359906e-01 -2.04084563e+00 -2.55822808e-01 2.84658045e-01 -8.68629292e-03 5.01661599e-01 1.15608490e+00 2.13557228e-01 6.32132232e-01 -5.86046651e-02 -1.02641857e+00 -1.06475532e-01 -7.37436235e-01 -1.56631291e-01 4.50774223e-01 2.42343277e-01 -6.85116887e-01 -1.09631345e-01 6.85307682e-01]
[4.49679708480835, 1.1443039178848267]
fb881052-c498-4d31-a3bf-cd0a87c6b422
text-in-context-token-level-error-detection
null
null
https://aclanthology.org/2021.inlg-1.25
https://aclanthology.org/2021.inlg-1.25.pdf
Text-in-Context: Token-Level Error Detection for Table-to-Text Generation
We present our Charles-UPF submission for the Shared Task on Evaluating Accuracy in Generated Texts at INLG 2021. Our system can detect the errors automatically using a combination of a rule-based natural language generation (NLG) system and pretrained language models (LMs). We first utilize a rule-based NLG system to generate sentences with facts that can be derived from the input. For each sentence we evaluate, we select a subset of facts which are relevant by measuring semantic similarity to the sentence in question. Finally, we finetune a pretrained language model on annotated data along with the relevant facts for fine-grained error detection. On the test set, we achieve 69% recall and 75% precision with a model trained on a mixture of human-annotated and synthetic data.
['Ondřej Dušek', 'Simon Mille', 'Zdeněk Kasner']
null
null
null
null
inlg-acl-2021-8
['table-to-text-generation']
['natural-language-processing']
[ 3.12496006e-01 6.67377710e-01 7.84232393e-02 -5.10635078e-01 -1.46684122e+00 -5.44446290e-01 6.80225194e-01 4.22363937e-01 -4.88990963e-01 1.30109882e+00 5.34566522e-01 -3.23402971e-01 4.11334485e-01 -1.01272023e+00 -1.07497561e+00 3.54584932e-01 4.37094361e-01 7.78765917e-01 2.69660681e-01 -5.50568402e-01 2.39219502e-01 -1.98833838e-01 -1.34072900e+00 1.11356270e+00 1.40460467e+00 9.06083822e-01 1.04935683e-01 9.57307816e-01 -2.22709537e-01 9.47300673e-01 -1.16054535e+00 -8.77866685e-01 -1.10228553e-01 -5.21535158e-01 -1.24244535e+00 -3.04632187e-01 4.18989867e-01 3.68102826e-03 1.22771725e-01 9.44288254e-01 4.03795362e-01 1.36050567e-01 7.35381305e-01 -8.14010501e-01 -6.91742420e-01 9.22967970e-01 2.12636128e-01 3.19981098e-01 9.44583535e-01 2.99252063e-01 9.94100392e-01 -1.21166635e+00 9.71768439e-01 1.31797874e+00 5.84122241e-01 9.72788572e-01 -1.08195448e+00 -2.54080355e-01 -1.04925312e-01 1.09525518e-02 -1.09265006e+00 -6.02978826e-01 2.75190413e-01 -2.09510103e-01 1.73674214e+00 1.40047207e-01 2.33348936e-01 1.28632843e+00 3.71899724e-01 5.68169296e-01 6.98403060e-01 -8.90420020e-01 3.33012432e-01 1.79681391e-01 -6.74979463e-02 7.67192960e-01 3.15627664e-01 -4.72493395e-02 -6.31493926e-01 -1.74569190e-01 -2.82192752e-02 -8.81980121e-01 -1.66530371e-01 5.53547025e-01 -1.08957303e+00 7.45688617e-01 8.72502476e-02 2.11552277e-01 -3.81978899e-01 1.05561735e-03 4.19201314e-01 3.99350435e-01 8.09119821e-01 1.21221995e+00 -6.84511721e-01 -2.16799632e-01 -1.18488228e+00 5.68822682e-01 1.01452339e+00 1.14277577e+00 4.65484709e-01 -6.89393878e-02 -9.35739219e-01 8.65473628e-01 2.63328433e-01 7.03509808e-01 8.69853437e-01 -7.05306888e-01 1.14696455e+00 7.02367485e-01 4.50141966e-01 -7.71152616e-01 -4.93530296e-02 -2.63078988e-01 -3.64389360e-01 -1.65260479e-01 2.65828580e-01 -5.16847789e-01 -7.69003808e-01 1.57200253e+00 -4.75971662e-02 -1.09988824e-01 5.86939514e-01 4.41781312e-01 8.86066794e-01 7.38151670e-01 2.56797701e-01 -4.23237085e-02 1.14588547e+00 -9.87075269e-01 -4.02363002e-01 -4.66415644e-01 9.43839729e-01 -7.75426030e-01 1.24221778e+00 3.00160289e-01 -1.16900909e+00 -7.03226447e-01 -1.14004540e+00 -1.65453106e-02 -3.64590049e-01 3.34682018e-01 -7.59152919e-02 3.06378037e-01 -1.13459909e+00 5.88878095e-01 -3.60371947e-01 -2.15926141e-01 1.39045104e-01 -1.37641504e-01 -1.41412988e-01 -3.43139432e-02 -1.75588632e+00 1.04698288e+00 6.96044266e-01 -3.95713180e-01 -9.71819282e-01 -7.69360781e-01 -1.10389161e+00 -3.12679186e-02 2.34574988e-01 -8.78140390e-01 1.85774624e+00 -7.28760481e-01 -1.20678008e+00 8.70575309e-01 -3.72600794e-01 -9.95848477e-01 5.26752412e-01 -2.93462574e-01 -7.89496601e-01 1.92754552e-01 5.86307645e-01 6.01798236e-01 3.99740189e-01 -9.41411138e-01 -7.78351009e-01 3.54801863e-01 -3.99479754e-02 2.74022296e-02 1.52385868e-02 1.67920381e-01 -2.13377066e-02 -6.78106606e-01 -3.94590467e-01 -5.12447238e-01 -1.82577834e-01 -6.77684247e-01 -8.98302078e-01 -3.78779739e-01 2.57468045e-01 -8.87322783e-01 1.38491976e+00 -1.63049805e+00 -4.60930705e-01 4.55437154e-02 -7.87728429e-02 5.01156449e-01 -4.34951872e-01 5.09388268e-01 -3.49368975e-02 4.28344905e-01 -7.03552961e-02 -4.34736192e-01 8.12530518e-02 -2.81052947e-01 -7.32921958e-01 -7.23387539e-01 7.54128933e-01 1.06895757e+00 -1.48310983e+00 -5.18644512e-01 -2.02735737e-01 -7.37179145e-02 -6.54486716e-01 3.13776076e-01 -9.18286562e-01 3.06391511e-02 -5.11491656e-01 1.49732560e-01 8.57594833e-02 -2.61762440e-01 -3.36184315e-02 7.23490119e-02 4.77624685e-01 1.11351633e+00 -8.29082906e-01 1.47062826e+00 -8.23842466e-01 3.11958313e-01 -8.24442565e-01 -4.28459108e-01 1.06045496e+00 4.79364693e-01 -3.15247327e-01 -6.28647745e-01 -2.58797407e-01 2.71688342e-01 -2.46397987e-01 -6.91821933e-01 8.34900379e-01 -2.19516948e-01 -3.61539960e-01 7.20402539e-01 4.80038732e-01 -5.41499257e-01 7.61725843e-01 5.56264997e-01 1.54221070e+00 -5.10293357e-02 2.91811377e-01 -7.93929473e-02 6.01027548e-01 3.43838006e-01 3.69253099e-01 9.77198124e-01 1.57962486e-01 6.61752284e-01 2.97761798e-01 -3.70562434e-01 -1.00625312e+00 -8.87563646e-01 3.00153196e-01 6.95963919e-01 -4.24636930e-01 -8.47267807e-01 -8.56976807e-01 -1.31709623e+00 -1.63008139e-01 1.68499148e+00 -4.40684050e-01 -3.70339632e-01 -3.02383125e-01 -4.76572335e-01 8.59514773e-01 4.10529464e-01 4.34826463e-01 -1.48172355e+00 -3.46824259e-01 4.33168292e-01 -7.76619732e-01 -1.31012702e+00 -5.23813486e-01 -3.62092435e-01 -4.23863113e-01 -1.19087505e+00 -3.97606157e-02 -6.45907581e-01 7.35484540e-01 -4.33405876e-01 1.79785800e+00 1.02788582e-01 1.51234150e-01 -1.25066787e-01 -6.21308386e-01 -5.73203087e-01 -1.38611722e+00 2.24348381e-01 -8.16000476e-02 -2.48201013e-01 5.37975729e-01 1.00486599e-01 -1.57720864e-01 -7.06522614e-02 -6.79907858e-01 2.61017472e-01 3.26026261e-01 8.26793909e-01 4.84945923e-01 7.21587017e-02 9.25954580e-01 -1.26988506e+00 1.22715139e+00 -3.37820143e-01 -3.61943573e-01 7.36321151e-01 -6.09973788e-01 3.67562383e-01 9.88099635e-01 9.16677713e-03 -1.38870513e+00 -3.02214921e-01 -2.88169533e-01 1.15698382e-01 -1.39724419e-01 6.65028453e-01 -2.21901178e-01 5.43833435e-01 1.30182040e+00 1.68369681e-01 -5.68214774e-01 -1.00637108e-01 5.02429843e-01 8.17909718e-01 5.52057743e-01 -9.31044400e-01 5.24347246e-01 -3.66714090e-01 -4.86882657e-01 -1.92725286e-01 -1.55377197e+00 2.55875051e-01 -2.69332319e-01 -7.36723095e-02 7.28983879e-01 -8.57524991e-01 5.53662442e-02 1.41770899e-01 -1.59041190e+00 -5.19050121e-01 -7.12704718e-01 2.06046641e-01 -5.71520567e-01 6.06363593e-03 -6.06110394e-01 -7.00660169e-01 -8.21771026e-01 -7.00019360e-01 1.28800833e+00 1.47700489e-01 -7.33139336e-01 -9.23160315e-01 2.45448425e-01 4.24104482e-01 2.28195921e-01 5.13424352e-02 1.00737107e+00 -9.98546064e-01 -5.45630753e-02 -4.70647097e-01 3.73228230e-02 6.26031041e-01 4.41994593e-02 1.26798958e-01 -6.93051279e-01 -3.03516490e-03 -3.19233716e-01 -7.58445084e-01 6.78102732e-01 -1.07827589e-01 9.11260843e-01 -5.67254245e-01 -2.92073548e-01 -5.42812757e-02 1.09107196e+00 5.79685066e-03 5.00377119e-01 1.20284021e-01 2.81324118e-01 4.96746302e-01 7.69131064e-01 2.50275284e-01 4.69238192e-01 4.79258418e-01 -3.26120168e-01 3.67434740e-01 -2.90938377e-01 -1.06923914e+00 3.42230290e-01 5.27191699e-01 3.25508386e-01 -6.46808445e-01 -1.06463778e+00 6.37146771e-01 -1.60714662e+00 -1.03470039e+00 5.98639399e-02 2.07870698e+00 1.45748460e+00 7.08610058e-01 -1.34535939e-01 -7.10032182e-03 7.17274129e-01 -3.13414246e-01 -1.51905432e-01 -5.94242632e-01 -2.52909154e-01 4.98423159e-01 -9.97890346e-03 8.93012226e-01 -8.63297045e-01 1.21102500e+00 6.69264507e+00 9.05336857e-01 -7.62215197e-01 8.75462592e-02 9.82921958e-01 -9.85225737e-02 -7.11664081e-01 -8.27076137e-02 -1.13327479e+00 7.57801473e-01 1.55676651e+00 -7.43836105e-01 1.97466865e-01 6.55804276e-01 4.37893808e-01 1.40347732e-02 -1.06366324e+00 3.01117480e-01 2.93877810e-01 -1.51949346e+00 5.05719662e-01 -5.46918392e-01 1.13157654e+00 -2.96819918e-02 -4.83361602e-01 6.98329866e-01 8.44898582e-01 -1.02724600e+00 8.49665880e-01 8.33572745e-01 1.03268015e+00 -5.70743322e-01 1.04466093e+00 6.08225882e-01 -6.01786017e-01 2.25025579e-01 -3.00946981e-01 -8.73742253e-02 2.62404561e-01 1.08216918e+00 -1.53845191e+00 6.36294067e-01 2.44955838e-01 3.70932072e-01 -8.78079653e-01 7.02823400e-01 -1.01919842e+00 8.47861111e-01 -8.15689787e-02 -3.62502068e-01 -8.54194760e-02 5.22781730e-01 4.04571444e-01 1.41057110e+00 3.82612795e-01 -5.04485611e-03 -4.42418009e-02 1.10666585e+00 -5.44289529e-01 5.24778701e-02 -5.65818131e-01 -1.59306660e-01 7.81710148e-01 1.05652118e+00 -8.95617977e-02 -9.28114712e-01 -2.45926887e-01 9.59531426e-01 6.62957907e-01 2.31950358e-01 -4.12698150e-01 -6.46194637e-01 2.65359674e-02 1.09358490e-01 -7.45545551e-02 3.05300027e-01 -2.54059911e-01 -1.34347427e+00 1.57729536e-01 -1.09218574e+00 3.79508913e-01 -1.20174849e+00 -1.40627515e+00 1.03631866e+00 -2.33053342e-01 -1.08704758e+00 -9.55902874e-01 -3.50627542e-01 -7.62438953e-01 1.26052022e+00 -1.19384253e+00 -7.09868371e-01 5.04039694e-03 1.38340682e-01 7.85098255e-01 -3.32317740e-01 9.91298914e-01 6.91699388e-04 -2.75036812e-01 7.91804373e-01 -5.12836933e-01 2.80239016e-01 7.67799556e-01 -1.46205175e+00 1.18927622e+00 1.15946758e+00 9.96406004e-02 5.92330813e-01 7.97126293e-01 -1.14278126e+00 -5.71425378e-01 -1.78567183e+00 1.93355274e+00 -8.80343378e-01 5.79692483e-01 -2.26971537e-01 -7.21034348e-01 6.47269428e-01 1.14719450e-01 -9.87120122e-02 5.55329263e-01 1.09613677e-04 -2.71130115e-01 2.28299528e-01 -1.35409737e+00 3.48180592e-01 9.56715882e-01 -6.16121769e-01 -1.11765969e+00 6.67327762e-01 1.01179397e+00 -6.66994750e-01 -5.66842675e-01 3.32997084e-01 -9.83252302e-02 -4.37914997e-01 4.66118813e-01 -1.15236199e+00 1.17918324e+00 -3.74614149e-01 -8.05760399e-02 -1.94467950e+00 1.47327306e-02 -5.09206414e-01 -9.29471031e-02 1.34392107e+00 1.25306356e+00 -3.76137525e-01 4.67861027e-01 7.09048212e-01 -3.12091202e-01 -7.61170506e-01 -7.27842569e-01 -8.15353215e-01 1.17885992e-01 -5.75558901e-01 7.90906370e-01 4.52516168e-01 3.97296578e-01 7.19931722e-01 1.00968078e-01 -6.58379495e-03 1.18317403e-01 -1.02227740e-01 5.58072925e-01 -7.91585684e-01 -4.46602583e-01 -1.37819439e-01 2.94815470e-02 -6.65304661e-01 3.12312335e-01 -1.13284039e+00 5.76454043e-01 -1.62917340e+00 7.93844834e-02 -1.56993955e-01 1.18578654e-02 4.90189433e-01 -6.68343186e-01 -4.77349125e-02 7.13439658e-02 -2.08955839e-01 -8.96278501e-01 4.85906482e-01 9.51879621e-01 -1.64425388e-01 5.73185571e-02 -9.09163952e-02 -9.94327128e-01 5.01370907e-01 7.44896531e-01 -5.01762986e-01 -3.87280494e-01 -5.23004949e-01 5.34812450e-01 1.38425365e-01 1.92151770e-01 -1.15163338e+00 -2.56866459e-02 -8.05888996e-02 3.57842803e-01 -2.05223739e-01 -2.38935828e-01 1.33532166e-01 -2.80611485e-01 2.67762721e-01 -1.02727616e+00 2.67129123e-01 2.38461480e-01 2.24613279e-01 -3.21493864e-01 -5.17938793e-01 4.80758250e-01 -2.17395902e-01 -3.74250710e-01 -1.99159980e-01 -2.42192984e-01 7.88791418e-01 4.95996535e-01 4.83710527e-01 -7.88119078e-01 -4.89946097e-01 -5.24564385e-01 2.67869413e-01 2.11234108e-01 4.67900723e-01 6.81386948e-01 -1.22322631e+00 -1.15747821e+00 1.95447221e-01 4.84950185e-01 -7.73100927e-02 -8.49567801e-02 2.81842779e-02 -4.06752825e-01 7.34163165e-01 2.75424421e-01 1.00641191e-01 -6.09586716e-01 3.71506959e-01 4.79549557e-01 -8.07342350e-01 -1.48315161e-01 1.08355820e+00 -3.72626424e-01 -6.03091776e-01 -2.84474224e-01 -4.70212400e-01 -7.76776597e-02 -4.94955838e-01 9.50998187e-01 1.68492630e-01 5.56682467e-01 -2.53249228e-01 -2.48629823e-01 -1.38081610e-01 -5.78877665e-02 -4.49018657e-01 8.33724856e-01 3.45086366e-01 2.69770056e-01 2.29250610e-01 6.90664232e-01 2.75044709e-01 -8.02033603e-01 -2.96493262e-01 2.05377012e-01 -1.15072601e-01 -2.38795832e-01 -1.63022983e+00 -4.33013856e-01 5.15245199e-01 -9.91217270e-02 2.90354073e-01 7.81011343e-01 -7.96259493e-02 9.45674181e-01 6.57194793e-01 5.17715335e-01 -1.20100605e+00 6.34834990e-02 8.98911536e-01 1.21159220e+00 -1.19279313e+00 -5.75044811e-01 -4.94166762e-01 -6.99392140e-01 9.37660575e-01 9.51405823e-01 -1.90994948e-01 1.99825615e-01 2.17001557e-01 8.13537538e-02 -6.40158728e-02 -1.39478219e+00 2.74083521e-02 5.12432218e-01 4.65506703e-01 7.13757217e-01 2.41972655e-01 -5.57658732e-01 9.96663988e-01 -8.87508452e-01 2.33691722e-01 5.14654398e-01 5.00557721e-01 -5.38808882e-01 -1.19989741e+00 -3.22613167e-03 8.48721623e-01 -2.51153499e-01 -5.77923059e-01 -6.22669756e-01 2.10214585e-01 1.68079048e-01 1.45338213e+00 -1.87367797e-01 -5.67243218e-01 7.28987694e-01 4.37882692e-01 3.09584945e-01 -1.36815667e+00 -6.65335119e-01 -7.32757568e-01 9.79696512e-01 -3.98390412e-01 2.14598238e-01 -6.71330094e-01 -1.37010467e+00 4.70931046e-02 -2.82256100e-02 5.93157709e-01 3.26971382e-01 1.36896431e+00 6.33660972e-01 5.77844083e-01 4.27262723e-01 -5.42864017e-02 -6.65914357e-01 -1.37555742e+00 -1.01385014e-02 6.57117307e-01 -6.25721291e-02 -1.27355382e-01 -1.58859804e-01 1.86415359e-01]
[11.485095977783203, 8.869659423828125]
1003b76e-fe23-432d-838a-e0381548c503
modeling-label-correlations-for-second-order
2204.03619
null
https://arxiv.org/abs/2204.03619v1
https://arxiv.org/pdf/2204.03619v1.pdf
Modeling Label Correlations for Second-Order Semantic Dependency Parsing with Mean-Field Inference
Second-order semantic parsing with end-to-end mean-field inference has been shown good performance. In this work we aim to improve this method by modeling label correlations between adjacent arcs. However, direct modeling leads to memory explosion because second-order score tensors have sizes of $O(n^3L^2)$ ($n$ is the sentence length and $L$ is the number of labels), which is not affordable. To tackle this computational challenge, we leverage tensor decomposition techniques, and interestingly, we show that the large second-order score tensors have no need to be materialized during mean-field inference, thereby reducing the computational complexity from cubic to quadratic. We conduct experiments on SemEval 2015 Task 18 English datasets, showing the effectiveness of modeling label correlations. Our code is publicly available at https://github.com/sustcsonglin/mean-field-dep-parsing.
['Kewei Tu', 'Songlin Yang']
2022-04-07
null
null
null
null
['semantic-dependency-parsing']
['natural-language-processing']
[ 5.24912812e-02 2.41363540e-01 -2.09493507e-02 -6.34397030e-01 -1.24129927e+00 -8.09484541e-01 2.21307293e-01 1.98119402e-01 -5.94239116e-01 5.47044516e-01 8.80300775e-02 -6.09060466e-01 -2.03982353e-01 -6.40153408e-01 -9.58956420e-01 -4.08441901e-01 -2.44867980e-01 5.02599895e-01 2.01714799e-01 2.02639475e-01 2.30352417e-01 3.43453772e-02 -1.03196371e+00 3.94078285e-01 7.18295753e-01 9.04581130e-01 1.72476247e-01 9.11988080e-01 -4.46756333e-01 9.85763311e-01 -2.23821476e-01 -9.68459845e-01 1.68104842e-01 -2.41774753e-01 -1.32666051e+00 -7.60496035e-02 8.20899844e-01 -1.29915744e-01 -2.35050276e-01 1.25150609e+00 1.92384928e-01 3.17415297e-01 1.36117771e-01 -8.86487842e-01 -2.14551389e-01 1.00151169e+00 -5.15858650e-01 2.68443912e-01 8.41382816e-02 -1.42122269e-01 1.56191325e+00 -7.01903403e-01 7.55497456e-01 1.32456183e+00 6.10351980e-01 4.92785305e-01 -1.35739315e+00 -4.05201733e-01 3.92198265e-01 3.25891942e-01 -9.70336914e-01 -3.19485515e-01 7.84754395e-01 -4.52863365e-01 1.18668401e+00 3.35137367e-01 2.95052439e-01 7.74541378e-01 -1.71117753e-01 1.04316366e+00 1.21913850e+00 -3.51264358e-01 1.30335867e-01 -3.75138611e-01 5.81490099e-01 1.15593314e+00 5.80646622e-04 -2.92494535e-01 -6.57105386e-01 -1.65952787e-01 3.14613611e-01 -1.74129859e-01 1.88585728e-01 8.49472638e-03 -1.14711177e+00 8.03845942e-01 5.05520463e-01 3.58113885e-01 -2.85362512e-01 5.43140411e-01 6.15159750e-01 1.54714495e-01 6.39052451e-01 2.40845487e-01 -7.61201203e-01 -5.11800051e-01 -9.72934246e-01 9.08704624e-02 6.09298289e-01 9.11503792e-01 9.05306816e-01 -3.14703017e-01 2.12020949e-02 1.01176798e+00 1.10361286e-01 5.53237975e-01 -1.93914324e-01 -1.52665544e+00 8.59476566e-01 4.81733471e-01 -1.41844839e-01 -8.68244588e-01 -5.29326797e-01 -3.73739541e-01 -6.92143321e-01 -2.36696407e-01 9.41979825e-01 -2.01003522e-01 -7.63564825e-01 1.83523130e+00 2.45335817e-01 -1.32935584e-01 -1.83061615e-01 8.23890626e-01 3.37234706e-01 4.44563776e-01 2.37488031e-01 -1.63041502e-02 1.76034677e+00 -1.11622405e+00 -6.46595597e-01 -5.76313257e-01 1.29253197e+00 -7.20904589e-01 1.07238877e+00 5.16315103e-01 -1.46588266e+00 -4.22490351e-02 -5.74277699e-01 -5.15260637e-01 -1.08764410e-01 2.58229107e-01 1.25595498e+00 3.77955645e-01 -1.04647589e+00 7.19282866e-01 -1.20748174e+00 2.68677212e-02 4.28036869e-01 2.58691877e-01 -4.45034415e-01 -3.00066710e-01 -1.02662826e+00 4.73491371e-01 1.84657827e-01 3.32634300e-01 -5.01893759e-01 -7.65682697e-01 -7.10159719e-01 1.06626242e-01 5.83950043e-01 -4.73747641e-01 1.28005636e+00 -3.47238839e-01 -1.15860188e+00 7.53297985e-01 -6.62680209e-01 -2.80703366e-01 3.04654121e-01 -1.98308051e-01 1.75372899e-01 1.42559618e-01 1.14019521e-01 5.69198191e-01 1.03757158e-01 -7.58032203e-01 -4.92919534e-01 -6.87873960e-01 4.75801289e-01 -1.21382251e-01 -3.20071906e-01 2.19463170e-01 -5.01376510e-01 -2.84492880e-01 6.46538675e-01 -1.14793146e+00 -4.37110066e-01 -3.00681442e-01 -4.51401889e-01 -3.98468256e-01 2.86302175e-02 -9.95857120e-01 1.11966443e+00 -2.18439031e+00 1.21278875e-01 -2.77873892e-02 3.52614224e-01 -7.81807899e-02 -8.27959254e-02 2.54380465e-01 1.87939674e-01 2.88893849e-01 -4.23318505e-01 -7.09205508e-01 2.52606839e-01 2.95521796e-01 -5.96084855e-02 1.48048893e-01 1.35216236e-01 6.81186616e-01 -8.89786482e-01 -6.23096168e-01 -2.69528538e-01 3.34055871e-01 -8.66952002e-01 -1.90681443e-01 -4.09147263e-01 2.34044284e-01 -5.14594853e-01 5.84550858e-01 8.34835529e-01 -5.22328496e-01 4.89651710e-01 -1.40039936e-01 6.41100388e-03 8.82616758e-01 -1.04660189e+00 2.06019115e+00 -6.45271599e-01 4.53282744e-01 2.45380506e-01 -9.46973145e-01 5.29150069e-01 -5.97631745e-03 4.27740246e-01 -6.25417352e-01 5.21422327e-02 3.37224603e-01 -7.65918568e-03 -2.32095614e-01 2.82466233e-01 -1.29770488e-01 -2.39279836e-01 3.72483015e-01 5.62052913e-02 2.42923453e-01 8.56827021e-01 5.15040696e-01 1.49089730e+00 1.12565337e-02 -5.13505042e-01 -2.37806514e-01 3.04146975e-01 1.52894512e-01 9.20346022e-01 5.54225802e-01 -3.88178490e-02 8.57268497e-02 1.18335044e+00 -3.87341201e-01 -9.34067070e-01 -9.40660417e-01 -1.01498708e-01 1.31589782e+00 -3.85139227e-01 -7.04988003e-01 -1.10556912e+00 -9.17567849e-01 -3.29533815e-01 6.21420503e-01 -5.35015464e-01 4.05889779e-01 -8.85199249e-01 -8.72118413e-01 5.85782409e-01 8.59210610e-01 3.24739516e-01 -6.85934961e-01 -4.33815122e-01 2.77678698e-01 -4.65019614e-01 -1.51249397e+00 -4.75143522e-01 3.20037603e-01 -1.16505861e+00 -9.20536757e-01 -2.71192580e-01 -6.18876100e-01 8.55545819e-01 -5.89168891e-02 1.30959988e+00 1.07504643e-01 -1.83437809e-01 -6.10509738e-02 -2.78183430e-01 -7.60787427e-02 -2.22240062e-03 2.45106384e-01 -3.30840617e-01 -3.75931263e-01 4.89062726e-01 -5.85352719e-01 -5.50954461e-01 -4.69447821e-02 -6.31930232e-01 1.18057631e-01 5.26070774e-01 9.57114339e-01 5.57734072e-01 -3.24826539e-02 -1.67036057e-01 -1.23837805e+00 1.52950764e-01 -4.79400679e-02 -9.73815858e-01 1.94312811e-01 -3.26780081e-01 4.02897567e-01 6.55105650e-01 -6.69979900e-02 -9.47467744e-01 5.85160889e-02 -1.91594288e-01 -5.72970472e-02 9.31190774e-02 7.49893248e-01 1.23588070e-01 2.79456466e-01 2.41242796e-01 -6.37886534e-03 -2.64739633e-01 -7.75870919e-01 3.13005477e-01 2.28807822e-01 1.19540431e-01 -9.24603283e-01 3.82168800e-01 5.55698633e-01 4.49348241e-01 -4.89724874e-01 -1.27402031e+00 -1.94775432e-01 -6.45677805e-01 1.46344393e-01 9.11722541e-01 -7.30113029e-01 -1.04662979e+00 3.05770159e-01 -1.47987378e+00 -4.82555956e-01 1.58839568e-01 5.69776833e-01 -3.11854273e-01 4.34220642e-01 -1.21555018e+00 -8.34380627e-01 -2.29286313e-01 -1.21178210e+00 1.03041863e+00 -3.18277419e-01 1.04335368e-01 -9.94853079e-01 -1.55983105e-01 9.96876478e-01 1.50259793e-01 -1.06343672e-01 1.11818779e+00 -4.07646626e-01 -8.81012738e-01 -1.77472550e-03 -4.19531107e-01 3.19791764e-01 -5.53904772e-01 -2.73604184e-01 -7.50114560e-01 -8.79536793e-02 -2.23225012e-01 -1.80547684e-01 9.70090866e-01 1.80691391e-01 1.16955101e+00 -3.49314332e-01 -9.09964368e-02 4.56411779e-01 1.30566120e+00 -1.65561542e-01 4.08157468e-01 6.13382384e-02 1.05833328e+00 7.57312059e-01 7.11258948e-01 3.33862811e-01 6.09629869e-01 6.42033696e-01 1.31315321e-01 -2.31938642e-02 -2.48847365e-01 -8.79908875e-02 3.85930896e-01 1.23777223e+00 -1.46666884e-01 5.26151806e-02 -1.11838627e+00 6.23945475e-01 -1.75906169e+00 -7.89673030e-01 -5.82085133e-01 1.84623182e+00 8.71325433e-01 3.32941711e-01 -2.10950617e-02 -1.27729320e-04 2.63416916e-01 1.94176435e-01 -2.21104562e-01 -7.22647846e-01 9.26812068e-02 2.63382226e-01 7.99585462e-01 9.63877082e-01 -9.41203892e-01 1.26691413e+00 5.41403770e+00 7.64764249e-01 -8.61256838e-01 5.66268682e-01 5.97083688e-01 -2.79341400e-01 -3.97000849e-01 4.25566673e-01 -9.61179078e-01 4.83482957e-01 1.07068098e+00 4.44422603e-01 6.46732092e-01 6.83913469e-01 -3.70098986e-02 -1.64376840e-01 -1.01697612e+00 6.33397222e-01 -2.97405690e-01 -1.10156333e+00 -4.86532241e-01 1.84266493e-01 4.27157819e-01 2.80271947e-01 -2.39539266e-01 3.49311709e-01 3.86248678e-01 -5.88958621e-01 4.83858317e-01 1.76469952e-01 4.87079173e-01 -6.61178887e-01 6.11012101e-01 3.25621396e-01 -1.10738063e+00 -6.32600905e-03 -5.36927044e-01 -1.77570909e-01 3.22549939e-01 1.06388891e+00 -7.96941817e-01 3.02981466e-01 8.67546260e-01 1.95090443e-01 -5.08174002e-01 2.82029420e-01 -3.17736566e-01 8.89211297e-01 -5.39715767e-01 -1.64034814e-01 5.32713473e-01 -2.85085052e-01 3.50098699e-01 1.24809313e+00 8.40020627e-02 1.20420635e-01 1.73995927e-01 7.62337565e-01 -2.29814500e-01 1.50380179e-01 -1.59067824e-01 -1.89868882e-01 2.17134684e-01 1.14107871e+00 -8.05314779e-01 -3.12474400e-01 -5.24042666e-01 8.71848881e-01 7.91551709e-01 4.10081334e-02 -8.65285516e-01 -3.22625130e-01 6.22147381e-01 1.24633610e-01 3.78576934e-01 -6.74310029e-01 -5.76956034e-01 -1.20263183e+00 5.25339186e-01 -4.79116350e-01 4.78340715e-01 -3.71387094e-01 -1.08587039e+00 4.42533970e-01 -4.99361530e-02 -4.45720494e-01 -2.49026746e-01 -9.18918014e-01 -4.61880751e-02 7.06786215e-01 -1.32835054e+00 -1.09541345e+00 8.47456381e-02 2.64985055e-01 3.88258368e-01 4.33650821e-01 8.30386996e-01 6.31124854e-01 -6.10695899e-01 8.75204861e-01 -5.57284243e-02 3.51081729e-01 2.45303795e-01 -1.47035444e+00 5.77875018e-01 1.03006411e+00 2.81563669e-01 7.16093063e-01 6.70730531e-01 -5.08418858e-01 -1.58475053e+00 -7.13509262e-01 1.58227539e+00 -5.26802897e-01 9.55491602e-01 -8.17113936e-01 -8.50706995e-01 6.87490523e-01 1.99811701e-02 2.51733303e-01 6.91314042e-01 8.82538497e-01 -7.96708345e-01 -1.22848451e-01 -7.96459913e-01 5.23339450e-01 1.42697036e+00 -6.45309865e-01 4.46163602e-02 7.73704648e-01 7.24950910e-01 -3.68749142e-01 -1.22956204e+00 3.59678537e-01 2.25110620e-01 -8.25394630e-01 7.06703246e-01 -4.61479753e-01 5.48359811e-01 -6.30164221e-02 -3.32710296e-01 -9.08431470e-01 -2.27509260e-01 -3.17030042e-01 7.75541179e-03 1.18185699e+00 8.91613722e-01 -5.68654776e-01 1.09552705e+00 9.14577544e-01 -1.00535370e-01 -8.98473442e-01 -9.74851489e-01 -8.19342911e-01 2.83319414e-01 -7.08231926e-01 3.40685815e-01 8.99197698e-01 1.29735187e-01 5.05139410e-01 -1.30584776e-01 2.06089601e-01 8.27474356e-01 6.60321191e-02 4.65036839e-01 -1.01254189e+00 -7.21631646e-01 -3.60806942e-01 -1.84849739e-01 -1.14695060e+00 4.03072566e-01 -1.16354609e+00 -6.76040947e-02 -1.42893589e+00 4.41866040e-01 -6.10074818e-01 -2.87797451e-01 8.28731418e-01 -1.70697883e-01 1.20753631e-01 4.10897672e-01 -8.84065256e-02 -8.68135273e-01 1.77111924e-01 1.17383671e+00 -6.05801642e-02 2.73786217e-01 -3.01943719e-01 -4.96246666e-01 7.11057305e-01 8.20912778e-01 -7.11463809e-01 -1.95214242e-01 -1.11489201e+00 5.97529650e-01 4.02137965e-01 2.76819885e-01 -7.03779459e-01 2.44569391e-01 6.80364445e-02 -1.27201006e-01 -4.94510472e-01 5.33124208e-01 -5.69519520e-01 -1.38489842e-01 2.73149848e-01 -4.45574343e-01 2.84750938e-01 -1.66354105e-02 2.85254478e-01 -2.80861288e-01 -4.69722211e-01 5.41900396e-01 -2.55366057e-01 -4.24466491e-01 2.39722043e-01 6.26761913e-02 2.61223048e-01 5.06230772e-01 3.21763724e-01 -5.23369491e-01 3.63523364e-02 -7.32811093e-01 1.33189291e-01 1.21046834e-01 3.40623371e-02 2.30398387e-01 -1.00779283e+00 -6.68390870e-01 -1.25769213e-01 -1.78069234e-01 1.89170763e-01 6.66718185e-01 1.21791148e+00 -5.95526874e-01 6.31798148e-01 3.55601519e-01 -6.01444721e-01 -1.29004085e+00 4.31192279e-01 -1.43598720e-01 -7.70058990e-01 -4.69484925e-01 1.30236828e+00 -4.55985889e-02 -6.57745481e-01 8.94559622e-02 -2.41743878e-01 2.83275127e-01 -2.28404347e-02 2.53938377e-01 5.18377483e-01 3.26695055e-01 -3.39415789e-01 -6.18510962e-01 4.98186678e-01 -3.86484802e-01 -3.29562873e-01 1.42483199e+00 1.31156459e-01 -5.43857992e-01 2.75457025e-01 1.55518758e+00 -2.46630814e-02 -1.14065838e+00 -2.70909816e-01 4.96056348e-01 -4.10160482e-01 1.77055985e-01 -7.71777391e-01 -1.22823668e+00 1.22424674e+00 2.68624365e-01 4.29300740e-02 9.77907658e-01 2.25213677e-01 1.10075247e+00 5.77063262e-01 6.41511858e-01 -1.12599945e+00 -1.57722488e-01 8.92217338e-01 3.28275681e-01 -1.08221781e+00 -1.94979146e-01 -7.70478308e-01 -4.95300770e-01 7.29776621e-01 3.67687404e-01 -9.75092053e-02 5.97416878e-01 2.64293224e-01 2.63797008e-02 -3.05016607e-01 -9.75128174e-01 -7.89353326e-02 -4.00691070e-02 -4.15057177e-03 8.54367077e-01 3.60659331e-01 -6.97895288e-01 4.28799421e-01 -4.30619091e-01 -2.54999638e-01 3.01275283e-01 9.26047146e-01 -2.12672442e-01 -1.53906798e+00 -5.95764928e-02 3.41407776e-01 -1.00051260e+00 -4.35645312e-01 -3.80503088e-02 2.75964439e-01 -7.64161497e-02 1.10345316e+00 -1.04563599e-02 -2.48378098e-01 2.30262116e-01 2.86934525e-01 7.26462483e-01 -3.29323649e-01 -4.11593616e-01 7.08841458e-02 3.60978961e-01 -7.73680985e-01 -2.56844401e-01 -9.12569523e-01 -1.49318326e+00 -5.22560298e-01 -1.07943259e-01 2.41027400e-01 1.03624892e+00 8.27426791e-01 4.43104893e-01 4.98134166e-01 3.40210527e-01 -2.23864108e-01 -6.67176664e-01 -1.09507716e+00 -4.32403803e-01 3.57741952e-01 -2.71378197e-02 -5.02962470e-01 -3.82098854e-01 -4.79917079e-02]
[10.346654891967773, 9.620504379272461]
c45b0bf2-c9bc-4b61-b571-b703fcd73a02
real-time-clustering-and-multi-target
1807.02851
null
http://arxiv.org/abs/1807.02851v1
http://arxiv.org/pdf/1807.02851v1.pdf
Real-time clustering and multi-target tracking using event-based sensors
Clustering is crucial for many computer vision applications such as robust tracking, object detection and segmentation. This work presents a real-time clustering technique that takes advantage of the unique properties of event-based vision sensors. Since event-based sensors trigger events only when the intensity changes, the data is sparse, with low redundancy. Thus, our approach redefines the well-known mean-shift clustering method using asynchronous events instead of conventional frames. The potential of our approach is demonstrated in a multi-target tracking application using Kalman filters to smooth the trajectories. We evaluated our method on an existing dataset with patterns of different shapes and speeds, and a new dataset that we collected. The sensor was attached to the Baxter robot in an eye-in-hand setup monitoring real-world objects in an action manipulation task. Clustering accuracy achieved an F-measure of 0.95, reducing the computational cost by 88% compared to the frame-based method. The average error for tracking was 2.5 pixels and the clustering achieved a consistent number of clusters along time.
['Francisco Barranco', 'Eduardo Ros', 'Cornelia Fermuller']
2018-07-08
null
null
null
null
['event-based-vision']
['computer-vision']
[ 2.12996811e-01 -4.38114464e-01 1.18428811e-01 -2.72353049e-02 -4.52534080e-01 -4.92586553e-01 4.54649866e-01 5.74827850e-01 -7.55895376e-01 3.76622051e-01 -5.08755505e-01 8.90460089e-02 -2.30482936e-01 -3.50997597e-01 -4.77371514e-01 -1.01321232e+00 -2.21435025e-01 4.34716344e-01 9.52566326e-01 3.55004430e-01 3.76741439e-01 9.34608340e-01 -1.79625630e+00 1.99440897e-01 5.07288754e-01 9.12417769e-01 6.02117121e-01 8.84101033e-01 1.01037964e-01 7.45602727e-01 -7.47059166e-01 4.31633353e-01 3.71004492e-01 -2.73689330e-01 -1.25654399e-01 5.24060667e-01 2.22803921e-01 6.96505457e-02 -1.08280055e-01 8.03525090e-01 3.47103626e-01 3.14474523e-01 3.86334211e-01 -1.40693927e+00 2.28661656e-01 4.74684052e-02 -7.90403128e-01 3.52649838e-01 4.27780151e-01 2.57699996e-01 8.80873874e-02 -4.20236766e-01 9.65449274e-01 1.03787112e+00 7.82407701e-01 4.96221215e-01 -1.25762808e+00 -5.00282526e-01 -1.50105283e-02 3.23532522e-01 -1.36367226e+00 -3.17360997e-01 3.74492139e-01 -8.20761025e-01 7.10813105e-01 -3.42228264e-02 8.82206619e-01 5.51904917e-01 5.50542533e-01 4.50991213e-01 1.20748734e+00 -5.14141083e-01 5.76367378e-01 -5.31614050e-02 9.71824080e-02 4.85846072e-01 4.26027954e-01 1.70276195e-01 -3.65304023e-01 -6.63011670e-02 9.81258988e-01 5.12055457e-01 -5.01581840e-02 -6.40984893e-01 -1.65386498e+00 6.36144221e-01 -1.38376700e-02 5.06635070e-01 -6.22305870e-01 2.80519783e-01 3.55247617e-01 -1.34958103e-01 3.98655534e-02 1.29576042e-01 -1.64778531e-01 -3.48611087e-01 -8.81340861e-01 6.50973246e-02 7.43694961e-01 9.06096697e-01 4.08262372e-01 -1.69774801e-01 -1.96335271e-01 1.08103804e-01 2.55193383e-01 5.04204631e-01 4.08337444e-01 -1.29919636e+00 -2.14507684e-01 6.74361050e-01 3.04012150e-01 -9.18999732e-01 -7.30658889e-01 2.74356514e-01 -4.48030770e-01 7.97924757e-01 8.02232265e-01 -2.23482877e-01 -9.15124714e-01 1.16665852e+00 7.37923980e-01 3.36990982e-01 -2.07869127e-01 8.05422425e-01 3.76365095e-01 4.68476266e-01 -1.00679532e-01 -7.89689422e-01 1.15790784e+00 -5.29055774e-01 -1.14194560e+00 4.38196719e-01 1.41987473e-01 -9.33953285e-01 4.74122375e-01 8.33465636e-01 -9.64755297e-01 -6.07702792e-01 -8.65764558e-01 5.62297285e-01 -4.50358331e-01 -3.70679274e-02 5.19209981e-01 5.37425399e-01 -9.25617993e-01 5.94537437e-01 -1.37522805e+00 -8.76005948e-01 2.58015990e-01 6.20459259e-01 -2.82339841e-01 1.38532832e-01 -2.35354438e-01 8.80396783e-01 5.97841322e-01 -1.18494682e-01 -8.26889455e-01 -4.75136817e-01 -5.66539526e-01 -4.15346473e-01 4.25512165e-01 -2.79953212e-01 1.15060616e+00 -5.77415466e-01 -1.67806864e+00 6.36157036e-01 -2.28371203e-01 -5.82540572e-01 6.06374145e-01 -1.58225909e-01 -2.64554858e-01 5.60075879e-01 1.01907775e-02 5.85182786e-01 8.82031858e-01 -1.18305743e+00 -1.01138651e+00 -3.03038836e-01 -3.19037706e-01 -1.27526313e-01 2.80749118e-05 1.92008615e-01 -3.44624132e-01 -1.89842731e-01 4.90162149e-03 -9.92754817e-01 -5.40407658e-01 7.15386495e-02 9.55187976e-02 -2.04232797e-01 1.26351798e+00 -8.91682580e-02 1.02371788e+00 -2.05254412e+00 -1.13660544e-01 2.10924998e-01 1.57396212e-01 1.97488517e-01 2.62921751e-01 6.15461051e-01 1.07029684e-01 -6.85849488e-01 -4.62192409e-02 -5.51981404e-02 -4.42522287e-01 2.39629552e-01 1.23553626e-01 9.78179097e-01 -2.01974005e-01 4.15194392e-01 -1.12528217e+00 -7.43148088e-01 1.00409603e+00 3.61006677e-01 -8.03414807e-02 -5.49205169e-02 -1.72417328e-01 6.90497935e-01 -1.12944782e-01 5.50827920e-01 4.49324131e-01 -7.93116689e-02 6.12467378e-02 -9.80611444e-02 -6.51141346e-01 -5.17335653e-01 -1.75970602e+00 1.83791685e+00 3.51752639e-02 8.04169357e-01 2.44772583e-01 -8.01730454e-01 7.80258894e-01 2.48895288e-01 1.31532264e+00 -4.82879072e-01 2.88602948e-01 -1.27595454e-01 -1.08559944e-01 -7.18303442e-01 5.44584453e-01 3.05284858e-01 2.16715574e-01 2.88994908e-01 -2.76648730e-01 -1.08780243e-01 5.93553841e-01 2.25079939e-01 1.27510345e+00 1.42752662e-01 4.19790417e-01 -3.03652197e-01 2.36332119e-01 6.12481117e-01 4.51099813e-01 7.60961115e-01 -4.84883159e-01 5.39478719e-01 -9.62576121e-02 -2.88030237e-01 -6.05909884e-01 -9.58487391e-01 -1.49541736e-01 5.07397115e-01 4.66418326e-01 -3.99684608e-01 -7.00402439e-01 -2.00164646e-01 1.41022429e-02 3.44854325e-01 -3.87507498e-01 1.40910715e-01 -5.72761476e-01 -4.10733998e-01 1.47913648e-02 4.13005531e-01 8.66162300e-04 -9.26444411e-01 -1.65741122e+00 6.67359054e-01 2.59345323e-01 -1.20103621e+00 -9.65148658e-02 8.00001025e-02 -1.04173172e+00 -1.61932397e+00 -3.49210709e-01 -5.56723833e-01 7.29172885e-01 4.94828105e-01 6.54731035e-01 -2.76288837e-01 -9.07177687e-01 1.00124681e+00 -3.86251986e-01 -8.43615711e-01 -3.08123291e-01 -6.41881824e-01 2.61838287e-01 1.03585059e-02 6.08120441e-01 -4.34271693e-02 -6.34252369e-01 3.91422629e-01 -9.54186738e-01 -3.67716014e-01 1.93004280e-01 5.78027606e-01 8.74953508e-01 3.45312923e-01 3.22277278e-01 -3.30020338e-01 2.81624883e-01 -1.53326109e-01 -1.11523354e+00 -7.95222074e-03 -3.18334550e-01 -6.84937775e-01 3.14702928e-01 -8.30201209e-01 -7.42471933e-01 9.83213484e-01 6.16008997e-01 -8.84427428e-01 -4.95688528e-01 5.67466281e-02 2.45023027e-01 -1.47041246e-01 4.94619757e-01 -1.31385028e-01 5.00884712e-01 -3.97503003e-02 4.47093099e-01 6.01668119e-01 7.92762399e-01 -1.86405599e-01 2.59703040e-01 1.06565523e+00 2.40958050e-01 -1.07962060e+00 1.17933620e-02 -1.09577477e+00 -8.36121976e-01 -8.51081729e-01 1.05025005e+00 -6.73420429e-01 -1.30243957e+00 7.03203857e-01 -1.22311795e+00 -2.88399905e-01 -5.63730061e-01 9.47858870e-01 -7.37344325e-01 3.64161730e-01 -5.48000991e-01 -1.03826821e+00 -2.37420294e-02 -1.09392321e+00 9.80556607e-01 3.70950162e-01 -2.07292199e-01 -8.79981577e-01 5.71411140e-02 -4.29377370e-02 1.57177463e-01 8.53539228e-01 -5.14062867e-03 -4.57155466e-01 -7.24996448e-01 -4.13182974e-01 2.17463553e-01 -1.45907998e-01 4.06364799e-01 3.83567095e-01 -8.24536622e-01 -4.35474247e-01 3.47266942e-01 3.32354754e-01 5.55254936e-01 9.75275874e-01 9.16486800e-01 3.23845237e-01 -7.78493345e-01 -1.12216853e-01 1.54690623e+00 8.50625396e-01 5.83581388e-01 3.07373971e-01 5.67681432e-01 5.29586375e-01 1.13393760e+00 6.90173864e-01 3.47345509e-02 7.38671303e-01 4.36713636e-01 -1.70945302e-02 -4.10348140e-02 4.09725338e-01 4.25506175e-01 3.55065584e-01 -1.39647469e-01 5.99062219e-02 -8.31610382e-01 7.57319987e-01 -2.18761325e+00 -1.21191752e+00 -6.28442466e-01 2.40523791e+00 5.11884093e-01 -2.99330987e-02 4.47990447e-01 1.81988537e-01 1.08405912e+00 -5.48723638e-01 -3.78328294e-01 -6.47783354e-02 3.92897844e-01 7.13081807e-02 7.18950093e-01 3.19784522e-01 -1.25013065e+00 5.17902315e-01 6.29237556e+00 6.29616916e-01 -1.12019324e+00 2.32653078e-02 -1.99371085e-01 -4.02682185e-01 7.46728837e-01 -2.12089866e-01 -8.56322408e-01 7.22342432e-01 1.00877500e+00 -1.36063859e-01 5.32783121e-02 6.12091541e-01 6.24620497e-01 -7.74806023e-01 -1.24431860e+00 1.11286402e+00 1.19144239e-01 -1.21497011e+00 -6.29444182e-01 1.15607224e-01 5.26861191e-01 -3.03646307e-02 -4.47505742e-01 -4.35563564e-01 2.52612025e-01 -6.05281472e-01 6.84073985e-01 7.37425327e-01 3.18028569e-01 -5.81898808e-01 3.65855008e-01 3.49904209e-01 -1.26482058e+00 -1.21599566e-02 -1.01797268e-01 -1.11298203e-01 4.64516938e-01 8.07274282e-01 -1.05719674e+00 3.57750088e-01 8.82507741e-01 8.26620400e-01 -2.75291294e-01 1.55795920e+00 2.87799209e-01 1.48289233e-01 -7.83318281e-01 -6.21726587e-02 -4.87977266e-03 -3.45471352e-01 7.59706438e-01 1.36244321e+00 2.16326728e-01 1.20524004e-01 4.69867468e-01 4.42101568e-01 6.54826045e-01 -2.82030702e-01 -8.86246145e-01 1.70752168e-01 5.42083502e-01 1.42421913e+00 -1.37478173e+00 -3.93011451e-01 -3.71386141e-01 6.74526870e-01 -4.84615892e-01 1.02902353e-01 -8.69820356e-01 -5.80986321e-01 2.72821218e-01 1.25830963e-01 7.44884551e-01 -5.73421240e-01 -2.18902081e-01 -5.64308405e-01 -1.41589627e-01 -4.99543935e-01 3.88539642e-01 -6.04935288e-01 -9.73457932e-01 1.36545196e-01 2.92394847e-01 -1.60101593e+00 -3.26025039e-01 -6.03908837e-01 -4.10204500e-01 2.93008924e-01 -9.85947669e-01 -7.98867583e-01 -5.39293468e-01 8.73779297e-01 6.64036870e-01 6.15155734e-02 5.70615888e-01 2.54701793e-01 -4.97874498e-01 -1.61123738e-01 3.74758333e-01 -1.79737896e-01 6.92242444e-01 -1.13660192e+00 -3.10639232e-01 1.02188444e+00 -8.64098817e-02 5.78572392e-01 7.59275615e-01 -7.39723802e-01 -1.58744264e+00 -1.04194295e+00 3.23126465e-01 -5.18215477e-01 8.23541164e-01 -1.26562610e-01 -7.56197393e-01 5.63235819e-01 4.23926383e-01 1.25254188e-02 5.05740047e-01 -6.21561825e-01 4.67209280e-01 -2.04825655e-01 -1.40216494e+00 3.01594019e-01 6.27013326e-01 -4.15652618e-03 -6.42980874e-01 5.03077328e-01 3.04140836e-01 -7.28996992e-01 -1.06328881e+00 1.78640336e-01 5.21530092e-01 -9.44668770e-01 8.83418679e-01 4.47059572e-02 -2.48420313e-01 -9.04820323e-01 1.43727511e-01 -9.77140248e-01 -1.73645481e-01 -8.86319816e-01 -1.95220411e-01 1.03153670e+00 -2.63214856e-01 -6.00880504e-01 6.29246831e-01 4.58300889e-01 -1.34822607e-01 -2.07707465e-01 -1.04091311e+00 -9.73793089e-01 -6.31789148e-01 -3.08351934e-01 -8.30935538e-02 7.12498724e-01 1.09148562e-01 -1.02493793e-01 1.65735051e-01 1.54498935e-01 9.17793989e-01 1.84432730e-01 8.06142151e-01 -1.48046851e+00 7.53381327e-02 -2.19653443e-01 -8.27469170e-01 -4.70913231e-01 -2.12810919e-01 -3.50227356e-01 3.23346764e-01 -1.48220944e+00 5.36792949e-02 -2.31966004e-01 -6.43321946e-02 2.97646910e-01 1.26957148e-01 5.85974054e-03 3.70662481e-01 2.74581283e-01 -9.44395006e-01 -1.72417805e-01 8.40071917e-01 -1.18922433e-02 -4.17827994e-01 -5.34490831e-02 1.82674989e-01 8.82920146e-01 7.43681371e-01 -5.46261072e-01 -2.30305150e-01 9.66835767e-02 -4.45020705e-01 -1.08104749e-02 4.40795600e-01 -1.35348165e+00 9.00417566e-01 -2.19214126e-01 3.70647430e-01 -1.03215408e+00 3.36140335e-01 -1.50212026e+00 5.93487263e-01 9.25024629e-01 6.94998652e-02 1.63843766e-01 4.96472061e-01 8.99010003e-01 -1.20122924e-01 -1.31435871e-01 9.42474246e-01 -2.62432396e-01 -1.13365769e+00 -1.48168981e-01 -7.73113489e-01 -4.57949013e-01 2.03438902e+00 -7.07566559e-01 -2.23141089e-01 8.42462108e-02 -8.45187366e-01 2.20552817e-01 7.28551626e-01 2.07070604e-01 7.03207970e-01 -1.06349730e+00 -2.78939486e-01 1.51571706e-01 2.60336045e-02 1.14102013e-01 -6.07279837e-02 1.33219135e+00 -5.77148259e-01 2.74424940e-01 -3.70283365e-01 -1.47016096e+00 -1.65347457e+00 7.36831367e-01 1.01429164e-01 2.35791966e-01 -8.46407712e-01 2.00187862e-01 -3.39682281e-01 2.48965666e-01 4.29753214e-01 -5.66795051e-01 -1.76717505e-01 2.07808837e-01 6.67719662e-01 9.99220967e-01 1.63415805e-01 -3.56099337e-01 -5.17165363e-01 9.07735646e-01 1.36421934e-01 -1.19292215e-01 1.06613672e+00 -2.24792510e-01 -1.25522077e-01 7.34492600e-01 6.22670174e-01 -1.57812700e-01 -1.41263902e+00 1.75302520e-01 1.57204911e-01 -4.72598612e-01 -1.16942748e-01 -4.60560590e-01 -9.13698137e-01 5.52653134e-01 1.12918377e+00 4.80837733e-01 1.12582922e+00 -1.88029800e-02 5.07216334e-01 4.08191711e-01 8.47620070e-01 -1.26686585e+00 6.51599169e-02 3.05909872e-01 4.04997349e-01 -1.20981753e+00 1.09099351e-01 -3.89867902e-01 -5.50377846e-01 1.17935097e+00 3.40858608e-01 -2.22831979e-01 4.55992252e-01 6.15509331e-01 1.82768986e-01 -4.32491064e-01 -4.32739556e-01 -4.84280050e-01 -2.40139812e-01 9.29813385e-01 -2.16243248e-02 4.43371385e-02 -3.36477518e-01 -2.73380429e-01 4.91340607e-01 5.28237343e-01 7.63402343e-01 1.49256122e+00 -7.41988420e-01 -6.50580943e-01 -8.47194910e-01 2.49266744e-01 -3.21839958e-01 6.09456837e-01 -9.63345021e-02 1.00399113e+00 1.06210150e-01 1.49345374e+00 3.80358160e-01 -1.41820535e-01 5.77265561e-01 -1.69308752e-01 6.96954250e-01 -4.93917346e-01 -5.94041109e-01 3.63871157e-01 -4.13093567e-01 -1.05643892e+00 -1.19639540e+00 -1.11546540e+00 -1.72943783e+00 4.92662676e-02 -3.16258997e-01 -1.50113320e-02 9.79064345e-01 6.04419827e-01 4.66460764e-01 6.11760676e-01 4.66417015e-01 -1.19239521e+00 1.42451059e-02 -6.44538462e-01 -6.86717391e-01 4.08269733e-01 3.79481405e-01 -8.58843982e-01 -1.57302737e-01 6.85357928e-01]
[8.471587181091309, -1.2698675394058228]
aa1abf1f-f7cc-41ce-a5b7-e83cec2f3457
assessment-of-reinforcement-learning
2305.05812
null
https://arxiv.org/abs/2305.05812v1
https://arxiv.org/pdf/2305.05812v1.pdf
Assessment of Reinforcement Learning Algorithms for Nuclear Power Plant Fuel Optimization
The nuclear fuel loading pattern optimization problem has been studied since the dawn of the commercial nuclear energy industry. It is characterized by multiple objectives and constraints, with a very high number of candidate patterns, which makes it impossible to solve explicitly. Stochastic optimization methodologies are used by different nuclear utilities and vendors to perform fuel cycle reload design. Nevertheless, hand-designed solutions continue to be the prevalent method in the industry. To improve the state-of-the-art core reload patterns, we aim to create a method as scalable as possible, that agrees with the designer's goal of performance and safety. To help in this task Deep Reinforcement Learning (RL), in particular, Proximal Policy Optimization is leveraged. RL has recently experienced a strong impetus from its successes applied to games. This paper lays out the foundation of this method and proposes to study the behavior of several hyper-parameters that influence the RL algorithm via a multi-measure approach helped with statistical tests. The algorithm is highly dependent on multiple factors such as the shape of the objective function derived for the core design that behaves as a fudge factor that affects the stability of the learning. But also an exploration/exploitation trade-off that manifests through different parameters such as the number of loading patterns seen by the agents per episode, the number of samples collected before a policy update, and an entropy factor that increases the randomness of the policy trained. Experimental results also demonstrate the effectiveness of the method in finding high-quality solutions from scratch within a reasonable amount of time. Future work must include applying the algorithms to wide range of applications and comparing them to state-of-the-art implementation of stochastic optimization methods.
['Koroush Shirvan', 'Paul Seurin']
2023-05-09
null
null
null
null
['stochastic-optimization']
['methodology']
[-1.44353092e-01 -6.74876720e-02 -3.14537793e-01 6.16917619e-03 -3.98497015e-01 -5.86329937e-01 5.72762311e-01 9.26865339e-02 -6.71660960e-01 9.78685856e-01 4.06721085e-02 -2.93785036e-01 -6.97717786e-01 -8.21299613e-01 -4.47447479e-01 -9.96689022e-01 -7.16194808e-02 7.60846794e-01 2.25260556e-01 -4.26775932e-01 4.62860554e-01 5.37861764e-01 -1.45149529e+00 -1.96280628e-01 5.96514225e-01 8.57310712e-01 4.37973201e-01 3.70732218e-01 1.78133503e-01 4.62298214e-01 -5.83143413e-01 1.10627569e-01 4.44627643e-01 -4.26295400e-01 -5.20430744e-01 -3.57258469e-02 -6.42959058e-01 5.39009571e-02 1.07258651e-02 1.00554037e+00 6.59545600e-01 3.25813293e-01 6.34078920e-01 -1.00532591e+00 8.54750276e-02 6.57408237e-01 -4.06657696e-01 2.97360331e-01 5.08751683e-02 5.86534202e-01 8.73149037e-01 -2.93882817e-01 4.88963336e-01 9.89968121e-01 2.00341120e-01 3.44160318e-01 -1.02183843e+00 -4.64960814e-01 -7.24008828e-02 2.62816936e-01 -1.26359761e+00 -1.47014474e-02 7.00310528e-01 -4.13699001e-01 9.58138466e-01 1.72245622e-01 8.88698220e-01 7.75513232e-01 3.44545960e-01 3.93646687e-01 1.32318223e+00 -4.54866469e-01 8.99451196e-01 3.06492060e-01 -3.82084459e-01 4.44874913e-01 3.06398869e-01 7.19571412e-01 -1.19169243e-01 -3.18340072e-03 4.31724459e-01 -4.36013699e-01 -1.07064836e-01 -6.36360168e-01 -5.37281513e-01 1.16023481e+00 1.02694191e-01 5.42151392e-01 -4.17465329e-01 2.41076931e-01 4.15430576e-01 2.62419820e-01 -6.78740516e-02 9.70882475e-01 -4.01279122e-01 -6.21016860e-01 -9.53210592e-01 5.39820969e-01 9.12311554e-01 2.28872150e-01 3.94381166e-01 1.85220331e-01 -1.26419470e-01 5.61192334e-01 4.18771207e-01 2.39921987e-01 4.96455997e-01 -7.12924182e-01 3.84555846e-01 5.31834245e-01 2.38536417e-01 -6.67485178e-01 -4.96672302e-01 -7.66291559e-01 -3.40227962e-01 8.35626364e-01 4.58143234e-01 -5.02595663e-01 -5.11444986e-01 1.58914614e+00 4.54220682e-01 -4.07388836e-01 -6.52054623e-02 1.02379441e+00 -2.41108939e-01 6.25361383e-01 -1.13567404e-01 -4.64350998e-01 1.09807444e+00 -6.29058003e-01 -5.10452986e-01 -1.43573940e-01 2.89865673e-01 -5.77671170e-01 8.38776886e-01 5.18228412e-01 -1.09517276e+00 -2.22158998e-01 -1.39023554e+00 9.11399245e-01 -4.01058704e-01 -1.18972421e-01 4.29851294e-01 1.10368752e+00 -6.49011612e-01 8.98321152e-01 -8.14010441e-01 -3.00961107e-01 2.07428753e-01 7.09992349e-01 2.35700458e-01 2.20247388e-01 -1.12012041e+00 1.35851824e+00 7.43609667e-01 1.08534895e-01 -1.09220803e+00 -5.93491971e-01 -3.56429964e-01 1.95313439e-01 9.08665240e-01 -6.02944791e-01 1.16658759e+00 -9.95896339e-01 -1.98319626e+00 2.19880179e-01 5.82983613e-01 -5.25904834e-01 7.44912326e-01 2.06743091e-01 -1.77222043e-01 -1.83515295e-01 -2.10909858e-01 1.90960974e-01 7.27854073e-01 -1.14278626e+00 -6.91024184e-01 -1.77256316e-01 1.12288758e-01 3.30825329e-01 -1.25935137e-01 -3.89652215e-02 3.59646976e-02 -3.35150152e-01 -4.72138792e-01 -1.05919290e+00 -5.55668056e-01 -6.48492038e-01 -1.32726267e-01 -1.44439474e-01 6.95705771e-01 -2.16301188e-01 1.30415356e+00 -1.79573214e+00 5.31007767e-01 6.34351850e-01 -3.67040902e-01 1.39912784e-01 2.61265725e-01 6.77410066e-01 -2.64167394e-02 7.46246278e-02 -1.35007188e-01 1.37850072e-03 2.67714083e-01 2.88151622e-01 6.93605766e-02 6.29340231e-01 -5.18589802e-02 4.94322628e-01 -8.53151262e-01 -4.37366292e-02 2.70674855e-01 -1.28569072e-02 -5.89424551e-01 2.01091170e-01 -4.99823689e-01 2.18571797e-01 -8.09215367e-01 3.23110640e-01 2.35172644e-01 8.65560248e-02 3.37986141e-01 -6.16409332e-02 -6.01536810e-01 -1.94476128e-01 -1.32023776e+00 1.32106721e+00 -4.16337878e-01 7.93339834e-02 8.43100548e-02 -1.21439779e+00 8.21437001e-01 2.17295334e-01 7.84431040e-01 -8.29258502e-01 4.85889852e-01 2.90990382e-01 3.92238855e-01 -3.89760524e-01 4.56525624e-01 -3.04527611e-01 -3.23428884e-02 4.69573230e-01 -1.78309932e-01 -3.84077013e-01 6.44740105e-01 -3.22560012e-01 1.11576211e+00 2.35452503e-01 3.74224037e-01 -6.81463599e-01 5.69282413e-01 1.16664954e-01 5.88470697e-01 5.20378888e-01 -5.43440059e-02 2.39456937e-01 7.56897688e-01 -3.95453662e-01 -1.28850639e+00 -6.83572292e-01 -2.22272649e-02 5.83340466e-01 1.41928401e-02 -1.46246970e-01 -6.83254838e-01 -5.28046250e-01 3.03621106e-02 1.02455902e+00 -5.58677554e-01 -3.29663873e-01 -5.50374806e-01 -1.06147885e+00 1.63170710e-01 1.28912196e-01 2.51157790e-01 -1.12557864e+00 -1.15999174e+00 5.35863698e-01 5.07826626e-01 -7.91453063e-01 -1.29966408e-01 5.23828447e-01 -7.38803923e-01 -1.01936221e+00 -4.18692440e-01 -1.00461990e-01 4.21483546e-01 -4.10860658e-01 1.07848442e+00 -1.38678417e-01 -4.17255878e-01 3.17379117e-01 -3.06071579e-01 -4.69377130e-01 -6.37830138e-01 2.93114275e-01 6.96147531e-02 -1.69856519e-01 -4.90084179e-02 -5.28052449e-01 -5.52100003e-01 5.03002286e-01 -9.71863747e-01 -2.87441462e-01 7.29434609e-01 7.27060080e-01 3.73457700e-01 6.56905353e-01 6.46131039e-01 -5.08079886e-01 9.00398850e-01 -5.28154492e-01 -1.15417743e+00 2.11142987e-01 -9.78392482e-01 5.12214422e-01 7.30557024e-01 -3.41143996e-01 -8.67966056e-01 -1.03451625e-01 2.73896623e-02 -2.51107514e-01 1.91522613e-01 4.76793557e-01 -5.07547706e-02 -6.82651326e-02 4.71158028e-01 -7.96150789e-03 1.51079744e-01 -1.63232967e-01 9.36190486e-02 2.54936337e-01 -1.06008671e-01 -8.41646433e-01 9.27290261e-01 5.57047091e-02 2.25336939e-01 -6.20373666e-01 -4.44917679e-01 -9.69872698e-02 -6.51220698e-03 -5.72512448e-01 7.92289257e-01 -3.60320419e-01 -9.84900594e-01 1.49231419e-01 -6.37821674e-01 -4.95970160e-01 -4.67370540e-01 4.43293691e-01 -5.81789553e-01 -5.26706502e-03 -3.12671587e-02 -1.06351388e+00 -1.91392466e-01 -1.46584380e+00 2.78672576e-01 5.32330692e-01 -2.12210983e-01 -8.68026316e-01 2.40803748e-01 2.04701960e-01 6.72453225e-01 3.13036919e-01 1.03308976e+00 -3.45864356e-01 -6.70572400e-01 8.63464624e-02 2.88569272e-01 2.55082190e-01 -1.21707149e-01 -4.23467048e-02 -3.70025426e-01 -5.00330269e-01 2.17715979e-01 -2.47888461e-01 3.72993261e-01 5.08233249e-01 7.98413754e-01 -1.02920376e-01 -1.40184775e-01 2.68600106e-01 1.75372326e+00 8.04256976e-01 6.11420989e-01 8.86747658e-01 1.45899385e-01 3.94900799e-01 7.30077028e-01 7.75932789e-01 -2.73219943e-02 8.78106177e-01 8.24963570e-01 1.96877033e-01 2.59770334e-01 4.57604602e-02 4.17928606e-01 3.50915074e-01 -1.41245097e-01 -2.66049981e-01 -7.69682229e-01 1.38243914e-01 -1.85101318e+00 -9.87910330e-01 5.70756853e-01 2.38393140e+00 5.50222993e-01 4.67725068e-01 3.87625396e-01 2.84618884e-01 4.64007914e-01 7.83429071e-02 -6.33664072e-01 -7.86779225e-01 1.52127609e-01 1.26549557e-01 9.22057569e-01 4.57069695e-01 -6.19902790e-01 4.73254055e-01 5.68614101e+00 1.05151224e+00 -1.08073056e+00 -1.04641460e-01 6.66810274e-01 -3.33213061e-01 -3.72576922e-01 1.61593094e-01 -7.06888556e-01 7.86311984e-01 8.95914018e-01 -2.36544162e-01 9.47473705e-01 8.18789303e-01 6.93516076e-01 -5.37675440e-01 -7.88166761e-01 6.61352754e-01 -2.38182485e-01 -1.25733173e+00 -5.51664948e-01 3.15336585e-01 8.46587360e-01 -5.52911870e-02 7.72678526e-03 3.38987321e-01 3.44692975e-01 -1.06775904e+00 9.71252799e-01 3.31105292e-01 1.23881727e-01 -1.12205946e+00 8.21770012e-01 3.68131191e-01 -9.26019847e-01 -7.37487078e-01 -8.51536766e-02 1.43339275e-03 2.94427693e-01 6.46664917e-01 -9.33634102e-01 5.93660593e-01 4.55985546e-01 -1.25487566e-01 -2.11047128e-01 1.07379651e+00 -2.57699080e-02 4.45196271e-01 -6.27104580e-01 -7.43354797e-01 6.86752915e-01 -6.03880703e-01 7.62859344e-01 7.28089511e-01 3.82554710e-01 -1.79136664e-01 3.12921137e-01 1.06233680e+00 4.87781227e-01 1.49258390e-01 -3.78108323e-01 -3.13609034e-01 4.24128562e-01 1.14425528e+00 -9.58276033e-01 3.70494872e-01 -1.49581805e-01 2.35780284e-01 1.62361130e-01 2.28388920e-01 -1.05868852e+00 2.58563645e-02 5.90313375e-01 3.13396335e-01 4.17055547e-01 -3.48848462e-01 -2.72836864e-01 -3.93008798e-01 -1.71128035e-01 -1.13709068e+00 3.22101295e-01 -1.94559917e-01 -1.02566016e+00 4.42522764e-01 2.61513233e-01 -1.03834867e+00 -4.51316714e-01 -5.35891235e-01 -7.14703858e-01 6.35594428e-01 -1.27693009e+00 -4.71244544e-01 2.19992906e-01 2.33283550e-01 5.41689515e-01 -6.34979248e-01 3.52798939e-01 1.80496886e-01 -7.58645773e-01 1.98470250e-01 3.19413185e-01 -5.12883306e-01 1.39770791e-01 -1.22806144e+00 -2.79277980e-01 6.42602444e-01 -2.61858612e-01 1.42270833e-01 1.21379924e+00 -4.95654732e-01 -1.80071735e+00 -3.39343995e-01 9.18669999e-02 9.26349089e-02 8.44890952e-01 -1.88060720e-02 -2.08958805e-01 -1.06474936e-01 3.47433239e-01 -4.63865727e-01 2.35484719e-01 5.63301183e-02 5.53404093e-01 -3.98867697e-01 -1.21533144e+00 6.49595737e-01 4.41314489e-01 4.51799482e-02 -2.41236970e-01 1.19589634e-01 1.48259178e-01 -3.15579742e-01 -6.29820466e-01 4.06433493e-01 2.90847600e-01 -1.08810568e+00 7.80775666e-01 -2.29578167e-01 1.15695417e-01 -3.43524545e-01 5.21008670e-02 -1.53111613e+00 -2.74282098e-01 -8.21579397e-01 -1.15247779e-01 1.03660285e+00 6.75383866e-01 -6.47349358e-01 1.00880587e+00 5.75375319e-01 6.53975159e-02 -1.32794511e+00 -1.12479174e+00 -8.98967385e-01 4.44651656e-02 -5.02460003e-02 4.07715321e-01 4.10059065e-01 -7.76455328e-02 1.77629039e-01 -2.31975421e-01 4.45824042e-02 5.63948929e-01 -9.69161186e-03 4.49182868e-01 -7.59667397e-01 -8.14801395e-01 -6.19283915e-01 -1.82479754e-01 -3.67497504e-01 -1.78250611e-01 -5.01755714e-01 -8.23624805e-02 -1.00137627e+00 -6.65234327e-02 -6.76962495e-01 -2.77519494e-01 5.37441149e-02 2.23188505e-01 -5.71956754e-01 4.05630112e-01 -2.59703219e-01 -2.68802166e-01 7.66569018e-01 1.08095896e+00 -6.46090955e-02 -4.87633586e-01 1.97570220e-01 -3.99796963e-01 4.49285328e-01 1.00757694e+00 -6.52361035e-01 -6.47517622e-01 1.77651998e-02 6.07128501e-01 2.34079614e-01 -6.45933673e-02 -1.35828280e+00 1.88262522e-01 -3.94426405e-01 1.72378018e-01 -3.33689600e-01 2.09757984e-01 -1.08037364e+00 5.55781841e-01 7.71242917e-01 -4.20456491e-02 4.15980548e-01 4.08877619e-02 5.47405243e-01 -1.03687651e-01 -9.03117061e-01 9.51511860e-01 -3.40484291e-01 -4.50196952e-01 9.93461683e-02 -5.19595742e-01 7.83590749e-02 1.29420173e+00 -2.71017790e-01 2.26553261e-01 -1.41922474e-01 -4.13575858e-01 4.58020449e-01 2.68197268e-01 4.12250131e-01 6.20004302e-03 -8.95972013e-01 -5.27202666e-01 -1.25814244e-01 -5.32817781e-01 -3.91633421e-01 1.28077846e-02 5.87205708e-01 -4.93784219e-01 2.56508768e-01 -2.34453663e-01 -3.72878134e-01 -9.54069018e-01 5.36781013e-01 6.63318992e-01 -8.36037457e-01 -2.77419746e-01 2.61703521e-01 -5.14778435e-01 -6.89882189e-02 1.70805588e-01 -1.69181377e-01 -1.45298958e-01 3.02619636e-01 1.93915024e-01 5.33907294e-01 2.20957860e-01 1.11395158e-01 -1.97991714e-01 4.91239846e-01 1.60891175e-01 -3.58254850e-01 1.48986232e+00 1.82178259e-01 2.81078815e-01 1.81285411e-01 8.03988338e-01 -2.01742157e-01 -1.33105433e+00 3.41105342e-01 1.58830121e-01 -3.49308580e-01 3.57052773e-01 -9.85598445e-01 -1.09693253e+00 2.31081739e-01 8.34164441e-01 3.40132564e-01 9.40529108e-01 -4.58644390e-01 4.33902085e-01 3.07298213e-01 5.67731202e-01 -1.75421000e+00 2.53474712e-01 4.45559442e-01 7.89108217e-01 -9.23021436e-01 2.31323019e-01 1.13757886e-01 -6.91596746e-01 1.17921793e+00 4.44684654e-01 -7.31312707e-02 4.73879725e-01 6.53950930e-01 -2.43685782e-01 -1.17535308e-01 -6.25837922e-01 -1.98707402e-01 -1.30174354e-01 1.72002181e-01 -7.09493309e-02 4.56883684e-02 -7.20535100e-01 4.47695732e-01 -1.30343204e-02 5.63678285e-03 2.81653702e-01 1.02103531e+00 -6.05997384e-01 -1.53362894e+00 -4.59922671e-01 2.50656098e-01 -4.68287230e-01 3.49595577e-01 4.61153127e-02 1.01467419e+00 1.97631478e-01 7.19234467e-01 -3.48333716e-01 -7.11704493e-02 3.37790996e-01 -1.30741715e-01 4.84543562e-01 -2.37635583e-01 -9.54474986e-01 -1.85275629e-01 2.26791695e-01 -4.75531429e-01 -5.63785098e-02 -7.89989710e-01 -1.19827688e+00 -7.65467510e-02 -3.54473144e-01 5.31961262e-01 9.06006575e-01 1.13651013e+00 -2.36138124e-02 6.83058083e-01 8.55277956e-01 -9.19090331e-01 -1.12151659e+00 -5.61580837e-01 -6.26611412e-01 -1.03592565e-02 -2.57833958e-01 -9.85468924e-01 -4.15071368e-01 -6.83911622e-01]
[5.04980993270874, 2.5197184085845947]
3950adef-aee6-46ab-9d66-cc004e544ec4
fast-low-rank-column-wise-compressive-sensing-1
2212.09664
null
https://arxiv.org/abs/2212.09664v1
https://arxiv.org/pdf/2212.09664v1.pdf
Fast Low Rank column-wise Compressive Sensing for Accelerated Dynamic MRI
This work develops a novel set of algorithms, alternating Gradient Descent (GD) and minimization for MRI (altGDmin-MRI1 and altGDmin-MRI2), for accelerated dynamic MRI by assuming an approximate low-rank (LR) model on the matrix formed by the vectorized images of the sequence. The LR model itself is well-known in the MRI literature; our contribution is the novel GD-based algorithms which are much faster, memory efficient, and general compared with existing work; and careful use of a 3-level hierarchical LR model. By general, we mean that, with a single choice of parameters, our method provides accurate reconstructions for multiple accelerated dynamic MRI applications, multiple sampling rates and sampling schemes. We show that our methods outperform many of the popular existing approaches while also being faster than all of them, on average. This claim is based on comparisons on 8 different retrospectively under sampled multi-coil dynamic MRI applications, sampled using either 1D Cartesian or 2D pseudo radial under sampling, at multiple sampling rates. Evaluations on some prospectively under sampled datasets are also provided. Our second contribution is a mini-batch subspace tracking extension that can process new measurements and return reconstructions within a short delay after they arrive. The recovery algorithm itself is also faster than its batch counterpart.
['Namrata Vaswani', 'Sajan Goud Lingala', 'Silpa Babu']
2022-12-19
null
null
null
null
['compressive-sensing']
['computer-vision']
[ 3.11360240e-01 -2.31317297e-01 -9.15513039e-02 -4.47234541e-01 -1.22015667e+00 -3.82367760e-01 3.31169724e-01 -1.73725665e-01 -6.43559635e-01 5.89543402e-01 6.50734425e-01 -3.93488735e-01 -5.25726914e-01 1.08650893e-01 -6.04380727e-01 -8.78092945e-01 -8.50114524e-01 7.21436560e-01 1.15501583e-01 -1.34677038e-01 -5.61885498e-02 6.07534170e-01 -6.40023470e-01 2.36035436e-01 4.88455653e-01 6.45891905e-01 5.19861698e-01 8.95080209e-01 4.68158424e-01 1.03547716e+00 1.19805492e-01 7.57786483e-02 5.95651150e-01 -4.67601866e-01 -9.97785270e-01 1.68118417e-01 2.57538855e-01 -6.88672185e-01 -6.49037957e-01 9.12790656e-01 1.12494600e+00 3.75378460e-01 4.09541249e-01 -5.43296099e-01 4.02845368e-02 6.02715194e-01 -8.71125102e-01 7.79202163e-01 2.70987421e-01 9.38933436e-03 2.09938601e-01 -1.24448705e+00 9.30532455e-01 6.44164801e-01 1.04852509e+00 5.09800971e-01 -1.47434354e+00 -3.00120592e-01 -6.26250431e-02 2.47508720e-01 -1.33702433e+00 -6.67651653e-01 5.80280066e-01 -5.53013980e-01 7.55751133e-01 5.10479748e-01 5.13049126e-01 7.14422047e-01 3.44336331e-01 6.73398018e-01 1.79634404e+00 -2.18719855e-01 2.23300993e-01 -2.96220630e-01 3.31587613e-01 5.60632944e-01 -2.82698385e-02 1.64663538e-01 -3.99030924e-01 -6.04625583e-01 8.05481374e-01 -1.35798827e-01 -7.23252237e-01 -7.07873106e-01 -1.73667860e+00 7.22221732e-01 9.11962688e-02 5.93364716e-01 -9.74912345e-01 -2.46383846e-02 8.14185917e-01 2.76730210e-01 5.44677973e-01 -7.24309981e-02 -2.84771398e-02 -8.75808820e-02 -1.53093112e+00 9.59641114e-02 7.50585556e-01 6.86854362e-01 7.99101740e-02 1.90335289e-01 -2.23943710e-01 8.72071505e-01 1.08618073e-01 6.15508139e-01 9.22097683e-01 -9.73423302e-01 1.43619582e-01 -6.22045457e-01 -1.57703519e-01 -9.97339904e-01 -9.49070930e-01 -8.34071577e-01 -1.26264775e+00 -1.80269122e-01 2.65517365e-02 -1.06814392e-01 -6.65178597e-01 1.60495806e+00 5.27935207e-01 5.75807571e-01 -3.10792685e-01 1.41427493e+00 7.75889635e-01 2.50350416e-01 -2.27332711e-01 -1.06413281e+00 1.25956583e+00 -9.36253369e-01 -1.08634794e+00 6.73027933e-02 6.28387928e-01 -8.04399073e-01 5.56264400e-01 6.46237969e-01 -1.36598492e+00 -2.70048410e-01 -9.18600321e-01 3.60649139e-01 4.09633398e-01 1.24775477e-01 7.70809710e-01 6.69454098e-01 -1.61611784e+00 8.41283381e-01 -1.25434172e+00 -1.85847608e-03 -1.14384405e-01 5.15132368e-01 -4.98083174e-01 -3.17343414e-01 -9.42523837e-01 1.14072168e+00 -7.46590719e-02 3.67183089e-01 -1.13387370e+00 -8.65246534e-01 -5.41923046e-01 -7.80748188e-01 8.71968344e-02 -5.68640411e-01 1.18591762e+00 -5.29773116e-01 -1.35110092e+00 8.40006828e-01 -1.48659706e-01 -6.41481221e-01 8.81095827e-01 -1.48283288e-01 -4.49320704e-01 6.19201720e-01 1.36300340e-01 3.00592512e-01 8.27872276e-01 -9.37544107e-01 3.94832343e-01 -5.11485577e-01 -5.03138959e-01 5.38121402e-01 1.52487442e-01 2.41421178e-01 -3.42095703e-01 -7.19698489e-01 7.62876928e-01 -1.32448435e+00 -8.48989666e-01 -2.70585477e-01 -3.05982918e-01 5.26462793e-01 2.13686749e-01 -1.26174068e+00 1.26420546e+00 -1.90843236e+00 3.18261474e-01 3.62830281e-01 5.25808036e-01 -6.06976524e-02 -1.86680369e-02 1.44881815e-01 -7.99848378e-01 -5.77055097e-01 -5.83007097e-01 -1.72893330e-01 -5.94217062e-01 1.10549226e-01 -1.47436276e-01 1.28040290e+00 -8.20759714e-01 6.18623853e-01 -1.11392593e+00 -3.54417473e-01 5.72262183e-02 4.73073661e-01 -4.40506518e-01 1.84661690e-02 8.26083720e-01 9.41508770e-01 -2.15723395e-01 2.59610504e-01 9.84967828e-01 -2.13113740e-01 4.99510825e-01 -8.73295009e-01 -7.87348524e-02 -6.31327555e-02 -1.46954775e+00 2.15888071e+00 -3.36035252e-01 3.28310907e-01 5.86224437e-01 -1.36262512e+00 3.08107734e-01 6.84391320e-01 1.39462519e+00 -3.86434585e-01 -1.46688655e-01 6.33274078e-01 6.01189174e-02 -5.42372882e-01 2.22536534e-01 -3.75540286e-01 4.62449551e-01 7.43545175e-01 -3.41811553e-02 2.39428878e-02 1.32814825e-01 5.67588687e-01 1.24645030e+00 2.84753740e-02 1.56664968e-01 -7.62973130e-01 5.14923453e-01 -1.66237995e-01 2.92938650e-01 1.15782893e+00 -3.82744670e-01 8.40366602e-01 -2.58850958e-02 -4.38057661e-01 -1.01011229e+00 -9.33357358e-01 -6.54249310e-01 6.68770254e-01 -2.18952864e-01 -2.80890077e-01 -6.34587824e-01 -5.61604023e-01 -5.17911553e-01 1.62428871e-01 -3.64013672e-01 2.87564546e-01 -1.04022276e+00 -1.51165485e+00 3.67266744e-01 3.07610661e-01 1.12135649e-01 -3.70719016e-01 -6.18069470e-01 6.61986887e-01 -5.70698798e-01 -1.31709576e+00 -6.30891860e-01 1.93229869e-01 -1.60993958e+00 -8.83226812e-01 -1.39049590e+00 -3.10780406e-01 7.69980192e-01 3.95067066e-01 1.07142460e+00 -1.81399465e-01 -3.86480808e-01 7.63377190e-01 -2.43092790e-01 3.42479408e-01 -2.38180071e-01 -3.79567564e-01 6.66852534e-01 2.65724566e-02 -4.25959229e-01 -8.39020967e-01 -9.06212687e-01 8.56108963e-02 -9.22758222e-01 2.38345757e-01 5.82932413e-01 1.12423182e+00 9.14044619e-01 -5.56934476e-01 2.62599409e-01 -1.08733678e+00 5.67659855e-01 -6.57032847e-01 -4.19389382e-02 6.11827185e-04 -7.95482874e-01 9.49047282e-02 1.47499993e-01 -4.94608045e-01 -7.43400455e-01 2.01064676e-01 -3.50950718e-01 -5.91976941e-01 3.07871580e-01 6.40915334e-01 7.38496363e-01 -5.43343484e-01 7.85679162e-01 6.62542880e-01 3.85594159e-01 -6.49652064e-01 4.80082095e-01 3.33728373e-01 5.94206989e-01 -5.15416086e-01 4.02420819e-01 8.63162637e-01 2.78354645e-01 -8.66149664e-01 -4.64054108e-01 -6.96256101e-01 -7.47057140e-01 -5.19965947e-01 5.08201480e-01 -9.13556635e-01 -1.33853137e-01 4.26522076e-01 -6.93212807e-01 -2.17834577e-01 -3.52083951e-01 1.16820776e+00 -8.03954542e-01 9.44953203e-01 -1.07377493e+00 -5.67798376e-01 -7.77571440e-01 -1.58015788e+00 8.47662270e-01 -5.71663439e-01 -3.58610861e-02 -9.46453214e-01 3.95892739e-01 1.41945764e-01 7.97904193e-01 3.86349410e-01 5.00739753e-01 -5.37143409e-01 -1.44943997e-01 1.20339934e-02 1.18097253e-01 3.45694184e-01 -1.34171098e-01 -1.03670597e+00 -5.03657281e-01 -9.58203733e-01 7.91298687e-01 -9.25536156e-02 5.63444078e-01 9.03712451e-01 1.07250154e+00 -2.51037598e-01 -1.51160464e-01 9.31304872e-01 1.40978134e+00 -1.73544362e-01 5.88712037e-01 1.88806996e-01 5.01598835e-01 3.86381477e-01 4.36165810e-01 4.68867719e-01 3.00980628e-01 7.98170984e-01 1.76520068e-02 -2.90248543e-01 -3.32872123e-01 4.62020636e-01 3.63978028e-01 1.73553455e+00 -3.73513758e-01 8.90242636e-01 -9.72176313e-01 5.83250701e-01 -1.69849062e+00 -7.19973385e-01 -5.30288279e-01 2.28505158e+00 9.80434716e-01 -4.60138679e-01 2.28258252e-01 5.29349707e-02 5.48906982e-01 1.99748501e-01 -4.80474949e-01 2.96176281e-02 -1.14452519e-01 4.35814947e-01 8.22378874e-01 5.25765181e-01 -9.80342388e-01 1.50617212e-01 7.29628181e+00 8.20212781e-01 -1.21210468e+00 9.39999878e-01 6.31944597e-01 -4.50407386e-01 -6.06843084e-02 -2.61283126e-02 -2.78307319e-01 1.33803174e-01 1.15886855e+00 -9.55813844e-03 6.20594501e-01 5.89656889e-01 3.58454883e-01 -1.92394286e-01 -7.90644825e-01 1.22337341e+00 3.11272532e-01 -1.23476923e+00 -3.60134512e-01 6.83317110e-02 6.80760086e-01 6.43776000e-01 -3.54034752e-02 -9.23733925e-04 -1.45068929e-01 -7.27901280e-01 4.53994334e-01 6.43013477e-01 1.08937621e+00 -2.83542931e-01 5.19192040e-01 2.64490843e-01 -9.15556431e-01 4.21365440e-01 -1.85484335e-01 3.68605465e-01 5.29509068e-01 9.52789247e-01 -5.33464611e-01 9.22817469e-01 7.62976408e-01 5.60983777e-01 -3.55750948e-01 1.13235056e+00 1.44375801e-01 7.21177518e-01 -4.18309689e-01 6.58250332e-01 1.52935147e-01 -3.12250108e-01 1.11087549e+00 1.48933625e+00 1.57627419e-01 4.61476326e-01 3.29357654e-01 1.91392958e-01 6.10803366e-01 2.94227630e-01 -4.10081059e-01 6.01241052e-01 -2.77749170e-02 1.56307948e+00 -7.00801373e-01 -4.71709251e-01 -2.62681693e-01 9.72838521e-01 7.99102057e-03 2.80342907e-01 -6.39438927e-01 4.53400388e-02 -3.36179044e-03 1.79815933e-01 1.04354452e-02 -4.50043797e-01 -1.65769141e-02 -1.39277661e+00 2.64008231e-02 -1.20943403e+00 5.80519617e-01 -6.04993641e-01 -1.19800854e+00 8.04325283e-01 3.66993189e-01 -1.31801963e+00 -6.06792450e-01 -2.72412747e-01 2.17441050e-03 8.93152773e-01 -1.23233545e+00 -5.88188231e-01 -3.58808460e-03 9.52113807e-01 3.46847117e-01 1.91835433e-01 7.38139808e-01 7.65180647e-01 -2.07606018e-01 2.74000615e-01 3.80008698e-01 -9.54926014e-02 4.47450519e-01 -1.07357979e+00 -1.04454234e-01 9.78245854e-01 -5.55505827e-02 8.70312989e-01 8.53513837e-01 -6.39537930e-01 -1.86571908e+00 -6.58746064e-01 4.35602903e-01 -6.19859211e-02 6.17266476e-01 -7.27212280e-02 -8.98772478e-01 6.45462215e-01 1.77285045e-01 4.14210171e-01 7.40297854e-01 -3.47147435e-01 2.44330317e-01 -1.25925750e-01 -1.28586233e+00 2.41916329e-01 1.13311815e+00 -4.80675817e-01 -5.58730364e-01 9.43724036e-01 1.05365887e-01 -1.09599066e+00 -1.29357922e+00 6.11939192e-01 4.86603290e-01 -7.33152986e-01 1.21192086e+00 -3.38295639e-01 -1.24251910e-01 -3.81439984e-01 -1.81650802e-01 -1.25628185e+00 -4.33237612e-01 -8.49662483e-01 -4.58051980e-01 3.38189214e-01 1.46506634e-02 -5.06949544e-01 4.52497989e-01 3.94062847e-01 -2.80535638e-01 -8.58958602e-01 -1.31868267e+00 -8.11486125e-01 -6.43258095e-02 -6.50734305e-01 -5.40887974e-02 1.05714750e+00 -1.13555521e-01 5.46212010e-02 -7.23688066e-01 1.29579142e-01 1.33105838e+00 6.50618505e-03 2.16485277e-01 -4.28789049e-01 -7.20770776e-01 -8.50328058e-03 -2.47834280e-01 -1.07302260e+00 -2.39932358e-01 -1.26615453e+00 -1.61488682e-01 -1.23234916e+00 3.98217738e-01 -4.80515540e-01 -3.63275945e-01 1.08158454e-01 6.32480234e-02 6.24183238e-01 4.10273187e-02 7.09781170e-01 -4.53355998e-01 1.75815150e-01 1.28239846e+00 1.54522702e-01 1.09636202e-01 -1.48550555e-01 -3.32923174e-01 5.31132996e-01 1.29712194e-01 -5.18521130e-01 -3.34492087e-01 -3.95574689e-01 -5.99275343e-02 7.11464763e-01 2.55653590e-01 -1.02762973e+00 5.43537498e-01 4.99004632e-01 2.37244248e-01 -7.34067261e-01 1.98237821e-01 -4.98782426e-01 5.86127400e-01 8.22068453e-01 -2.89765924e-01 4.07104105e-01 -8.69709533e-03 3.27262789e-01 -1.04869328e-01 -3.32153618e-01 8.66812706e-01 -4.84769732e-01 -5.12896299e-01 4.78734553e-01 -4.79821742e-01 1.06350519e-01 6.08422220e-01 4.95152548e-02 4.61477637e-01 -5.16850412e-01 -1.44039130e+00 3.14954808e-03 -1.27028868e-01 -7.55742341e-02 7.97885835e-01 -1.33181012e+00 -1.10930359e+00 4.72389422e-02 -4.55487043e-01 -4.79891092e-01 8.84788096e-01 2.13160276e+00 -5.63101232e-01 2.32645303e-01 4.03426141e-02 -1.17764413e+00 -9.71468627e-01 4.74792331e-01 6.35701656e-01 -7.66690135e-01 -1.12433779e+00 6.05017722e-01 7.68253626e-03 -5.98627269e-01 -5.97411133e-02 2.49365434e-01 -6.66418821e-02 -1.62792981e-01 7.85465240e-01 5.88800848e-01 5.90657353e-01 -9.27639782e-01 -5.52785397e-01 6.19354606e-01 -1.79000959e-01 -6.20543718e-01 1.56061494e+00 -3.37925911e-01 -3.59323204e-01 3.17370266e-01 1.21561813e+00 -4.74354587e-02 -8.71172726e-01 -4.81401414e-01 -2.72041932e-02 -2.66570687e-01 6.05099320e-01 -6.73849940e-01 -1.25887966e+00 5.63962042e-01 1.25232601e+00 -2.61859804e-01 1.28496933e+00 -2.88235635e-01 7.11862087e-01 2.85447435e-03 7.31148601e-01 -7.43445277e-01 -3.88805091e-01 3.24220181e-01 1.17717564e+00 -8.70149136e-01 5.41866481e-01 -1.79297492e-01 -6.81204855e-01 1.05401754e+00 -1.17148116e-01 -3.00978214e-01 7.29843080e-01 4.88901615e-01 1.61323324e-02 -3.69008124e-01 -3.76882285e-01 1.80072382e-01 4.48247015e-01 4.94937152e-01 7.03367412e-01 1.29625142e-01 -1.03673196e+00 3.02526593e-01 3.49633358e-02 1.25962242e-01 5.52684784e-01 1.03273964e+00 1.59601659e-01 -9.72060084e-01 -5.20990789e-01 5.64561784e-01 -8.51142883e-01 -3.09702873e-01 4.84927684e-01 5.65572262e-01 -3.63644034e-01 4.96502012e-01 -5.84032178e-01 -8.80879462e-02 3.54277432e-01 -3.21057290e-01 8.81230414e-01 -3.40238839e-01 -7.13714838e-01 3.70864749e-01 8.11358169e-02 -1.04275668e+00 -7.34938622e-01 -1.18189180e+00 -1.03282964e+00 3.79218534e-02 -2.28659287e-01 3.44073296e-01 9.30992126e-01 7.31268167e-01 2.67334461e-01 3.28769743e-01 8.94051135e-01 -9.93906736e-01 -1.04653060e+00 -9.78292465e-01 -8.15921843e-01 4.68330741e-01 2.94879466e-01 -5.19919157e-01 -2.45492816e-01 -1.95851907e-01]
[13.419368743896484, -2.389461040496826]
28e7ced3-7104-4445-8f80-19591bca822f
transfer-learning-for-the-efficient-detection
2307.02975
null
https://arxiv.org/abs/2307.02975v1
https://arxiv.org/pdf/2307.02975v1.pdf
Transfer Learning for the Efficient Detection of COVID-19 from Smartphone Audio Data
Disease detection from smartphone data represents an open research challenge in mobile health (m-health) systems. COVID-19 and its respiratory symptoms are an important case study in this area and their early detection is a potential real instrument to counteract the pandemic situation. The efficacy of this solution mainly depends on the performances of AI algorithms applied to the collected data and their possible implementation directly on the users' mobile devices. Considering these issues, and the limited amount of available data, in this paper we present the experimental evaluation of 3 different deep learning models, compared also with hand-crafted features, and of two main approaches of transfer learning in the considered scenario: both feature extraction and fine-tuning. Specifically, we considered VGGish, YAMNET, and L\textsuperscript{3}-Net (including 12 different configurations) evaluated through user-independent experiments on 4 different datasets (13,447 samples in total). Results clearly show the advantages of L\textsuperscript{3}-Net in all the experimental settings as it overcomes the other solutions by 12.3\% in terms of Precision-Recall AUC as features extractor, and by 10\% when the model is fine-tuned. Moreover, we note that to fine-tune only the fully-connected layers of the pre-trained models generally leads to worse performances, with an average drop of 6.6\% with respect to feature extraction. %highlighting the need for further investigations. Finally, we evaluate the memory footprints of the different models for their possible applications on commercial mobile devices.
['Elena Pagani', 'Franca Delmastro', 'Mattia Giovanni Campana']
2023-07-06
null
null
null
null
['transfer-learning']
['miscellaneous']
[ 3.74251574e-01 8.44169259e-02 -6.72214702e-02 5.86248413e-02 -6.07505083e-01 -2.34483153e-01 6.31765246e-01 4.68945682e-01 -1.05746078e+00 7.89969563e-01 -1.52805150e-01 -4.20212984e-01 -4.51874435e-01 -8.48282576e-01 -6.23350382e-01 -7.21607625e-01 -2.53552407e-01 7.45520711e-01 1.37398064e-01 -1.58756614e-01 -1.25490785e-01 5.65170050e-01 -1.61556625e+00 3.39946508e-01 7.69253790e-01 9.47799146e-01 1.46771759e-01 8.70657444e-01 1.87098786e-01 4.83164608e-01 -8.39119852e-01 -2.32680023e-01 -1.10250890e-01 2.69271154e-02 -5.74794114e-01 -3.29028308e-01 6.97956532e-02 -2.16968879e-01 3.37130547e-01 5.79764783e-01 7.65527427e-01 -1.39893845e-01 6.81074083e-01 -8.20241928e-01 9.63960961e-02 3.22219789e-01 -1.38236329e-01 4.35659021e-01 1.68666869e-01 3.01170945e-01 4.73889321e-01 -5.90616703e-01 5.30805945e-01 7.30829298e-01 1.11488307e+00 4.30318832e-01 -1.03914380e+00 -4.33184832e-01 -2.78817475e-01 2.73740321e-01 -1.50049865e+00 -3.76672655e-01 2.16125399e-01 -6.82456434e-01 1.24223650e+00 4.57214177e-01 5.25822222e-01 1.34813988e+00 1.75984219e-01 4.44108814e-01 1.10023952e+00 -3.49428713e-01 2.35624894e-01 6.24479592e-01 1.73191145e-01 5.58809519e-01 5.58549881e-01 -3.78818922e-02 1.09411469e-02 -4.31037843e-01 3.47545564e-01 1.56987190e-01 -1.75751895e-01 -1.53492845e-03 -1.01339316e+00 8.31219912e-01 3.34101230e-01 8.97397280e-01 -7.34538853e-01 -1.30379483e-01 5.77242613e-01 6.74689710e-02 5.62096000e-01 5.58316231e-01 -7.90786386e-01 -9.14620683e-02 -9.83827174e-01 1.35679826e-01 6.27919495e-01 3.43553007e-01 4.55918670e-01 -9.27124098e-02 -3.86516094e-01 6.06867552e-01 2.79779397e-02 6.45364285e-01 7.41767168e-01 -8.98976624e-02 4.22849119e-01 7.15069532e-01 1.34909190e-02 -1.07134235e+00 -1.02079642e+00 -7.75999784e-01 -1.02404082e+00 -4.38339084e-01 4.62056637e-01 -7.41238296e-01 -6.10805154e-01 1.43419135e+00 3.10158879e-01 4.08086777e-01 2.13830676e-02 3.09989303e-01 8.11587870e-01 5.88401139e-01 3.19746494e-01 -2.07588732e-01 1.55971408e+00 -5.87978899e-01 -5.45615911e-01 5.75425336e-03 8.20740759e-01 -4.89150941e-01 9.68817353e-01 4.94739443e-01 -7.01870859e-01 -6.47721350e-01 -9.64424253e-01 4.14166510e-01 -8.76107633e-01 4.69113857e-01 2.25946948e-01 9.42149460e-01 -1.11681676e+00 5.89299142e-01 -7.81012654e-01 -7.09833562e-01 4.10120070e-01 8.46492052e-01 -9.92187187e-02 2.39638150e-01 -1.30483365e+00 9.19800103e-01 4.11749154e-01 1.22172728e-01 -5.06105602e-01 -5.53245783e-01 -3.81515592e-01 1.74993843e-01 3.48912716e-01 -8.98383737e-01 8.92413557e-01 -6.13943875e-01 -1.28143477e+00 9.08968806e-01 1.66887954e-01 -8.34619462e-01 7.36548722e-01 -4.38339949e-01 -5.38122237e-01 1.47803828e-01 -2.39963636e-01 3.81213874e-01 7.23898590e-01 -8.30233097e-01 -6.94095314e-01 -3.75046223e-01 -2.17121184e-01 -1.82586327e-01 -6.65904939e-01 -9.13208500e-02 -3.51495028e-01 -4.89095688e-01 -8.26854229e-01 -1.00497210e+00 -2.98758209e-01 -5.62290609e-01 -3.99815470e-01 -1.15122542e-01 5.93923390e-01 -7.75323331e-01 1.35927486e+00 -1.74617684e+00 -8.50102082e-02 4.14504498e-01 1.73210934e-01 1.09136569e+00 8.83185416e-02 4.30638075e-01 1.53011844e-01 -5.35500161e-02 -2.54344970e-01 -8.61463398e-02 -4.37471837e-01 1.88661683e-02 2.80108601e-01 3.68392706e-01 1.69222340e-01 9.70148683e-01 -6.68417454e-01 -4.17599946e-01 5.57779670e-01 1.05334282e+00 -4.53184813e-01 1.56086922e-01 -2.09004551e-01 4.78351802e-01 -5.06255746e-01 2.08946183e-01 4.65115249e-01 -3.87205511e-01 4.69230056e-01 -1.21851541e-01 6.54266775e-03 5.19093014e-02 -8.54530156e-01 1.06770968e+00 -7.76063800e-01 2.65412778e-01 -2.59179711e-01 -1.12035024e+00 7.17719436e-01 5.49782455e-01 7.12714076e-01 -6.33808136e-01 5.14446318e-01 2.29304954e-01 1.49318650e-02 -8.62789392e-01 2.25678295e-01 2.18435824e-01 3.40813361e-02 1.45488396e-01 1.25113472e-01 6.29343390e-01 7.76273757e-02 -4.33086097e-01 9.91950512e-01 -2.74107873e-01 6.02024138e-01 -3.29874933e-01 8.83998811e-01 -1.69466764e-01 -1.44855743e-02 8.20477962e-01 5.23403697e-02 2.51668751e-01 3.68310571e-01 -5.12575209e-01 -4.64494258e-01 -5.04570425e-01 -2.03172475e-01 1.07720327e+00 -4.80381638e-01 -2.20048726e-01 -1.22749531e+00 -9.00259316e-01 -2.12977886e-01 6.12995684e-01 -9.19163406e-01 -9.30166692e-02 -6.46793187e-01 -1.42552316e+00 8.32093775e-01 2.69708097e-01 3.99432957e-01 -1.42622542e+00 -1.22505522e+00 2.93324769e-01 -1.27538845e-01 -1.01511419e+00 1.64582163e-01 3.58958304e-01 -9.02188301e-01 -1.27018106e+00 -8.01221728e-01 -4.48809177e-01 3.58508587e-01 -1.26991808e-01 1.03156519e+00 3.08826983e-01 -2.61177570e-01 3.16422939e-01 -3.33253950e-01 -6.76382422e-01 -5.56861997e-01 7.69207001e-01 -1.88257575e-01 1.38785586e-01 6.88307226e-01 -2.18070060e-01 -6.86995685e-01 5.47905266e-02 -8.31116319e-01 -2.84908921e-01 6.78493500e-01 7.04759955e-01 3.98567080e-01 -4.93578464e-02 6.70905828e-01 -1.32532728e+00 5.96096873e-01 -7.53693879e-01 -4.62301612e-01 1.96780086e-01 -7.29273677e-01 -1.79556787e-01 6.81205332e-01 -2.91193217e-01 -7.42243230e-01 3.88767831e-02 -7.01051474e-01 9.41561721e-03 -5.19481182e-01 3.12229067e-01 7.88057297e-02 2.03628346e-01 9.23411191e-01 8.27339068e-02 -4.21410389e-02 -6.67306900e-01 4.65532951e-02 9.00386393e-01 1.16949029e-01 -7.90933818e-02 3.90806377e-01 4.45268095e-01 -5.46262600e-02 -1.25486398e+00 -6.07504487e-01 -4.74681854e-01 -4.25314039e-01 -2.19130004e-03 1.07824159e+00 -7.92616248e-01 -8.83373737e-01 4.63787913e-01 -1.01534629e+00 -3.48275810e-01 -1.39407128e-01 3.76842588e-01 -3.16883087e-01 5.28007708e-02 -5.35782158e-01 -7.14945912e-01 -9.54604685e-01 -1.10429263e+00 1.00011826e+00 1.02026515e-01 -3.98241132e-01 -1.10222387e+00 1.19803458e-01 3.53028208e-01 6.64641857e-01 3.52321297e-01 1.02469385e+00 -1.10718131e+00 1.49749443e-01 -3.69401842e-01 -2.04039022e-01 3.58537793e-01 2.10132614e-01 -2.34561741e-01 -1.36466026e+00 -3.66086930e-01 3.53334285e-02 1.07801870e-01 8.68209898e-01 4.90600586e-01 1.08672643e+00 -3.85394365e-01 -5.45150638e-01 3.43432635e-01 1.35466182e+00 3.24089348e-01 5.06316125e-01 4.19404387e-01 6.68468058e-01 5.22040367e-01 2.28846565e-01 4.58101541e-01 2.21406937e-01 8.46753955e-01 3.85140955e-01 -4.72444624e-01 2.18987539e-02 1.05072804e-01 1.86539575e-01 6.58180892e-01 -3.61557245e-01 -4.22610968e-01 -9.93342459e-01 4.95303601e-01 -1.62942886e+00 -7.80859709e-01 -2.09830120e-01 2.43821120e+00 5.00985503e-01 2.14651853e-01 5.17005026e-01 3.93267840e-01 6.07584774e-01 -1.59555465e-01 -2.33177587e-01 -5.01635849e-01 1.29126564e-01 6.53497100e-01 4.17793781e-01 2.32490376e-01 -1.28310978e+00 3.57628822e-01 5.49736834e+00 8.23308229e-01 -1.57747972e+00 3.69151920e-01 7.21580982e-01 -6.05270937e-02 2.81555206e-01 -7.46202111e-01 -8.23330641e-01 6.83527827e-01 1.41918051e+00 5.31474948e-01 2.49316096e-01 6.58025146e-01 3.01077157e-01 -2.52066627e-02 -7.58219898e-01 8.41556132e-01 -5.62578663e-02 -1.28495920e+00 -1.42661169e-01 1.25371858e-01 5.12540817e-01 3.35307032e-01 2.85174418e-02 4.47180957e-01 -4.20799434e-01 -9.43767607e-01 1.91222772e-01 4.27081615e-01 8.48078370e-01 -8.60629141e-01 1.22251487e+00 3.72886330e-01 -8.29920828e-01 -1.48855776e-01 1.02262706e-01 1.44604877e-01 -4.54408936e-02 7.30377853e-01 -1.43443441e+00 6.42153680e-01 7.09108591e-01 3.19181025e-01 -8.64354372e-01 8.39418232e-01 2.22723514e-01 8.65761936e-01 -4.26525950e-01 -3.07224452e-01 1.80381969e-01 2.46453255e-01 2.61152804e-01 1.77980268e+00 4.89970624e-01 -4.39746261e-01 -2.14384079e-01 3.82542908e-01 9.77732688e-02 4.53794032e-01 -4.93678272e-01 7.90104493e-02 7.20588192e-02 1.25471270e+00 -8.12382281e-01 -5.53172648e-01 -3.20390105e-01 7.60209858e-01 2.81319432e-02 1.60678685e-01 -9.17900085e-01 -4.36842889e-01 3.56435001e-01 5.79068422e-01 6.32269442e-01 2.95260549e-01 -2.78250519e-02 -8.29855800e-01 -2.34151557e-01 -9.99355435e-01 5.30329108e-01 -1.68408155e-01 -9.04272497e-01 9.10690188e-01 5.66287786e-02 -9.18250859e-01 -5.33819079e-01 -8.00441861e-01 -3.39131415e-01 6.56432390e-01 -1.37649751e+00 -8.65792096e-01 -2.82007933e-01 5.33638418e-01 2.91399121e-01 -4.64249998e-02 1.14814878e+00 6.77273989e-01 -6.16013646e-01 6.86433613e-01 2.94775873e-01 -2.01599166e-01 2.38992840e-01 -8.76621306e-01 3.90004009e-01 3.07365388e-01 8.10019076e-02 5.51559567e-01 4.78923649e-01 -4.21630383e-01 -9.92791235e-01 -1.20832264e+00 1.12116933e+00 -5.21853805e-01 1.62223846e-01 -3.32763821e-01 -9.67010617e-01 3.59195024e-01 1.15083508e-01 -3.23186845e-01 7.29669452e-01 5.15069999e-02 2.16955513e-01 -3.78668718e-02 -1.37014735e+00 4.44782645e-01 6.19759083e-01 -1.93436280e-01 -2.28221759e-01 3.61577272e-01 4.70307916e-01 -1.21771045e-01 -8.58247459e-01 6.58206582e-01 5.94065309e-01 -1.26695764e+00 1.03596234e+00 -4.18793201e-01 -4.91379015e-02 -4.18275520e-02 1.97173774e-01 -1.04017365e+00 1.14879515e-02 -3.36313277e-01 -1.77995056e-01 9.43741739e-01 4.55502450e-01 -7.30051160e-01 7.31034577e-01 -9.63226184e-02 4.48912501e-01 -1.16152656e+00 -7.33509541e-01 -4.05951768e-01 -2.03259781e-01 -5.01296401e-01 6.25137925e-01 9.29215193e-01 -4.48739052e-01 4.54377174e-01 -4.30022240e-01 -1.20685190e-01 -8.01606029e-02 -2.97724068e-01 7.33817160e-01 -1.35607684e+00 -5.67757726e-01 -3.20993721e-01 -2.00857103e-01 -3.90633404e-01 -3.69068414e-01 -7.04746485e-01 -2.89826989e-01 -1.25179768e+00 -7.05788806e-02 -4.67223138e-01 -5.88652253e-01 4.06305462e-01 -2.27048069e-01 3.98346663e-01 6.28164932e-02 -6.58757910e-02 -3.20737004e-01 2.71175965e-03 6.50759637e-01 3.53360474e-02 -5.67707837e-01 4.99780983e-01 -4.37095344e-01 7.69350350e-01 9.71731484e-01 -5.44368863e-01 -2.64507562e-01 -1.97160691e-01 1.82596892e-01 -8.23658779e-02 3.98561090e-01 -1.20699275e+00 -2.25947872e-01 2.63809949e-01 4.78307992e-01 -4.01837975e-01 2.44525805e-01 -9.50210690e-01 3.32715929e-01 9.51222897e-01 -1.03918374e-01 1.15839899e-01 4.50721115e-01 3.08086187e-01 2.66796678e-01 -2.44483143e-01 5.85756540e-01 -4.70184721e-02 -3.44628006e-01 -8.59495848e-02 -7.04511881e-01 -6.94577619e-02 9.26997244e-01 -1.44688889e-01 -1.14883400e-01 -8.83734822e-02 -6.69203639e-01 -2.55102962e-01 -4.22593392e-02 3.78614336e-01 1.22803181e-01 -6.36484563e-01 -5.48243105e-01 2.92833209e-01 8.61929804e-02 -2.76354492e-01 5.12024701e-01 1.12164152e+00 -5.71909308e-01 7.32249260e-01 -1.83470502e-01 -6.50058448e-01 -1.26724494e+00 7.98491418e-01 2.85159886e-01 -7.84520864e-01 -2.84467816e-01 3.62532943e-01 -8.63887146e-02 -5.29800057e-01 3.01177412e-01 -4.11262065e-01 -6.44420087e-01 3.83044153e-01 4.23708290e-01 8.27394068e-01 6.97941124e-01 -6.02181494e-01 -6.17860794e-01 5.68484545e-01 9.75126494e-03 3.44783723e-01 1.33092093e+00 2.80778259e-01 2.00480178e-01 2.52902776e-01 1.19720030e+00 -8.92586727e-03 -5.45801401e-01 -4.99232225e-02 1.66303858e-01 3.14990401e-01 -2.25725845e-01 -8.34855199e-01 -9.84800160e-01 9.59933162e-01 1.13851142e+00 2.40947053e-01 1.20263934e+00 -2.57165194e-01 6.53767705e-01 4.59850252e-01 1.30777016e-01 -7.52075493e-01 -2.75780410e-01 1.87397256e-01 4.54842061e-01 -1.04580164e+00 -4.99255396e-02 -1.23621538e-01 -3.93289536e-01 7.69553900e-01 2.95293659e-01 4.34587337e-02 7.03799367e-01 4.00933661e-02 1.36875242e-01 -3.37341458e-01 -4.53223586e-01 -3.12760830e-01 3.88963401e-01 7.24179387e-01 5.13696969e-01 1.60708666e-01 -4.77825344e-01 6.51937008e-01 3.57065611e-02 3.59547466e-01 6.29749745e-02 6.31427526e-01 -1.64776161e-01 -1.06696248e+00 -3.76259804e-01 7.94452608e-01 -8.31934512e-01 -7.16656446e-02 -2.31946394e-01 1.02279413e+00 6.99823320e-01 9.44029272e-01 -1.59658596e-01 -4.86690760e-01 5.02589643e-01 9.17845592e-02 2.99578696e-01 -4.98491019e-01 -1.27400029e+00 -3.82374525e-02 6.49850592e-02 -2.61429459e-01 -4.77986991e-01 -6.46367669e-01 -9.98230934e-01 -4.34400961e-02 -2.73434132e-01 2.46741474e-02 6.87819779e-01 9.80776846e-01 6.67395592e-01 7.00401187e-01 2.92579353e-01 -7.31578231e-01 -4.35102046e-01 -1.05258858e+00 -8.88895392e-02 2.35547394e-01 3.30059320e-01 -4.65091467e-01 -9.11937356e-02 -3.20671909e-02]
[15.462869644165039, -1.7567425966262817]
b0557f88-ecff-4f05-b2f3-7b4b7a68fdfb
relation-aware-subgraph-embedding-with-co
2307.01507
null
https://arxiv.org/abs/2307.01507v1
https://arxiv.org/pdf/2307.01507v1.pdf
Relation-aware subgraph embedding with co-contrastive learning for drug-drug interaction prediction
Relation-aware subgraph embedding is promising for predicting multi-relational drug-drug interactions (DDIs). Typically, most existing methods begin by constructing a multi-relational DDI graph and then learning relation-aware subgraph embeddings (RaSEs) of drugs from the DDI graph. However, most existing approaches are usually limited in learning RaSEs of new drugs, leading to serious over-fitting when the test DDIs involve such drugs. To alleviate this issue, We propose a novel DDI prediction method based on relation-aware subgraph embedding with co-contrastive learning, RaSECo. RaSECo constructs two heterogeneous drug graphs: a multi-relational DDI graph and a multi-attributes-based drug-drug similarity (DDS) graph. The two graphs are used respectively for learning and propagating the RaSEs of drugs, thereby ensuring that all drugs, including new ones, can aggregate effective RaSEs. Additionally, we employ a cross-view contrastive mechanism to enhance drug-pair (DP) embedding. RaSECo learns DP embeddings from two distinct views (interaction and similarity views) and encourages these views to supervise each other collaboratively to obtain more discriminative DP embeddings. We evaluate the effectiveness of our RaSECo on three different tasks using two real datasets. The experimental results demonstrate that RaSECo outperforms existing state-of-the-art prediction methods.
['Weiqiang Jin', 'Yuanchao Su', 'Biao Zhao', 'Guizhong Liu', 'Mengying Jiang']
2023-07-04
null
null
null
null
['contrastive-learning', 'contrastive-learning']
['computer-vision', 'methodology']
[ 3.29692475e-02 6.42060637e-02 -7.16909945e-01 -2.80351937e-01 -3.94180119e-01 -4.21174109e-01 6.95470870e-01 6.52852595e-01 3.12677622e-01 6.18349910e-01 4.74684358e-01 -2.43521824e-01 -5.00357568e-01 -9.02158856e-01 -7.00266242e-01 -8.74388158e-01 -1.21882655e-01 6.74325764e-01 1.32534519e-01 -4.71518822e-02 -2.82461778e-03 6.97759509e-01 -8.65378320e-01 3.27622503e-01 9.23418880e-01 3.66186827e-01 -5.83164021e-02 8.95401388e-02 -1.84649357e-03 7.93762386e-01 -3.36577982e-01 -3.77286732e-01 -3.03815156e-02 -3.31423134e-01 -5.48841178e-01 -2.40298435e-01 1.36612192e-01 -7.14546591e-02 -7.38622785e-01 8.46276522e-01 6.43787384e-01 -8.87018517e-02 9.24000025e-01 -1.41385734e+00 -1.02374709e+00 4.03413236e-01 -6.90994680e-01 1.68792754e-01 6.95724249e-01 1.39959931e-01 1.41111588e+00 -1.20683229e+00 9.44852531e-01 1.38516462e+00 4.29283410e-01 5.05363822e-01 -1.17073750e+00 -7.41723597e-01 2.15284497e-01 3.83776963e-01 -1.43013883e+00 1.30962640e-01 8.47842634e-01 -5.21637917e-01 1.21420753e+00 3.06973696e-01 6.38634324e-01 1.19542277e+00 5.65468848e-01 5.85322201e-01 9.21340823e-01 3.52379113e-01 9.14138034e-02 2.54955649e-01 1.32130161e-01 6.44132197e-01 4.13537681e-01 1.35156840e-01 -5.14083982e-01 -6.76158965e-01 6.40238881e-01 6.79360271e-01 -5.80328405e-01 -7.33646512e-01 -1.13006997e+00 9.56237912e-01 7.22727358e-01 3.48613560e-01 -4.39837396e-01 -3.18207234e-01 3.72842312e-01 1.93366304e-01 5.38890004e-01 4.27095383e-01 -5.84754705e-01 3.59492987e-01 -4.00409624e-02 1.96902454e-01 7.72179425e-01 7.57812917e-01 5.57462454e-01 -5.06028891e-01 -5.44730313e-02 8.99913192e-01 6.70177758e-01 -8.80124941e-02 3.69398385e-01 6.66396245e-02 6.08919919e-01 1.22079301e+00 -3.73871028e-01 -1.18516374e+00 -2.42129013e-01 -2.75061190e-01 -8.76354039e-01 -1.79737821e-01 -2.20008448e-01 1.39318869e-01 -7.74592221e-01 1.38243461e+00 7.40502656e-01 5.71182787e-01 3.17650884e-02 5.48303008e-01 1.11796582e+00 7.45593965e-01 3.47646743e-01 -2.09127933e-01 1.15965176e+00 -8.60944211e-01 -6.65592909e-01 2.19233051e-01 8.83882344e-01 -5.81506968e-01 5.95320344e-01 1.83998078e-01 -5.58820248e-01 -3.20291042e-01 -1.14528167e+00 1.79403841e-01 -5.87267935e-01 -2.83701092e-01 4.87574399e-01 1.56175777e-01 -5.02233922e-01 8.09176326e-01 -5.49445629e-01 -5.33709936e-02 7.87832320e-01 6.36688113e-01 -7.54903495e-01 -5.45706093e-01 -1.39790535e+00 9.75504041e-01 1.34339243e-01 -2.89078981e-01 -9.92815435e-01 -1.21018684e+00 -1.10955977e+00 -1.54949911e-02 3.42280477e-01 -6.18650496e-01 2.80092597e-01 -2.67819524e-01 -1.19464684e+00 5.85649133e-01 9.71452445e-02 5.28023653e-02 1.47557870e-01 9.63134598e-03 -6.70019567e-01 5.34675922e-03 -3.33832130e-02 2.50807613e-01 4.49704349e-01 -1.41161489e+00 -1.63017973e-01 -7.94991553e-01 1.34437785e-01 4.36204255e-01 -3.89535695e-01 -3.07394773e-01 -3.62944514e-01 -5.54873526e-01 -9.72861424e-02 -8.04002762e-01 -2.29664668e-01 -9.20033827e-03 -6.22452319e-01 -7.18962193e-01 1.04083741e+00 -4.22522277e-01 1.45711458e+00 -1.98483610e+00 6.94810987e-01 3.67887199e-01 9.78038549e-01 4.27251279e-01 -4.29091036e-01 8.87851715e-01 -7.98012674e-01 9.82174054e-02 -3.73721570e-02 1.36237308e-01 -4.77979422e-01 4.23640251e-01 -2.60861311e-03 6.53275967e-01 3.61083806e-01 8.00692320e-01 -1.30151832e+00 -3.89964283e-01 3.20285052e-01 7.82765985e-01 -4.49893981e-01 4.78619754e-01 -2.05097422e-01 4.36429799e-01 -8.08774531e-01 6.94718838e-01 7.96014130e-01 -6.09442651e-01 8.24165165e-01 -5.30481875e-01 3.20571959e-01 1.80689201e-01 -9.49331820e-01 1.17166328e+00 -1.62112013e-01 -1.99067712e-01 -7.96135783e-01 -1.12809515e+00 8.43796492e-01 5.00945985e-01 1.02634680e+00 -3.24682444e-01 -8.88678282e-02 1.68838382e-01 1.77764833e-01 -4.53934580e-01 -3.86283755e-01 3.09729129e-02 4.69924390e-01 1.86320141e-01 2.03186885e-01 1.68846950e-01 -1.83369473e-01 3.75733614e-01 1.17374253e+00 1.35296509e-01 6.72891200e-01 -1.43491961e-02 7.18952715e-01 -3.67813528e-01 6.29597843e-01 -2.40040183e-01 -2.37775445e-02 2.02206567e-01 7.76424468e-01 -6.21220291e-01 -7.57337451e-01 -1.08573091e+00 -2.89555877e-01 5.63881218e-01 2.99202144e-01 -7.77274549e-01 -3.33448887e-01 -1.43116558e+00 3.98242652e-01 1.82020247e-01 -7.93511271e-01 -1.86177000e-01 -3.51940930e-01 -8.71087611e-01 -1.31931484e-01 4.68132019e-01 -2.19807327e-01 -6.10224724e-01 5.79247177e-01 4.49760020e-01 4.32428688e-01 -6.83277667e-01 -8.14123571e-01 -5.99976331e-02 -7.71303952e-01 -1.73234451e+00 -6.40310824e-01 -8.75577986e-01 6.99599862e-01 4.42589134e-01 9.10417080e-01 6.91088736e-02 -3.56226861e-01 8.54454562e-02 -1.90059692e-01 -1.55592814e-01 -4.02656734e-01 -3.19695204e-01 1.33600324e-01 1.28701672e-01 7.67018378e-01 -8.34534049e-01 -8.16854119e-01 5.23579538e-01 -8.87361109e-01 -1.81398571e-01 4.53678191e-01 8.90253365e-01 9.86174881e-01 -2.78484374e-01 6.92035615e-01 -1.32615006e+00 6.40230596e-01 -9.43877757e-01 -4.89315271e-01 4.45516020e-01 -9.24478412e-01 5.97145148e-02 8.37399364e-01 -5.34657896e-01 -4.22546327e-01 2.40101442e-02 -1.30085513e-01 -6.65460408e-01 1.83207557e-01 7.54319429e-01 -4.02419060e-01 -1.92477852e-01 5.31604469e-01 1.44586682e-01 -3.29365651e-03 -2.96541959e-01 4.52549696e-01 8.04436982e-01 -3.07930171e-01 -1.95270613e-01 5.29335141e-01 3.14790197e-02 1.36586964e-01 -5.51000774e-01 -4.95177686e-01 -5.92745066e-01 -2.63053089e-01 1.70293689e-01 7.88792789e-01 -1.03875339e+00 -9.21338856e-01 1.63703501e-01 -1.01049912e+00 3.14298064e-01 -4.08376865e-02 6.39406800e-01 -1.82355344e-01 5.26898324e-01 -6.24029577e-01 -6.41269386e-02 -2.76628524e-01 -1.43850029e+00 8.45494390e-01 2.47771293e-02 -2.10360959e-01 -1.25567472e+00 7.74837375e-01 2.94849366e-01 -1.72059834e-01 5.20046592e-01 1.32371938e+00 -1.24108052e+00 -6.80318594e-01 -1.65214568e-01 -2.75356919e-01 1.20300375e-01 7.12717474e-01 1.22908503e-01 -8.98563087e-01 -3.58865172e-01 -3.19410145e-01 -1.49858192e-01 5.44717789e-01 3.05310398e-01 1.19346201e+00 -3.93628508e-01 -9.93091404e-01 4.78825688e-01 1.50700521e+00 5.54555535e-01 5.26484966e-01 -1.92520142e-01 1.21029973e+00 4.07919258e-01 4.21550661e-01 5.27851641e-01 6.17806911e-01 8.75573695e-01 4.54232574e-01 -2.20570654e-01 3.25657167e-02 -4.94257748e-01 2.09760159e-01 1.09553802e+00 -1.78840667e-01 -3.73467147e-01 -6.82075202e-01 5.13887584e-01 -1.84555209e+00 -6.77304685e-01 -1.66426703e-01 2.07775450e+00 1.17183483e+00 -2.32372373e-01 3.85148302e-02 -3.68586898e-01 6.43282771e-01 3.82790446e-01 -8.48936856e-01 -2.02850729e-01 1.04932941e-01 4.51091170e-01 2.25367352e-01 4.14132953e-01 -1.02598572e+00 6.14013970e-01 4.95359945e+00 8.53351474e-01 -8.36230636e-01 1.41036445e-02 7.55303025e-01 3.84470075e-01 -8.84770215e-01 -1.72172114e-01 -6.29027545e-01 5.22023141e-01 7.34942198e-01 -2.51801163e-01 1.24712691e-01 8.11266959e-01 -1.08018713e-02 5.84860146e-01 -1.48784816e+00 1.05159426e+00 5.64208962e-02 -1.71446860e+00 5.03785193e-01 4.24873292e-01 7.45460868e-01 -2.36770231e-02 4.14354578e-02 1.28690163e-02 5.02958477e-01 -1.19780731e+00 -6.13171875e-01 3.84045720e-01 8.76385391e-01 -7.78282881e-01 6.10192180e-01 -2.16143534e-01 -1.46373558e+00 2.87953317e-01 -2.81107396e-01 7.57568419e-01 -2.21544027e-01 5.13957918e-01 -8.53183329e-01 1.29701936e+00 1.71718106e-01 1.65946078e+00 -4.01308507e-01 8.22640538e-01 -2.09371716e-01 1.60145193e-01 1.39680222e-01 4.89915200e-02 2.24600121e-01 -6.36717200e-01 3.70812327e-01 7.35454619e-01 1.98577121e-02 2.56342381e-01 2.54279256e-01 5.53244710e-01 -4.48446363e-01 3.96875501e-01 -1.09712493e+00 -4.72063422e-01 5.09395540e-01 1.11710978e+00 -2.38237791e-02 -3.46597642e-01 -8.62981975e-01 1.08865356e+00 3.40257943e-01 2.32395470e-01 -8.98827076e-01 -3.14753056e-01 1.12014949e+00 1.16247572e-01 1.29564628e-01 3.12993169e-01 1.83541059e-01 -1.10087705e+00 -2.78700113e-01 -1.17515671e+00 5.99462569e-01 -4.12376434e-01 -1.71981454e+00 6.52473927e-01 -1.68464571e-01 -1.56118906e+00 1.48041263e-01 -7.56954610e-01 -4.90660310e-01 7.70296514e-01 -1.59380424e+00 -1.22890031e+00 1.13259234e-01 6.75409973e-01 3.53077412e-01 -4.08542484e-01 1.12781966e+00 5.22023737e-01 -7.16372252e-01 6.74676239e-01 2.73274422e-01 -1.60237670e-01 7.46843219e-01 -1.17274535e+00 2.61594176e-01 -8.81272182e-02 2.60501206e-01 7.59116292e-01 2.31466740e-01 -8.98606896e-01 -1.68809438e+00 -1.29034913e+00 8.57403755e-01 -6.09391987e-01 7.93147147e-01 -1.87642947e-01 -1.14350152e+00 6.21617734e-01 9.81009230e-02 3.76082778e-01 1.32706332e+00 1.85214221e-01 -6.76595867e-01 -1.32393956e-01 -1.12534928e+00 6.17352664e-01 1.17005432e+00 -5.58603466e-01 -5.30709386e-01 9.25152004e-01 8.37764382e-01 -3.72487754e-01 -1.59075701e+00 4.94751632e-01 4.80051607e-01 -5.94462991e-01 1.28517985e+00 -1.09963095e+00 6.11307383e-01 -3.55549127e-01 6.66858479e-02 -1.49959254e+00 -4.31977421e-01 -1.95690125e-01 -3.45732421e-01 6.99918091e-01 4.81236309e-01 -9.52514648e-01 5.54597080e-01 1.96031094e-01 2.41240337e-02 -1.70404160e+00 -7.77618051e-01 -6.54280841e-01 -1.43196881e-01 3.01131964e-01 7.66565800e-01 1.53073835e+00 2.21194267e-01 6.57714188e-01 -3.82499456e-01 3.72826606e-01 3.31676722e-01 1.64241880e-01 6.15422845e-01 -1.28764939e+00 -4.74290371e-01 -1.90358087e-01 -7.97754824e-01 -6.89268291e-01 -1.26177832e-01 -1.15774584e+00 -6.77758813e-01 -1.56168592e+00 6.36385500e-01 -1.73892781e-01 -6.48780882e-01 4.58950937e-01 -4.12940472e-01 -1.85524762e-01 -2.54225314e-01 2.10814297e-01 -2.50946909e-01 6.71774566e-01 1.48872447e+00 -5.21520078e-01 -3.08846503e-01 -2.42204383e-01 -6.86263442e-01 4.87625778e-01 3.33341539e-01 -5.15764415e-01 -8.12832594e-01 1.57086805e-01 1.61195993e-02 2.44956948e-02 -5.83223216e-02 -3.87395263e-01 2.42971303e-03 -3.17434311e-01 3.12117904e-01 -4.60750878e-01 1.51251778e-01 -8.34416568e-01 4.97133762e-01 6.07557178e-01 -8.92353524e-03 -8.75844210e-02 -2.86432561e-02 1.10321021e+00 -3.89816344e-01 1.97193012e-01 6.33952320e-01 1.03452206e-01 -2.19618097e-01 1.23073435e+00 1.64882079e-01 -6.19391575e-02 1.36706150e+00 -1.19227536e-01 -4.36696887e-01 1.16856255e-01 -7.80373752e-01 2.39982337e-01 2.54892588e-01 5.87141275e-01 1.13383734e+00 -1.61999190e+00 -5.12580454e-01 2.32356399e-01 6.88856483e-01 -1.50928557e-01 3.34205359e-01 6.89746857e-01 -3.81331593e-01 2.84037322e-01 1.12490028e-01 -4.42137033e-01 -1.61042356e+00 9.97440457e-01 2.43994519e-01 -7.03643560e-01 -6.42322898e-01 8.79951417e-01 6.76121950e-01 -5.46846569e-01 4.90681008e-02 -5.52641749e-02 -4.82572824e-01 1.88454077e-01 4.06048447e-01 7.38697941e-04 -6.85716374e-03 -5.15000284e-01 -7.83842981e-01 7.54668474e-01 -6.80922925e-01 7.66770184e-01 1.68940806e+00 6.06437325e-01 -2.42117897e-01 1.05098024e-01 1.59534788e+00 -2.23092642e-02 -7.96147168e-01 -3.94405663e-01 -1.66832864e-01 -6.88550353e-01 -2.10579857e-01 -7.12995231e-01 -1.06950915e+00 7.16412127e-01 6.98111951e-01 -9.40212160e-02 7.03656554e-01 1.27699465e-01 7.63079941e-01 8.18793327e-02 1.91001400e-01 -7.22977757e-01 3.04965973e-01 2.11238079e-02 1.08052957e+00 -1.18415344e+00 4.23041761e-01 -7.58508801e-01 -8.33659530e-01 1.09952247e+00 6.00903273e-01 -4.12421413e-02 1.01680005e+00 -1.48792177e-01 -2.32437715e-01 -8.91677499e-01 -7.37438798e-01 1.99022442e-01 5.93288422e-01 5.74941754e-01 6.29405200e-01 2.47716397e-01 -4.60508883e-01 2.54529953e-01 7.54716039e-01 -6.01284951e-02 4.56930734e-02 6.90280914e-01 4.35439199e-02 -1.68305504e+00 2.66461611e-01 6.42416000e-01 -6.99291676e-02 -2.98685312e-01 -6.37013257e-01 7.36322343e-01 1.66380823e-01 5.36764264e-01 -4.13456857e-01 -7.69454837e-01 4.33107406e-01 -1.16861343e-01 4.61014301e-01 -9.73239839e-01 -4.59965289e-01 8.58319253e-02 -1.32081091e-01 -6.36635602e-01 -3.80812407e-01 -2.72796184e-01 -1.11492741e+00 -2.91708916e-01 -4.04433697e-01 -4.81647905e-03 1.03254788e-01 7.40736902e-01 7.64924109e-01 6.11831665e-01 1.30644429e+00 -1.64382979e-01 -4.06872302e-01 -6.28641009e-01 -6.37947083e-01 6.87264621e-01 9.30247083e-02 -8.45482886e-01 -1.03152819e-01 -3.29149306e-01]
[5.2361931800842285, 5.897984504699707]
4aea4007-a525-461d-bca5-7d6d4cac51d5
speech-privacy-leakage-from-shared-gradients
2302.10441
null
https://arxiv.org/abs/2302.10441v1
https://arxiv.org/pdf/2302.10441v1.pdf
Speech Privacy Leakage from Shared Gradients in Distributed Learning
Distributed machine learning paradigms, such as federated learning, have been recently adopted in many privacy-critical applications for speech analysis. However, such frameworks are vulnerable to privacy leakage attacks from shared gradients. Despite extensive efforts in the image domain, the exploration of speech privacy leakage from gradients is quite limited. In this paper, we explore methods for recovering private speech/speaker information from the shared gradients in distributed learning settings. We conduct experiments on a keyword spotting model with two different types of speech features to quantify the amount of leaked information by measuring the similarity between the original and recovered speech signals. We further demonstrate the feasibility of inferring various levels of side-channel information, including speech content and speaker identity, under the distributed learning framework without accessing the user's data.
['Jian Liu', 'Jiaxin Zhang', 'Zhuohang Li']
2023-02-21
null
null
null
null
['keyword-spotting']
['speech']
[ 1.14029616e-01 5.31267673e-02 -1.54656097e-01 -4.85474914e-01 -1.42492878e+00 -9.66125011e-01 4.35748875e-01 -4.75465413e-03 -2.62963027e-01 4.99802411e-01 5.02533972e-01 -4.71322298e-01 1.44721299e-01 -4.48794127e-01 -6.09342396e-01 -9.85397339e-01 -2.90969074e-01 -4.57202941e-01 -3.63615692e-01 2.48245433e-01 -1.32369190e-01 5.10901392e-01 -1.02383399e+00 4.62247282e-01 1.21404633e-01 1.22244132e+00 -4.03881848e-01 6.86192274e-01 -7.96359256e-02 7.49576271e-01 -6.98356748e-01 -8.21756363e-01 7.46431112e-01 -3.76254082e-01 -5.71482718e-01 -1.23662479e-01 3.60076070e-01 -7.41551280e-01 -8.49684596e-01 1.48994899e+00 7.08072603e-01 -2.81302363e-01 -3.67742032e-02 -1.47660708e+00 -3.49629313e-01 6.19464219e-01 -1.40059143e-01 1.35607004e-01 2.46399909e-01 2.61777908e-01 9.68194783e-01 -7.11090863e-01 5.69744825e-01 9.81256127e-01 5.59636891e-01 5.63134670e-01 -8.64105403e-01 -8.81725311e-01 -6.95910156e-02 2.37735480e-01 -1.54682422e+00 -9.06723499e-01 7.20680714e-01 -7.08204061e-02 5.36299944e-01 8.48027408e-01 1.44572914e-01 1.11522341e+00 -1.24977432e-01 1.05486667e+00 1.21312261e+00 -2.50610352e-01 4.15933281e-01 6.82850897e-01 -3.48783880e-01 5.54829478e-01 6.34583533e-02 1.56344846e-01 -1.13193834e+00 -8.96008909e-01 6.02626316e-02 7.01679811e-02 -6.28256440e-01 -3.47353816e-01 -7.47373044e-01 8.17104042e-01 4.09444571e-02 1.64165832e-02 -1.72125608e-01 1.12061925e-01 6.34333670e-01 6.42466962e-01 6.96294546e-01 -2.29516014e-01 -5.29767811e-01 7.48599544e-02 -5.48828423e-01 9.21621174e-02 1.20535946e+00 9.36801136e-01 9.80234623e-01 -2.51574695e-01 -1.31645739e-01 4.20536965e-01 3.21165144e-01 5.52340984e-01 4.77266461e-01 -6.88642263e-01 6.38906896e-01 1.35708349e-02 2.07674038e-02 -1.22187555e+00 4.14375663e-01 -7.14729279e-02 -6.96704209e-01 -9.90068391e-02 2.96507418e-01 -6.12450540e-01 -1.88078359e-01 1.83307540e+00 5.83586395e-01 2.03662217e-01 3.57955098e-01 1.05542541e+00 2.85845697e-01 2.77818263e-01 -9.12912115e-02 -1.35081068e-01 1.10739565e+00 -7.62184918e-01 -7.35510230e-01 -8.55642259e-02 6.02671504e-01 -3.70692790e-01 7.01459646e-01 2.02386603e-01 -8.55775177e-01 3.01922411e-01 -9.47011709e-01 -1.42444879e-01 -3.79103541e-01 -2.93063343e-01 8.81823123e-01 1.28715205e+00 -1.21014559e+00 2.57752895e-01 -6.64339244e-01 -7.10875019e-02 8.07503045e-01 3.49884331e-01 -6.43477559e-01 -3.42417918e-02 -1.34503388e+00 1.30870283e-01 -3.58133674e-01 3.02655008e-02 -1.20836556e+00 -6.14567816e-01 -6.67325199e-01 3.11502814e-01 2.24860892e-01 -4.58408087e-01 1.33844984e+00 -8.83864462e-01 -1.40453684e+00 9.05905426e-01 3.20987925e-02 -7.45117486e-01 8.18978548e-01 2.09441543e-01 -5.02629161e-01 1.68967888e-01 -1.79269060e-01 -9.07302424e-02 1.24022353e+00 -7.91522920e-01 -6.98812842e-01 -9.36274350e-01 -1.62501574e-01 1.61872000e-01 -9.66276705e-01 2.80984491e-01 -6.47457838e-02 -5.06210625e-01 -1.20963782e-01 -8.01902711e-01 -2.36464769e-01 6.02007568e-01 -4.68481153e-01 3.23355794e-01 1.37101698e+00 -8.11144114e-01 8.51288855e-01 -2.68222284e+00 -2.22790375e-01 3.47263157e-01 3.80584329e-01 6.93213344e-02 1.82389960e-01 4.71752644e-01 3.53338391e-01 1.96166739e-01 -2.39415914e-01 -8.35557878e-01 7.72699192e-02 -9.55637991e-02 -7.48190880e-01 1.03803933e+00 -3.40584308e-01 7.50732422e-01 -6.98372781e-01 -1.95820913e-01 -2.25124627e-01 6.09188437e-01 -2.45705292e-01 3.34948242e-01 2.57370062e-02 4.69249696e-01 -5.91229141e-01 6.24701321e-01 8.21966529e-01 -2.06406899e-02 5.19166589e-01 1.84166104e-01 1.74896389e-01 2.98040360e-01 -9.31787729e-01 1.75987434e+00 -3.81666750e-01 7.35068679e-01 1.09630346e+00 -6.38580739e-01 6.38280511e-01 7.48443127e-01 3.70400101e-01 -3.95720840e-01 3.23019102e-02 8.03263485e-02 -4.42482054e-01 -5.22993743e-01 2.23877400e-01 -5.64836664e-03 -2.33484671e-01 9.03793633e-01 -1.11871243e-01 3.29215497e-01 -1.07032812e+00 3.64564508e-01 1.45045161e+00 -6.91138685e-01 5.46350963e-02 -6.09532632e-02 4.01737154e-01 -6.77336395e-01 4.95811105e-01 7.44307220e-01 -7.86052704e-01 2.73354620e-01 3.56044263e-01 -2.46552527e-01 -7.13276327e-01 -7.38190830e-01 8.06884542e-02 1.10819459e+00 1.48547307e-01 -5.02389848e-01 -8.86823893e-01 -1.05477250e+00 3.89970869e-01 5.69112599e-02 -1.38472140e-01 -1.39133558e-01 -1.27008125e-01 -4.42742288e-01 1.23148632e+00 9.11514163e-02 5.02778172e-01 -5.79584241e-01 -4.59968418e-01 -2.74785936e-01 -1.47454977e-01 -1.26442039e+00 -9.61635470e-01 3.66360438e-03 -5.14502883e-01 -8.00562143e-01 -3.89585197e-01 -3.69599074e-01 5.12571275e-01 4.78189766e-01 3.82305771e-01 -1.25160694e-01 -5.36449432e-01 9.40213084e-01 8.13022777e-02 -4.51786667e-01 -4.61708814e-01 -2.22936913e-01 -1.04386002e-01 7.68288314e-01 2.71339357e-01 -4.93989229e-01 -7.55607486e-01 3.69705707e-02 -9.32019114e-01 -4.43890393e-01 2.12633386e-01 5.22867620e-01 3.00503403e-01 -1.03071913e-01 5.56258261e-01 -1.01670063e+00 8.42895329e-01 -6.99406266e-01 -3.28407586e-01 2.37004578e-01 -5.44721067e-01 2.76031648e-03 5.89419484e-01 -3.66826743e-01 -1.11421585e+00 1.30343512e-01 -4.06480627e-03 -5.84374607e-01 -2.01954424e-01 2.37284556e-01 -6.71491802e-01 -6.40123069e-01 4.73756373e-01 2.84556389e-01 3.26961726e-01 -4.45724845e-01 6.30696893e-01 1.33015466e+00 3.36982340e-01 -4.66806680e-01 4.51017410e-01 8.88433814e-01 -3.64798278e-01 -8.41659665e-01 -1.79098606e-01 -5.84725857e-01 1.29152015e-01 1.87196136e-01 2.35987663e-01 -1.10928500e+00 -7.79726505e-01 8.42100561e-01 -1.01988864e+00 -4.30067480e-02 -3.47922176e-01 3.20794314e-01 -5.08968771e-01 7.27566063e-01 -7.96310246e-01 -1.11160278e+00 -8.63041461e-01 -1.06627750e+00 9.39664006e-01 -1.32398278e-01 2.56022543e-01 -8.27276230e-01 -5.49776107e-03 4.57503498e-01 7.63249576e-01 5.09679653e-02 5.51523983e-01 -9.54757512e-01 -8.10898542e-01 -6.09162569e-01 8.58642235e-02 4.10500586e-01 2.30949596e-01 -6.92824721e-01 -1.59419513e+00 -6.07339919e-01 6.38629615e-01 -3.55448455e-01 5.15078723e-01 -1.18419707e-01 1.30397809e+00 -1.13625014e+00 -1.62340492e-01 9.41380084e-01 1.29491508e+00 -2.52391875e-01 5.00889003e-01 -1.95304677e-01 5.84257483e-01 6.42863631e-01 -3.28299999e-02 9.28139150e-01 2.90150166e-01 3.55913311e-01 4.07449394e-01 1.90693483e-01 7.84979388e-02 -5.14601529e-01 4.03876066e-01 4.87071276e-01 7.90389240e-01 -2.43563518e-01 -5.01956463e-01 6.34822845e-01 -1.71308768e+00 -8.18228066e-01 5.24891138e-01 2.41492009e+00 8.71392846e-01 -6.40062213e-01 5.75333573e-02 -1.53022543e-01 7.68290162e-01 5.08362353e-01 -8.86653066e-01 -3.16100687e-01 -2.52155483e-01 5.93654253e-03 9.74496186e-01 3.58238280e-01 -1.04174709e+00 8.20155919e-01 6.36961651e+00 5.20729005e-01 -1.32693422e+00 5.33709526e-01 9.42960083e-01 -2.94877231e-01 -5.34289479e-01 1.11058243e-01 -5.41547596e-01 4.68042791e-01 1.06620228e+00 -6.04833484e-01 7.16398239e-01 9.58853960e-01 -1.03752136e-01 5.33339262e-01 -1.04485142e+00 1.21576381e+00 2.11563855e-02 -1.23191214e+00 -3.25440228e-01 4.63804513e-01 3.47153455e-01 2.93305367e-01 2.81710297e-01 -2.20902994e-01 3.96215826e-01 -7.60141373e-01 5.39786458e-01 1.07062407e-01 9.67771769e-01 -6.55569017e-01 1.33326292e-01 5.00643730e-01 -7.60324180e-01 -3.08432460e-01 -2.67116696e-01 1.93024203e-01 -1.40475065e-01 5.16219914e-01 -8.03495467e-01 2.20045224e-01 9.20277059e-01 2.19207272e-01 -1.73172116e-01 5.02697647e-01 -1.81330442e-01 7.20787466e-01 -5.50822794e-01 -6.90467283e-02 -2.38733701e-02 -3.14281955e-02 6.23543501e-01 1.07977664e+00 2.12208807e-01 -1.34432927e-01 -9.53196213e-02 8.21028650e-01 -6.84966862e-01 3.24512661e-01 -1.04946196e+00 -2.64182650e-02 7.73885787e-01 1.34007907e+00 3.85937393e-02 4.88548167e-02 -4.99290913e-01 1.29213512e+00 3.00347298e-01 4.77382630e-01 -4.42937016e-01 -2.63275146e-01 1.36098480e+00 -1.61588654e-01 1.77187428e-01 -1.20592572e-01 -1.88615382e-01 -1.33367646e+00 4.10298020e-01 -1.18093657e+00 5.46356499e-01 -1.19398795e-01 -1.37040639e+00 2.41559401e-01 -4.70161349e-01 -7.81033993e-01 -2.62931675e-01 -5.85598983e-02 -4.15051281e-01 9.91334021e-01 -1.37691176e+00 -1.17002273e+00 2.02109158e-01 1.22868538e+00 -2.67514624e-02 -3.92736286e-01 1.11155963e+00 2.09540680e-01 -3.71233791e-01 1.25523853e+00 2.89600343e-01 4.25189406e-01 5.71694791e-01 -7.45273888e-01 5.43642581e-01 9.18486357e-01 4.52163070e-01 6.35079682e-01 4.02012616e-01 -3.88936192e-01 -2.20384431e+00 -9.83727336e-01 6.90834701e-01 -1.97325945e-02 6.26420915e-01 -7.26756513e-01 -7.60851920e-01 7.79647052e-01 2.02026904e-01 6.92567110e-01 1.18579316e+00 -9.95738804e-02 -8.81471395e-01 -3.29588950e-01 -1.85243666e+00 3.48318517e-01 7.37703443e-01 -1.57016754e+00 3.89046997e-01 4.01517600e-01 8.39349747e-01 -1.82564840e-01 -8.02102506e-01 -2.71090955e-01 5.25926173e-01 -9.16386366e-01 8.14837873e-01 -6.45802200e-01 -4.35275257e-01 7.35711604e-02 -4.94385213e-01 -1.02352822e+00 3.71740133e-01 -1.29867339e+00 -3.85194898e-01 1.37256432e+00 2.26198032e-01 -9.96036053e-01 1.27674007e+00 1.17452788e+00 5.06531954e-01 -2.51012057e-01 -1.54995155e+00 -4.34339941e-01 -6.26606867e-02 -4.19674397e-01 9.58710134e-01 9.53272283e-01 3.47434253e-01 -1.69458687e-01 -6.40168726e-01 5.62629819e-01 9.33118463e-01 -1.18298762e-01 8.97062242e-01 -3.81518215e-01 -4.19594169e-01 1.59229845e-01 -6.69241428e-01 -7.85661161e-01 5.39886355e-01 -9.31469619e-01 -1.04804650e-01 -5.15009940e-01 1.19336195e-01 -4.58343148e-01 -4.22528863e-01 5.98697484e-01 1.02003917e-01 -1.71715230e-01 1.44294098e-01 3.11207920e-01 -3.44180197e-01 4.06954706e-01 5.48870087e-01 -3.12402546e-01 1.86716899e-01 2.33177572e-01 -9.67229903e-01 2.48780534e-01 8.83971274e-01 -5.94147205e-01 -5.15291750e-01 -5.84557593e-01 -2.33102832e-02 3.47072035e-01 4.82387424e-01 -5.11107028e-01 7.09151268e-01 2.30405722e-02 -2.30459750e-01 9.89000946e-02 1.98004842e-01 -1.09507263e+00 1.13343365e-01 7.83188269e-02 -6.75409794e-01 -2.88720340e-01 -1.12069249e-01 9.56240356e-01 -3.74397635e-01 1.64791599e-01 4.58297431e-01 -2.15916499e-01 -4.10055727e-01 7.01056540e-01 -1.12618253e-01 -9.80498493e-02 1.02402616e+00 1.60312340e-01 -3.04673523e-01 -8.29023480e-01 -5.88592052e-01 -6.47567492e-03 5.07335365e-01 2.87793010e-01 7.94605792e-01 -9.68276381e-01 -5.88822305e-01 6.67245984e-01 2.57728714e-02 -4.13809597e-01 2.76086271e-01 4.60628986e-01 -6.83532842e-03 3.37308556e-01 3.92120093e-01 -2.46558875e-01 -1.40556657e+00 6.99466109e-01 4.46774453e-01 1.26859695e-01 -7.69834638e-01 7.38954782e-01 6.53404519e-02 -3.66480291e-01 6.00610256e-01 2.14886084e-01 7.69109488e-01 -2.44290411e-01 8.91945601e-01 3.79089445e-01 4.17249382e-01 -5.08796930e-01 -3.88153404e-01 -3.19664508e-01 -1.70185789e-02 -4.98354018e-01 9.42812026e-01 -5.42652607e-01 -1.53762281e-01 -1.39115989e-01 1.91113710e+00 3.62334996e-01 -1.28044915e+00 -4.80708957e-01 4.24950644e-02 -9.04706180e-01 2.40052059e-01 -4.80385751e-01 -1.39856172e+00 7.70661712e-01 8.20696294e-01 1.49779826e-01 9.86816585e-01 4.16340679e-03 1.04810393e+00 3.23657572e-01 8.87643635e-01 -7.38152802e-01 -5.04670203e-01 -1.66681767e-01 4.47182059e-01 -1.14880216e+00 -3.33041281e-01 -1.90981761e-01 -5.68748891e-01 6.12661064e-01 -4.51880731e-02 3.99807602e-01 9.45208728e-01 6.79787517e-01 4.02579755e-01 4.41736393e-02 -8.06650937e-01 3.69035095e-01 -4.08055753e-01 6.51676178e-01 -9.59036648e-02 3.28149199e-01 5.33662550e-02 6.29169822e-01 1.81093886e-02 -1.29785642e-01 3.50039989e-01 1.27581620e+00 1.94051117e-02 -1.04211617e+00 -2.35162273e-01 6.95190132e-02 -1.08572042e+00 -9.33702886e-02 -6.69177830e-01 -6.58240840e-02 -3.89116436e-01 1.05007696e+00 -2.81202048e-01 -4.35352534e-01 -7.35003874e-02 9.67646539e-02 2.03542621e-03 -3.63282591e-01 -8.05308402e-01 -1.03781521e-01 -9.16271582e-02 -7.44405270e-01 -9.44868773e-02 -7.23559678e-01 -1.00784063e+00 -4.71592575e-01 -3.45625192e-01 2.63727397e-01 1.22531688e+00 5.60711443e-01 7.92816579e-01 -4.44093853e-01 1.32006991e+00 -2.24876374e-01 -1.29171145e+00 -2.77772218e-01 -9.58451986e-01 4.87927049e-01 7.40035832e-01 3.62545133e-01 -9.51371014e-01 -2.31125832e-01]
[5.869373798370361, 6.690192699432373]
467c5e7c-86c3-44ad-92c3-53ce2a00c90c
the-story-in-your-eyes-an-individual
2106.14183
null
https://arxiv.org/abs/2106.14183v1
https://arxiv.org/pdf/2106.14183v1.pdf
The Story in Your Eyes: An Individual-difference-aware Model for Cross-person Gaze Estimation
We propose a novel method on refining cross-person gaze prediction task with eye/face images only by explicitly modelling the person-specific differences. Specifically, we first assume that we can obtain some initial gaze prediction results with existing method, which we refer to as InitNet, and then introduce three modules, the Validity Module (VM), Self-Calibration (SC) and Person-specific Transform (PT)) Module. By predicting the reliability of current eye/face images, our VM is able to identify invalid samples, e.g. eye blinking images, and reduce their effects in our modelling process. Our SC and PT module then learn to compensate for the differences on valid samples only. The former models the translation offsets by bridging the gap between initial predictions and dataset-wise distribution. And the later learns more general person-specific transformation by incorporating the information from existing initial predictions of the same person. We validate our ideas on three publicly available datasets, EVE, XGaze and MPIIGaze and demonstrate that our proposed method outperforms the SOTA methods significantly on all of them, e.g. respectively 21.7%, 36.0% and 32.9% relative performance improvements. We won the GAZE 2021 Competition on the EVE dataset. Our code can be found here https://github.com/bjj9/EVE_SCPT.
['Jun Yu', 'Buyu Liu', 'Jun Bao']
2021-06-27
null
null
null
null
['gaze-estimation', 'eye-tracking']
['computer-vision', 'computer-vision']
[ 1.73735797e-01 9.00337845e-02 -1.65000916e-01 -5.03930807e-01 -4.25995767e-01 -1.91739872e-01 4.00567114e-01 -5.38554072e-01 -4.00631934e-01 7.96575069e-01 2.95065884e-02 1.13778621e-01 -3.20735388e-02 -1.14613496e-01 -7.63005614e-01 -5.70922434e-01 4.17365372e-01 2.01985285e-01 1.53798744e-01 4.05436195e-03 3.73054206e-01 -6.56603500e-02 -2.00206470e+00 1.73720047e-01 1.04186511e+00 1.16432345e+00 -2.04337090e-01 5.68386614e-01 3.65665436e-01 7.27449596e-01 -4.30449516e-01 -4.72656846e-01 2.98180848e-01 -4.27690625e-01 -7.09030509e-01 1.34326696e-01 8.01153004e-01 -4.29971576e-01 6.43887892e-02 9.99839962e-01 6.25189543e-01 -7.05314279e-02 3.72374117e-01 -1.84624803e+00 -6.97225094e-01 2.61975199e-01 -1.15228844e+00 1.04536854e-01 3.88411760e-01 5.50039768e-01 6.56440318e-01 -8.38734686e-01 3.95765781e-01 1.24182987e+00 6.60496056e-01 1.08956194e+00 -1.05836356e+00 -1.08614945e+00 2.35733360e-01 5.03335357e-01 -1.47544193e+00 -9.00493860e-01 5.65513790e-01 -4.38996762e-01 6.04749918e-01 4.18076485e-01 3.23181599e-01 1.24954855e+00 -9.18606371e-02 8.48965466e-01 1.52546823e+00 -4.07262474e-01 -2.91707158e-01 3.97204489e-01 5.38241506e-01 4.50457990e-01 -5.99840693e-02 2.74855793e-01 -7.39160120e-01 1.03244372e-01 2.43957087e-01 -1.81258731e-02 -7.71128893e-01 -2.14805737e-01 -8.64829659e-01 2.63037801e-01 5.47248602e-01 -1.88836604e-01 -1.29704103e-01 -9.25988331e-02 6.61509931e-02 1.43073633e-01 5.07771671e-01 -5.77716269e-02 -6.50253654e-01 -1.43485874e-01 -9.88382757e-01 2.04192206e-01 4.85958755e-01 1.09537446e+00 8.81429851e-01 -5.98773062e-01 -4.66849476e-01 9.34357762e-01 5.34906626e-01 5.10261655e-01 5.33401489e-01 -8.22273195e-01 3.73500586e-01 6.91719472e-01 3.91347051e-01 -5.70913136e-01 -4.56930280e-01 -2.97214091e-01 -6.24486804e-01 1.61220461e-01 5.49710393e-01 -2.53622472e-01 -9.94101822e-01 2.04711747e+00 3.85947078e-01 5.64666331e-01 -4.20032799e-01 9.22613978e-01 8.72028232e-01 2.40852222e-01 9.02638733e-02 -4.38051283e-01 1.37986505e+00 -1.14199662e+00 -7.52504230e-01 -2.23659098e-01 5.49795747e-01 -6.43878460e-01 1.25115490e+00 4.47252482e-01 -1.16284037e+00 -7.20608950e-01 -7.93361664e-01 -2.31850386e-01 -1.36917233e-01 5.12214601e-01 1.83971167e-01 7.75128841e-01 -1.31254137e+00 4.70238805e-01 -5.43750226e-01 -4.61685240e-01 6.38022780e-01 6.81812584e-01 -1.73820198e-01 4.68734391e-02 -8.80964518e-01 7.27507770e-01 4.09189574e-02 1.40302792e-01 -5.18809855e-01 -1.01749122e+00 -6.06014729e-01 -3.69927548e-02 4.86124188e-01 -7.31606662e-01 1.23427260e+00 -1.25830090e+00 -1.46367168e+00 1.09812772e+00 -8.73078644e-01 -2.51311094e-01 7.40070224e-01 -4.58553702e-01 -4.96917188e-01 -3.67116302e-01 -5.99599779e-02 9.57547545e-01 9.60411489e-01 -1.19846964e+00 -9.32787359e-01 -5.10295868e-01 -7.04724714e-02 2.17128128e-01 -2.58093327e-01 3.85424495e-01 -8.34674597e-01 -1.57084122e-01 -3.93513471e-01 -1.15164518e+00 5.30435264e-01 1.35870308e-01 -5.50907135e-01 -6.24236703e-01 6.46855295e-01 -8.98607850e-01 1.51882267e+00 -2.20326424e+00 3.50348726e-02 1.70412377e-01 6.05090022e-01 3.62896562e-01 -3.98162976e-02 -2.05707967e-01 -3.38351488e-01 7.99452290e-02 -4.97307209e-03 -1.00085998e+00 -5.19510545e-02 -2.73904622e-01 -1.02328947e-02 3.11294317e-01 1.83173135e-01 9.77755189e-01 -5.24680734e-01 -3.92882109e-01 -1.05477720e-02 5.40067255e-01 -4.98058408e-01 6.54740557e-02 -1.89555138e-02 5.33396304e-01 2.63950471e-02 5.33430994e-01 1.00726974e+00 -5.08948743e-01 -1.50562704e-01 -2.20327318e-01 -1.42768413e-01 1.87014654e-01 -1.00849903e+00 1.15977979e+00 -1.36848539e-01 6.97411954e-01 -2.83523053e-01 -6.23679347e-02 5.47670841e-01 9.38556790e-02 1.38738245e-01 -7.42535651e-01 3.38112891e-01 -1.72637358e-01 1.34145692e-01 -4.10253644e-01 4.30178940e-01 2.84787029e-01 5.32352030e-01 3.69015485e-01 6.23851502e-03 8.19582820e-01 1.09125011e-01 1.74386501e-02 5.87354898e-01 3.80882353e-01 1.09826431e-01 -1.10143155e-01 7.83497095e-01 -4.70464021e-01 5.88887215e-01 3.40587705e-01 -6.33880973e-01 7.33071029e-01 6.32863998e-01 -5.28787859e-02 -7.20434129e-01 -6.22221887e-01 -1.21937498e-01 1.13191140e+00 2.52603829e-01 -5.31539202e-01 -1.08767331e+00 -7.15959847e-01 -1.26385137e-01 7.04104543e-01 -1.03173816e+00 -2.16531917e-01 -3.50997776e-01 -7.08408058e-01 2.07762271e-01 3.76530170e-01 6.85221493e-01 -9.42104101e-01 -2.34272048e-01 -5.53302050e-01 -3.08783531e-01 -8.66019309e-01 -7.35940278e-01 -5.30958772e-01 -5.34963727e-01 -1.26473618e+00 -6.22405410e-01 -2.55958408e-01 6.56983256e-01 3.15243512e-01 1.01295233e+00 4.34901863e-01 5.36422618e-02 1.22591600e-01 -4.59022075e-02 -6.07319951e-01 5.63326068e-02 1.50691047e-01 1.44938573e-01 4.42933291e-01 9.18277085e-01 -1.25177860e-01 -8.22549582e-01 6.48382902e-01 -2.34133095e-01 2.88479716e-01 3.87360990e-01 7.55217075e-01 4.59932923e-01 -3.67703766e-01 2.21118674e-01 -1.00632489e+00 4.92513895e-01 -4.04446781e-01 -5.60892165e-01 3.94139677e-01 -9.74569678e-01 -1.38445795e-01 5.03136776e-02 -4.52401578e-01 -1.13156724e+00 -2.42308095e-01 -1.86676048e-02 -6.48532450e-01 -3.87911469e-01 -4.64307852e-02 -3.21999222e-01 -2.08569542e-02 7.68231928e-01 2.53738910e-02 1.92655995e-02 -4.96179283e-01 -4.09318246e-02 9.15207207e-01 4.23448473e-01 -2.32114509e-01 6.50863409e-01 2.81154037e-01 -2.08585039e-01 -3.28081071e-01 -1.05376470e+00 -4.65664595e-01 -5.18080711e-01 -3.36480767e-01 6.32472515e-01 -1.07696247e+00 -1.18743849e+00 7.92162895e-01 -9.80055273e-01 -3.23663563e-01 -2.98992787e-02 2.75317848e-01 -3.10651213e-01 9.90850404e-02 -2.97772437e-01 -8.83928537e-01 -4.03157175e-01 -1.00105882e+00 1.05138159e+00 6.81972980e-01 -1.66435868e-01 -6.89702213e-01 3.26309912e-02 5.33246577e-01 1.75084233e-01 4.63523678e-02 2.52450913e-01 -4.43612248e-01 -5.64437807e-01 4.66995575e-02 -5.31760693e-01 4.33908552e-01 8.77214968e-02 2.36414760e-01 -1.43904293e+00 -4.35011357e-01 -1.74705699e-01 -1.75672933e-01 8.69723618e-01 4.30455208e-01 1.24680889e+00 -8.15788582e-02 -5.86265087e-01 8.24219942e-01 1.10432541e+00 -2.18602419e-01 9.34236467e-01 2.10231602e-01 9.67944086e-01 6.93224072e-01 6.61448181e-01 2.93424487e-01 8.03484023e-01 7.93632269e-01 3.25063229e-01 1.11373737e-01 -3.72223735e-01 -2.21868977e-01 3.89311492e-01 1.13296181e-01 -5.58416724e-01 -3.31296146e-01 -9.14958656e-01 3.99511486e-01 -1.95939744e+00 -8.57335389e-01 -4.13517326e-01 2.35772109e+00 1.02929878e+00 -5.00448467e-03 5.38545966e-01 1.33709624e-01 9.12386775e-01 -3.65115464e-01 -7.91052401e-01 1.04052305e-01 5.03633693e-02 3.61522399e-02 4.82447743e-01 5.01123428e-01 -9.39548731e-01 8.65069211e-01 5.15756845e+00 5.32290220e-01 -1.03351557e+00 2.60735244e-01 7.65215755e-01 -5.90475738e-01 7.77511150e-02 -1.66956365e-01 -1.27729249e+00 9.24614131e-01 1.09423065e+00 1.69848241e-02 5.97235262e-01 5.05356789e-01 3.04189533e-01 -1.94472313e-01 -1.41962981e+00 1.28722966e+00 3.39093506e-01 -8.60474944e-01 -3.24639678e-01 2.03960061e-01 5.89437604e-01 9.41371024e-02 3.11699152e-01 3.09090674e-01 -1.38529241e-01 -9.98878419e-01 7.46919811e-01 1.09328997e+00 9.43097532e-01 -5.76129735e-01 6.69982731e-01 3.29625100e-01 -6.82968915e-01 -1.94090262e-01 -1.64983660e-01 3.52112949e-02 -1.67500153e-01 1.26012966e-01 -5.09102345e-01 4.32564795e-01 1.11338282e+00 9.51768875e-01 -1.13058424e+00 1.28633904e+00 -3.18857938e-01 6.28775001e-01 -3.26841354e-01 2.36172467e-01 -4.45771009e-01 5.90838082e-02 4.66997087e-01 5.21283984e-01 2.44271040e-01 -8.33554119e-02 -6.07583761e-01 9.01446044e-01 -2.91905463e-01 -1.68414861e-01 7.62950405e-02 6.03386223e-01 4.98164475e-01 1.19657898e+00 9.69454274e-02 -2.25542575e-01 -5.66805065e-01 8.79328549e-01 4.59285736e-01 5.00341296e-01 -1.11434722e+00 -1.50621701e-02 7.68502474e-01 4.09254611e-01 9.71273407e-02 4.72259015e-01 -2.86371708e-01 -1.22088957e+00 1.98806345e-01 -8.90475392e-01 3.45290691e-01 -1.22886086e+00 -1.16710889e+00 5.29523849e-01 2.00548861e-02 -1.23144341e+00 -2.48126447e-01 -4.85962361e-01 -4.76357311e-01 1.37837529e+00 -1.87815940e+00 -1.38522828e+00 -8.35245192e-01 9.21482325e-01 2.15877652e-01 -5.04051745e-02 3.99052501e-01 3.59923363e-01 -1.07437277e+00 1.27178252e+00 -1.89315170e-01 -1.52968779e-01 1.40445542e+00 -1.13024676e+00 3.13679576e-01 7.47336745e-01 -2.76960462e-01 8.21651101e-01 5.90563536e-01 -4.94677305e-01 -7.96339035e-01 -1.02692223e+00 1.13866270e+00 -1.03822553e+00 3.33680838e-01 -3.47044855e-01 -9.78896499e-01 1.00646877e+00 3.29118520e-01 2.32584425e-03 6.31834507e-01 3.58461410e-01 -3.37435603e-01 -2.51977503e-01 -1.20587945e+00 7.20984578e-01 1.27215028e+00 -1.86557904e-01 -3.35363477e-01 6.67954683e-02 3.91622573e-01 -5.62038839e-01 -5.86772501e-01 3.29979300e-01 7.41846621e-01 -1.27285433e+00 7.40380526e-01 -4.95775610e-01 2.59710819e-01 -4.59805608e-01 3.44270647e-01 -1.04952013e+00 -3.49413812e-01 -7.08616972e-01 -4.27522302e-01 1.49095130e+00 2.56020457e-01 -8.01671922e-01 6.58378303e-01 1.12425458e+00 2.55276322e-01 -6.37028098e-01 -5.23144722e-01 -4.39145088e-01 -7.53807947e-02 -2.72274703e-01 1.03458679e+00 7.50258029e-01 -2.89805323e-01 3.59900922e-01 -7.85447240e-01 1.89279556e-01 7.39930570e-01 -2.07617119e-01 1.14467657e+00 -1.37750983e+00 -1.04258314e-01 -3.14416468e-01 -1.02293804e-01 -1.15616655e+00 1.73719928e-01 -3.81888658e-01 -1.58022866e-01 -8.36723626e-01 5.33849955e-01 -2.01941729e-01 -4.92762953e-01 6.83274209e-01 -7.45880902e-01 4.76705074e-01 3.05119723e-01 5.81873715e-01 -7.22218156e-01 3.13904047e-01 1.08740973e+00 2.49751598e-01 -3.62340093e-01 2.52045929e-01 -9.83898520e-01 6.25117302e-01 8.35379124e-01 -3.27800900e-01 -3.14880162e-01 -3.53388578e-01 2.16384098e-01 -5.19589663e-01 7.52957225e-01 -9.58575070e-01 3.93437207e-01 -3.43790883e-03 6.23918533e-01 -6.27831817e-01 1.54136270e-01 -6.73132420e-01 5.15698381e-02 6.80234581e-02 -2.39982024e-01 -3.09106141e-01 3.66895586e-01 4.29169148e-01 1.52903542e-01 1.13765001e-01 8.84455502e-01 4.39913899e-01 -5.41688561e-01 4.70120221e-01 5.25302708e-01 -4.64888215e-02 1.01129234e+00 -4.51073587e-01 -7.84860909e-01 -2.13979244e-01 -8.16414416e-01 6.01504982e-01 7.73733020e-01 6.49680436e-01 2.60099620e-01 -1.18681598e+00 -6.04083300e-01 4.15900439e-01 4.80477452e-01 -3.01617593e-01 4.43372339e-01 1.34505844e+00 1.79788843e-01 2.85969019e-01 -3.07318747e-01 -8.12217355e-01 -1.63879478e+00 5.12294888e-01 5.31525314e-01 2.76462398e-02 -8.31824820e-03 1.06222713e+00 4.27481979e-01 -1.13276444e-01 1.87507331e-01 -2.45870590e-01 -4.62496787e-01 -9.31070149e-02 7.66847372e-01 4.85917896e-01 -9.52991396e-02 -9.88930523e-01 -5.13422549e-01 7.57529557e-01 -3.14442724e-01 2.79886186e-01 1.09796023e+00 -5.67732275e-01 -1.30094051e-01 3.27750415e-01 9.37480986e-01 -2.54187807e-02 -1.58284521e+00 -3.67732286e-01 -2.39423081e-01 -5.99776208e-01 -8.37045312e-02 -1.19872737e+00 -1.06636930e+00 7.77899086e-01 9.94131684e-01 -1.42079331e-02 1.52967036e+00 2.73461137e-02 5.19079447e-01 -2.69925892e-01 8.95623490e-02 -9.08577323e-01 -3.23167801e-01 2.02403277e-01 8.80555868e-01 -1.65121090e+00 -9.52572599e-02 -5.39826095e-01 -7.69721270e-01 5.84083736e-01 9.84739125e-01 1.48897052e-01 7.45182514e-01 -2.65073001e-01 -7.39816055e-02 -7.79926851e-02 -9.33107018e-01 -3.00865084e-01 8.40073109e-01 6.50032938e-01 3.55212182e-01 -1.42831430e-01 -1.96929276e-01 8.99292350e-01 -3.00388664e-01 4.99646693e-01 2.10211739e-01 4.40552175e-01 3.86150628e-02 -1.02381504e+00 -4.17241633e-01 5.38394272e-01 -4.06438500e-01 -1.59867510e-01 -4.06464875e-01 7.92297304e-01 4.96354640e-01 9.33049023e-01 2.12914288e-01 -5.78866363e-01 4.36449647e-01 2.36921966e-01 4.87653285e-01 -4.43081886e-01 -5.51207423e-01 2.89226826e-02 -3.96664292e-02 -7.97302186e-01 -6.55524671e-01 -1.01259112e+00 -7.10311830e-01 -6.93602085e-01 -4.29251283e-01 -3.31047893e-01 2.42262051e-01 7.86354303e-01 7.41534352e-01 4.20457780e-01 5.63853264e-01 -7.17392921e-01 -2.61442780e-01 -1.26636577e+00 -3.37448806e-01 5.38940549e-01 6.05701625e-01 -8.86301100e-01 -5.35188615e-01 3.11879873e-01]
[14.114008903503418, 0.0430297777056694]
e50e0cd5-bda3-4be3-8cfd-6d2e79f63e53
vertibench-advancing-feature-distribution
2307.02040
null
https://arxiv.org/abs/2307.02040v1
https://arxiv.org/pdf/2307.02040v1.pdf
VertiBench: Advancing Feature Distribution Diversity in Vertical Federated Learning Benchmarks
Vertical Federated Learning (VFL) is a crucial paradigm for training machine learning models on feature-partitioned, distributed data. However, due to privacy restrictions, few public real-world VFL datasets exist for algorithm evaluation, and these represent a limited array of feature distributions. Existing benchmarks often resort to synthetic datasets, derived from arbitrary feature splits from a global set, which only capture a subset of feature distributions, leading to inadequate algorithm performance assessment. This paper addresses these shortcomings by introducing two key factors affecting VFL performance - feature importance and feature correlation - and proposing associated evaluation metrics and dataset splitting methods. Additionally, we introduce a real VFL dataset to address the deficit in image-image VFL scenarios. Our comprehensive evaluation of cutting-edge VFL algorithms provides valuable insights for future research in the field.
['Bingsheng He', 'Junyi Hou', 'Zhaomin Wu']
2023-07-05
null
null
null
null
['feature-importance']
['methodology']
[-2.02152021e-02 -4.93086606e-01 -6.36993527e-01 -4.79347944e-01 -9.78763163e-01 -8.28045249e-01 3.50838035e-01 4.32289928e-01 -1.75988093e-01 9.54791665e-01 1.76585630e-01 -1.55573189e-01 -4.76394564e-01 -8.90066683e-01 -4.38022405e-01 -7.26243615e-01 -2.08602816e-01 1.81701303e-01 -3.01399291e-01 1.87234521e-01 2.11007297e-02 5.45588195e-01 -1.85440612e+00 6.47044778e-01 8.41489613e-01 1.16087866e+00 -3.80562752e-01 2.67765909e-01 -3.37815344e-01 6.46463394e-01 -6.33765638e-01 -5.18779397e-01 8.18247676e-01 -3.61639172e-01 -7.74664104e-01 -8.61323699e-02 3.37566584e-01 -1.92954078e-01 -1.12457812e-01 1.04280758e+00 5.26967406e-01 8.49814154e-04 3.83357733e-01 -1.87010908e+00 -4.77201581e-01 6.22629344e-01 -3.54453206e-01 -9.36418474e-02 4.08677310e-01 3.96566182e-01 1.09852016e+00 -7.62520254e-01 9.88953769e-01 7.84131944e-01 7.50704169e-01 3.88150692e-01 -1.21901751e+00 -6.63448334e-01 -5.58447973e-05 2.19517320e-01 -1.55910480e+00 -3.38580489e-01 9.07802820e-01 -3.48973960e-01 7.16255188e-01 6.86953008e-01 5.32359004e-01 9.96333838e-01 2.82801539e-01 8.24585736e-01 1.21151841e+00 -2.27130920e-01 4.09297675e-01 3.21969539e-01 1.73540741e-01 4.25350875e-01 7.17368722e-01 1.52841866e-01 -9.91841912e-01 -6.57662868e-01 2.18931511e-01 1.11105554e-01 -3.77758563e-01 -1.16606915e+00 -1.18657339e+00 7.88586855e-01 2.75445998e-01 8.54215994e-02 -2.64196754e-01 -4.44827676e-01 8.88997436e-01 6.51420057e-01 3.53372008e-01 3.57890785e-01 -7.49061823e-01 1.91498920e-02 -9.20359135e-01 4.53664958e-01 7.98910439e-01 8.57468843e-01 1.08069277e+00 -3.38059396e-01 -3.21300805e-01 5.74723899e-01 1.63054854e-01 2.35134244e-01 4.96740133e-01 -9.21015441e-01 5.01039386e-01 8.74283373e-01 1.52325138e-01 -1.15756917e+00 -2.67887712e-01 -5.37636876e-01 -7.73369253e-01 4.70506512e-02 5.32341838e-01 -6.94661662e-02 -4.08772469e-01 1.61233723e+00 6.38370454e-01 -2.72106320e-01 1.35419115e-01 9.87164915e-01 8.85616302e-01 7.52028376e-02 -2.33664900e-01 -2.92105138e-01 1.07414603e+00 -8.32640290e-01 -6.03208125e-01 3.23150158e-01 9.23198283e-01 -5.91689050e-01 1.20312750e+00 5.67134738e-01 -7.32584715e-01 -2.11119294e-01 -8.36275339e-01 2.12293759e-01 -6.79061532e-01 -9.10538286e-02 9.56682265e-01 9.73787487e-01 -8.98448467e-01 4.34435874e-01 -5.52500129e-01 -2.85259664e-01 8.77880096e-01 2.93250620e-01 -6.35305047e-01 -3.39464575e-01 -9.95418429e-01 2.75205135e-01 8.94382000e-02 -1.39998019e-01 -8.58343303e-01 -9.65504646e-01 -6.00063086e-01 -5.89816831e-02 1.97745919e-01 -9.37896252e-01 9.90385830e-01 -8.24221909e-01 -9.09443498e-01 8.17795813e-01 6.07780851e-02 -5.55806398e-01 8.80422592e-01 1.16396844e-02 -3.75803143e-01 -9.61133689e-02 1.12238869e-01 2.28018448e-01 8.04656804e-01 -1.33003998e+00 -6.37913227e-01 -5.04367650e-01 -3.40309460e-03 -2.64590625e-02 -5.59618533e-01 -1.68150142e-01 -7.78568238e-02 -5.66824317e-01 -1.57360092e-01 -4.89560276e-01 -4.88967240e-01 3.92159857e-02 -5.18805027e-01 -5.84660284e-02 8.51270795e-01 6.59648329e-02 1.30370867e+00 -2.25236845e+00 -3.31221431e-01 3.57291102e-01 4.40965623e-01 1.28751704e-02 -1.18571885e-01 7.54571199e-01 1.88461080e-01 3.05093646e-01 -1.01655655e-01 -1.68862447e-01 -8.33132789e-02 2.59626303e-02 -2.62709022e-01 8.06528270e-01 -1.99509040e-01 7.79409945e-01 -9.00471330e-01 -6.88394487e-01 2.08753511e-01 2.48376787e-01 -6.42695844e-01 -3.49007025e-02 -2.24769637e-01 4.02067602e-01 -6.86337054e-01 8.72021377e-01 9.20928836e-01 -1.42079756e-01 7.57626370e-02 -3.99080753e-01 -2.59955991e-02 -2.70649254e-01 -1.24529076e+00 1.94615388e+00 -2.46846437e-01 4.06918943e-01 6.75856471e-02 -7.71782458e-01 8.86923373e-01 1.48972258e-01 1.10083795e+00 -3.11205983e-01 3.85084838e-01 2.05986962e-01 -1.80984542e-01 -4.10805017e-01 3.78364712e-01 2.64314860e-01 -1.85422629e-01 6.22646391e-01 -8.26503057e-03 1.79063618e-01 1.34559959e-01 1.61059961e-01 1.32148659e+00 -1.68583483e-01 2.87754476e-01 -2.65529007e-01 4.79346514e-01 2.46149823e-01 7.56609499e-01 8.60187531e-01 -5.91415763e-01 7.23212659e-01 3.15720677e-01 -9.13627565e-01 -6.20399356e-01 -9.43204284e-01 -3.64988983e-01 8.27275872e-01 4.28661287e-01 -8.28695118e-01 -6.33269966e-01 -1.08065200e+00 3.89873892e-01 4.63865608e-01 -6.71433747e-01 -2.78905749e-01 -1.33546144e-01 -7.30971098e-01 6.30237043e-01 2.39498764e-01 4.56181914e-01 -1.00941586e+00 -9.48317647e-01 -1.22590899e-01 -3.88566479e-02 -7.20816255e-01 -2.54362613e-01 4.75701280e-02 -6.66570902e-01 -1.53808355e+00 -1.32877588e-01 -4.17359620e-01 5.79335868e-01 3.46902758e-01 1.31312811e+00 2.90196538e-02 -6.41511738e-01 3.41520399e-01 -3.48197460e-01 -3.52710932e-01 -1.32509455e-01 1.81485236e-01 3.65593284e-02 3.56158406e-01 5.55653274e-01 -2.76302367e-01 -7.84196913e-01 3.46847296e-01 -8.27630460e-01 -8.98358822e-02 4.01784658e-01 1.00184393e+00 7.78555155e-01 1.52634323e-01 5.59087813e-01 -1.27225876e+00 6.58069253e-01 -7.09606171e-01 -4.89301443e-01 3.88504952e-01 -8.34732175e-01 -1.40579477e-01 7.85529315e-01 -8.46724063e-02 -9.61161017e-01 1.12375282e-01 2.04287902e-01 -7.22165883e-01 -4.05104220e-01 4.68266964e-01 -4.84548330e-01 -1.99127525e-01 8.32633376e-01 2.10453331e-01 1.95369303e-01 -5.51725686e-01 5.09633601e-01 6.59729660e-01 3.55097681e-01 -7.19819546e-01 5.83192110e-01 6.13136172e-01 -7.64214844e-02 -2.74922192e-01 -4.44663942e-01 -4.34157908e-01 -4.28848028e-01 -5.48400767e-02 5.35847433e-02 -7.76308298e-01 -7.01429009e-01 3.23651999e-01 -6.25027239e-01 7.91355222e-02 -9.56095517e-01 4.15138572e-01 -5.79804778e-01 8.61823931e-02 -2.69814849e-01 -4.75908160e-01 -6.12867057e-01 -1.18204260e+00 7.79713929e-01 4.39181775e-02 -1.20426320e-01 -6.27300620e-01 1.77292570e-01 2.88204402e-01 6.74806952e-01 6.37834132e-01 7.95051634e-01 -6.48075759e-01 -4.55391645e-01 -5.24387717e-01 3.74018331e-03 1.47274524e-01 3.82696599e-01 1.24593079e-01 -1.00897908e+00 -7.22627163e-01 -9.98489633e-02 -6.65286481e-01 6.39490068e-01 2.54137427e-01 1.52313459e+00 -2.43382201e-01 -6.44041538e-01 9.45686579e-01 1.53974330e+00 -1.64919451e-01 8.85696560e-02 3.72970700e-01 4.54228818e-01 4.78516430e-01 9.01221693e-01 9.36896861e-01 2.56515741e-01 2.74074823e-01 6.04378641e-01 -3.42060300e-03 -1.48061588e-01 -2.46371642e-01 -1.13995932e-01 3.20504397e-01 3.91875416e-01 -4.46776092e-01 -9.84708250e-01 6.17763579e-01 -1.73732400e+00 -5.91081023e-01 3.50511894e-02 2.29042888e+00 5.57967365e-01 -2.91818798e-01 1.72154546e-01 2.18385130e-01 6.24864936e-01 1.99403629e-01 -9.57508683e-01 -2.32314408e-01 -4.06972677e-01 -1.20685846e-01 5.73750794e-01 -1.12021565e-01 -1.18191373e+00 5.22546470e-01 6.53025913e+00 8.03000808e-01 -1.29516017e+00 2.22022593e-01 5.45972288e-01 -3.79230410e-01 -6.16858304e-01 -1.86787859e-01 -4.61671889e-01 3.71980518e-01 6.61909461e-01 -7.40815401e-01 2.24598750e-01 9.88268733e-01 -1.52387008e-01 3.32210690e-01 -1.08990836e+00 1.22120476e+00 -1.81216404e-01 -1.50524580e+00 2.25162536e-01 2.02922091e-01 8.14709425e-01 4.28747267e-01 1.48988485e-01 7.86782950e-02 3.56980860e-01 -1.00706017e+00 4.58415389e-01 1.28080606e-01 1.09534740e+00 -8.49523842e-01 6.89626336e-01 3.06160063e-01 -1.16273856e+00 -3.22190076e-01 -2.43511006e-01 1.26702964e-01 -3.53732705e-01 6.57689154e-01 -5.03882110e-01 9.94523525e-01 9.92610812e-01 6.40070677e-01 -6.70070767e-01 1.33124232e+00 3.52034450e-01 5.97199798e-01 -3.05574477e-01 1.69349656e-01 -6.51663244e-02 5.22250645e-02 4.97312486e-01 8.73422801e-01 9.03613046e-02 -3.37069273e-01 2.29409978e-01 5.29265583e-01 -2.33152285e-01 6.44072711e-01 -1.03533173e+00 6.45048693e-02 8.06904137e-01 1.39241147e+00 -6.86111450e-01 2.31344298e-01 -7.37049878e-01 5.20321250e-01 3.34021479e-01 2.06150576e-01 -6.54387772e-01 -2.16076508e-01 1.13493061e+00 1.07444368e-01 1.27993450e-02 1.84317961e-01 -2.61422217e-01 -1.27942014e+00 5.71399890e-02 -1.17921925e+00 8.78175437e-01 1.46383364e-02 -1.72494876e+00 8.58163655e-01 -1.06926024e-01 -1.66167057e+00 -1.67703494e-01 -1.59604222e-01 -2.92508453e-01 4.77968514e-01 -1.30516672e+00 -1.33623838e+00 -6.26869798e-01 1.01964450e+00 3.20081055e-01 -4.34578091e-01 1.05196691e+00 2.84653038e-01 -6.04333043e-01 1.08650041e+00 5.02148509e-01 -7.58384988e-02 7.78925061e-01 -8.13960135e-01 1.64804786e-01 5.14020681e-01 3.26789021e-01 6.31888390e-01 5.32541394e-01 -3.97722572e-01 -1.83020604e+00 -1.39573216e+00 3.49575162e-01 -5.41663647e-01 2.14947701e-01 -4.07946944e-01 -7.52435148e-01 4.85196531e-01 1.18642762e-01 8.96906495e-01 1.10944164e+00 7.35775754e-02 -2.71369308e-01 -4.21089113e-01 -1.71706951e+00 2.35439137e-01 1.19665825e+00 -2.40011439e-01 1.03626959e-03 2.84997731e-01 4.94018048e-01 -3.22119951e-01 -1.06198740e+00 5.50004900e-01 6.10430777e-01 -1.29338050e+00 8.13254893e-01 -7.86287069e-01 1.23308323e-01 -3.28868061e-01 -3.66203487e-01 -1.21415699e+00 -1.77320510e-01 -6.27809227e-01 -9.94130224e-02 1.45959878e+00 1.03976406e-01 -9.76123154e-01 1.27016103e+00 4.74908292e-01 3.31647784e-01 -1.06529415e+00 -1.04152596e+00 -7.66545594e-01 -6.73992708e-02 -4.73489910e-01 1.37111521e+00 1.21649241e+00 -1.37677595e-01 -3.57925147e-01 -1.31187454e-01 -1.32964879e-01 8.59080136e-01 7.33060002e-01 1.00503147e+00 -1.17671561e+00 1.63269714e-02 -3.24123353e-01 -4.84274924e-01 -2.64768779e-01 2.41132393e-01 -1.24241138e+00 -3.58073324e-01 -1.11809313e+00 2.33731508e-01 -9.67686474e-01 -6.87202811e-01 5.63681722e-01 1.70780927e-01 2.81491935e-01 1.21956132e-01 2.93099254e-01 -7.23544419e-01 5.25310278e-01 1.01720965e+00 -2.27548778e-01 -7.82910362e-02 7.28111118e-02 -5.97968102e-01 5.13724089e-01 9.44375038e-01 -5.30571878e-01 -7.79632092e-01 -3.67448181e-01 -4.48876955e-02 -8.97473916e-02 2.03571945e-01 -7.78107762e-01 3.49605590e-01 -3.93457234e-01 3.65387112e-01 -6.19882941e-01 -1.72372520e-01 -1.07935226e+00 5.08219123e-01 3.76899987e-01 -2.97314405e-01 2.37134155e-02 8.51150379e-02 5.46651185e-01 -4.75063294e-01 2.21007064e-01 4.05875742e-01 -1.08051725e-01 -6.86548948e-01 5.63335001e-01 -3.94698009e-02 2.29663923e-01 1.40881908e+00 -3.22301328e-01 -3.34845006e-01 -5.01260944e-02 -2.75142610e-01 3.42173159e-01 9.08089101e-01 4.85233247e-01 5.69273114e-01 -1.30946159e+00 -7.47407138e-01 8.51784050e-01 3.55907947e-01 9.54665691e-02 3.20202410e-01 6.06777787e-01 -5.03359854e-01 4.08204705e-01 -3.48676950e-01 -6.57288194e-01 -1.22769463e+00 8.82303894e-01 2.88816065e-01 -1.92194596e-01 -6.35372400e-01 8.29437912e-01 -5.98054081e-02 -6.87134147e-01 2.19464868e-01 5.17575666e-02 2.11027130e-01 5.46312658e-03 3.33332360e-01 5.64811170e-01 4.33719367e-01 -6.12681389e-01 -7.72919476e-01 2.75939792e-01 3.28512266e-02 3.59818339e-01 1.17624009e+00 -1.52161390e-01 -6.53896760e-03 1.43103227e-01 1.30459738e+00 1.75731033e-01 -9.50068355e-01 -2.60390639e-01 -1.95309758e-01 -1.00585938e+00 1.36782564e-02 -7.09949136e-01 -1.47605753e+00 5.25234103e-01 7.29682684e-01 1.48149997e-01 1.38779795e+00 -1.96543619e-01 6.43777370e-01 1.09551713e-01 9.88982618e-01 -6.60757542e-01 -4.99562413e-01 -9.27984938e-02 6.50003433e-01 -1.27592254e+00 1.55557692e-01 -5.27936518e-01 -4.47966695e-01 9.04397011e-01 6.69940233e-01 2.94157177e-01 8.12180758e-01 5.15474260e-01 2.99203873e-01 -2.08789617e-01 -9.41665351e-01 2.07404837e-01 -4.93556187e-02 5.03066301e-01 2.28729665e-01 -2.05340348e-02 -4.53016728e-01 8.52871954e-01 -3.08649838e-01 3.01925331e-01 1.44076988e-01 1.09584165e+00 2.91777194e-01 -1.37675035e+00 -3.35204750e-01 9.64075685e-01 -5.64140558e-01 2.99793184e-01 -4.34045583e-01 6.90579772e-01 3.77013892e-01 8.92520487e-01 -1.58797264e-01 -5.02843499e-01 2.70792902e-01 -1.35100782e-01 3.35131258e-01 -2.13925183e-01 -9.16397870e-01 -3.92610699e-01 -2.86868960e-01 -8.45103085e-01 -4.60330062e-02 -8.15676570e-01 -1.04182935e+00 -3.58166546e-01 -4.31497127e-01 4.05766159e-01 4.77207243e-01 3.55668902e-01 8.05433750e-01 2.46198684e-01 9.81910586e-01 -5.47468901e-01 -8.20455253e-01 -3.95120829e-01 -7.53837466e-01 7.77379990e-01 3.41787368e-01 -5.97386718e-01 -2.97835439e-01 -2.86990643e-01]
[6.070060729980469, 6.3784918785095215]
bac2be6b-3b48-40b0-a750-bd486b80d3d6
exploring-neural-methods-for-parsing
1810.12579
null
http://arxiv.org/abs/1810.12579v1
http://arxiv.org/pdf/1810.12579v1.pdf
Exploring Neural Methods for Parsing Discourse Representation Structures
Neural methods have had several recent successes in semantic parsing, though they have yet to face the challenge of producing meaning representations based on formal semantics. We present a sequence-to-sequence neural semantic parser that is able to produce Discourse Representation Structures (DRSs) for English sentences with high accuracy, outperforming traditional DRS parsers. To facilitate the learning of the output, we represent DRSs as a sequence of flat clauses and introduce a method to verify that produced DRSs are well-formed and interpretable. We compare models using characters and words as input and see (somewhat surprisingly) that the former performs better than the latter. We show that eliminating variable names from the output using De Bruijn-indices increases parser performance. Adding silver training data boosts performance even further.
['Rik van Noord', 'Antonio Toral', 'Johan Bos', 'Lasha Abzianidze']
2018-10-30
exploring-neural-methods-for-parsing-1
https://aclanthology.org/Q18-1043
https://aclanthology.org/Q18-1043.pdf
tacl-2018-1
['drs-parsing']
['natural-language-processing']
[ 5.62470078e-01 9.07758594e-01 -3.55642140e-02 -7.21652210e-01 -9.18414354e-01 -9.65202570e-01 4.98296469e-01 2.41289809e-01 -2.21149907e-01 9.18526053e-01 6.89051867e-01 -7.88123667e-01 2.87172645e-01 -1.09419525e+00 -7.63382673e-01 -2.41700649e-01 1.14619128e-01 6.01584315e-01 3.25369388e-01 -4.90208656e-01 3.00084293e-01 5.25916815e-02 -1.33300090e+00 7.55798697e-01 6.52525961e-01 3.79826456e-01 3.25510591e-01 7.51381636e-01 -6.86589062e-01 1.19601524e+00 -9.63432372e-01 -4.71553147e-01 -2.02106088e-01 -5.96912920e-01 -1.45604849e+00 -4.59812582e-01 -8.12878250e-04 -1.49792522e-01 -3.49921882e-02 8.64434898e-01 4.22177050e-04 -8.22797269e-02 4.76445585e-01 -7.29203701e-01 -8.93268347e-01 1.13236082e+00 -1.37892216e-01 2.51690120e-01 6.61007404e-01 -2.15156391e-01 1.59580183e+00 -3.72131407e-01 1.16466749e+00 1.65848064e+00 5.34510493e-01 9.96062040e-01 -1.35984588e+00 -8.85643363e-02 2.05285341e-01 -7.56357536e-02 -9.53081846e-01 -4.23165083e-01 4.41462129e-01 -1.15299933e-01 1.81067169e+00 1.49814948e-01 2.83969641e-01 9.92399275e-01 -1.10339023e-01 7.55790532e-01 8.15979898e-01 -7.06076562e-01 3.62473488e-01 -3.64650011e-01 5.71580946e-01 8.63338411e-01 2.99507797e-01 -4.10081036e-02 -4.03470576e-01 2.44098771e-02 6.33168995e-01 -7.75185287e-01 1.07654640e-02 6.57129586e-02 -1.01470113e+00 1.40825427e+00 3.53603631e-01 5.45644641e-01 -1.42377213e-01 4.47328836e-01 7.29998052e-01 2.32852206e-01 1.54042974e-01 5.76187134e-01 -5.40360689e-01 -1.55085698e-01 -4.86677915e-01 3.43985140e-01 1.05248523e+00 7.73556292e-01 3.07852149e-01 2.54081011e-01 -6.09510615e-02 7.26031363e-01 -1.87328681e-02 1.74774081e-01 6.42956913e-01 -1.20205557e+00 5.68000376e-01 4.76661533e-01 9.59505606e-03 -6.99724257e-01 -3.45237672e-01 4.26855236e-02 -1.99235409e-01 -9.97892320e-02 6.15117431e-01 -2.36296833e-01 -8.74096513e-01 2.11597586e+00 -9.46205668e-03 -3.03614765e-01 7.58816898e-01 7.31468737e-01 8.77632558e-01 9.60214317e-01 6.11909747e-01 3.56952026e-02 1.58091354e+00 -6.00596130e-01 -6.29431546e-01 -7.43980050e-01 1.22031784e+00 -7.52848759e-02 9.78942394e-01 2.72213612e-02 -1.45758033e+00 -2.29882598e-01 -1.17883766e+00 -5.16732693e-01 -3.10555637e-01 -2.45488748e-01 9.98111486e-01 5.46752691e-01 -1.19325733e+00 6.66719496e-01 -8.16297472e-01 -2.46904895e-01 3.45741510e-01 3.09318513e-01 -4.39894229e-01 1.63493857e-01 -1.39576328e+00 1.22328627e+00 9.21925724e-01 -1.92530826e-01 -3.99859041e-01 -1.49005398e-01 -1.40768743e+00 2.03299791e-01 3.19748312e-01 -8.11730027e-01 1.79201007e+00 -1.35525227e+00 -1.36539030e+00 1.36337793e+00 -5.49770117e-01 -8.60551357e-01 -1.69530571e-01 -1.66944310e-01 -8.71001109e-02 3.36084992e-01 1.38575464e-01 8.87813330e-01 1.51156381e-01 -1.05028617e+00 -6.38594210e-01 -4.08338308e-01 5.25465727e-01 1.99448392e-01 2.10293710e-01 4.95634973e-01 2.63810214e-02 -4.84328210e-01 3.85403752e-01 -6.06032252e-01 -2.19108731e-01 -3.59611869e-01 -5.02786517e-01 -5.57282269e-01 1.99848741e-01 -9.23402429e-01 1.00046456e+00 -1.97755194e+00 2.44896755e-01 3.11608948e-02 1.05317436e-01 1.98021948e-01 -2.24533439e-01 2.25276589e-01 -3.31772983e-01 5.62654912e-01 -5.39006412e-01 -6.92152372e-03 1.68452069e-01 7.99145997e-01 -4.61607516e-01 -6.51800185e-02 8.22075784e-01 1.17286730e+00 -9.05513406e-01 -3.60230505e-01 -9.26603153e-02 3.47271487e-02 -6.26449943e-01 2.21829697e-01 -6.11246049e-01 6.35980349e-03 -5.69502473e-01 2.37007722e-01 1.19699866e-01 -2.70283520e-01 8.46780002e-01 5.04916251e-01 1.20878778e-01 1.21824694e+00 -8.53050292e-01 1.57013774e+00 -4.32989895e-01 6.59817934e-01 1.01436153e-01 -1.40730393e+00 9.05160427e-01 3.30650926e-01 -3.71386766e-01 -7.21163988e-01 9.22386646e-02 3.47035795e-01 2.63303816e-01 -4.34283257e-01 5.91570199e-01 -7.86072731e-01 -4.43860203e-01 4.64092284e-01 4.96522337e-02 -1.60689920e-01 2.95714647e-01 3.94758016e-01 1.19354784e+00 4.34245616e-01 6.47358179e-01 -2.88160652e-01 4.92766112e-01 5.78089178e-01 5.87083578e-01 6.27619445e-01 2.81162381e-01 4.93338615e-01 1.03344107e+00 -4.88962352e-01 -1.25320125e+00 -1.18624592e+00 2.17788339e-01 1.22102427e+00 -6.70040026e-02 -4.67195928e-01 -1.15697134e+00 -6.63864672e-01 -2.76095629e-01 1.35198951e+00 -5.05824387e-01 2.82651260e-02 -1.20367455e+00 -4.60493922e-01 9.66194212e-01 1.09297228e+00 -5.46822958e-02 -1.54241943e+00 -8.48793924e-01 4.35307443e-01 -2.45675132e-01 -1.15637732e+00 4.34851885e-01 6.51385963e-01 -9.56343353e-01 -9.57317591e-01 -1.48271531e-01 -1.21623409e+00 6.15667820e-01 -2.40612790e-01 1.39009714e+00 4.65744317e-01 -2.93007661e-02 -1.88325942e-01 -3.37497175e-01 -2.83956885e-01 -1.04196382e+00 1.30821094e-01 -5.09698689e-01 -9.71824288e-01 6.38094306e-01 -3.65908980e-01 1.74690753e-01 -5.00869632e-01 -8.73247206e-01 2.91574955e-01 2.31910929e-01 9.26541924e-01 2.44959369e-01 -4.24578249e-01 6.54325068e-01 -1.42335474e+00 7.86066890e-01 -4.33525920e-01 -5.06007493e-01 6.87444806e-02 -1.67877704e-01 7.58167803e-01 8.20742071e-01 1.33134872e-01 -1.16231310e+00 3.07822227e-03 -6.40040696e-01 5.90071857e-01 -4.24416989e-01 6.30679190e-01 -5.79569722e-04 7.21045375e-01 9.07084286e-01 -6.22300617e-02 1.49388954e-01 -3.72135699e-01 6.02739751e-01 4.19394433e-01 7.79744685e-01 -1.01708174e+00 6.71181917e-01 1.05734251e-01 -1.37453198e-01 -6.65253818e-01 -1.13238037e+00 1.66821852e-02 -5.04256487e-01 6.32678449e-01 1.17904663e+00 -8.51052225e-01 -4.55627292e-01 -6.21483177e-02 -1.72035980e+00 -4.20132935e-01 -4.61869448e-01 -6.06122762e-02 -8.07464719e-01 3.58808696e-01 -1.01194286e+00 -7.20777333e-01 -3.90672147e-01 -7.61412203e-01 8.73697102e-01 7.79683292e-02 -8.12460423e-01 -1.02773750e+00 -1.68224216e-01 1.44573495e-01 2.59960532e-01 3.00646305e-01 1.67433000e+00 -1.11823845e+00 -2.87796080e-01 2.05887809e-01 -3.10621023e-01 3.59601557e-01 -3.06959689e-01 -2.16372296e-01 -8.70567858e-01 3.10392916e-01 -1.04656801e-01 -5.07467866e-01 6.56794548e-01 3.11432034e-01 9.75798130e-01 -3.50669652e-01 -1.92936569e-01 2.43158907e-01 1.39846706e+00 2.02095300e-01 9.26674783e-01 6.87315822e-01 2.67148852e-01 9.81200576e-01 2.38606960e-01 -1.64590210e-01 5.19747138e-01 2.23388821e-01 2.10490376e-01 8.00916329e-02 -6.86678216e-02 -4.71739948e-01 5.27406156e-01 4.71411735e-01 1.71015382e-01 -2.06823096e-01 -1.01160061e+00 6.89437449e-01 -1.66673362e+00 -1.06014025e+00 -3.47100735e-01 1.60558867e+00 1.10154343e+00 4.49834347e-01 -1.18623868e-01 1.33591488e-01 6.92508101e-01 2.68155783e-01 -1.68400392e-01 -1.15305829e+00 -2.46708348e-01 9.17143583e-01 2.72995591e-01 7.97812760e-01 -7.28825271e-01 1.56138158e+00 7.40013504e+00 2.38018930e-01 -6.73500299e-01 8.69474001e-03 4.67093706e-01 3.72462630e-01 -6.41052246e-01 2.21520543e-01 -6.10295415e-01 -6.32276759e-02 1.34179974e+00 -8.18656981e-02 3.81576866e-01 9.80397046e-01 -9.28825587e-02 -1.78820327e-01 -1.22393548e+00 4.19924200e-01 -1.72030941e-01 -1.71883905e+00 1.42954454e-01 -5.44243395e-01 1.60956144e-01 -1.26693323e-01 -6.67570174e-01 3.81413102e-01 8.56184065e-01 -1.41616154e+00 8.86316359e-01 -9.17093270e-03 5.97037137e-01 -5.58815539e-01 7.95973957e-01 2.78875232e-01 -9.23075080e-01 -8.79565105e-02 -4.37918603e-01 -5.07330775e-01 3.79176646e-01 -4.01945673e-02 -8.14493418e-01 4.61574376e-01 1.90826148e-01 4.15649712e-01 -5.37157118e-01 2.40527645e-01 -8.85294020e-01 6.77232385e-01 -1.44095719e-01 -4.52781737e-01 4.63341892e-01 -3.93413976e-02 3.61550391e-01 1.53988230e+00 8.95029381e-02 5.37792444e-01 -1.40016228e-01 1.02264369e+00 -8.10748665e-04 -5.83539344e-02 -6.60122871e-01 -2.82202303e-01 4.68687505e-01 7.11064935e-01 -7.15293646e-01 -6.66284502e-01 -3.22151810e-01 9.21155035e-01 6.80288076e-01 1.52839646e-01 -4.00995493e-01 -4.45083529e-01 6.38579130e-01 -2.84595676e-02 3.95629555e-01 -3.04376483e-01 -4.76403058e-01 -9.32855189e-01 8.05736333e-02 -8.57086599e-01 4.94268239e-01 -1.16833460e+00 -1.07387042e+00 5.42286277e-01 4.21729162e-02 -5.79477549e-02 -8.45996082e-01 -1.09599662e+00 -5.90784073e-01 9.96638417e-01 -1.44007087e+00 -8.02407146e-01 2.90323853e-01 1.11438967e-01 5.69937110e-01 8.13024864e-02 1.30472457e+00 -3.41859311e-01 -2.33544439e-01 2.50949800e-01 -3.11211556e-01 6.24259531e-01 -8.12078342e-02 -1.52094460e+00 1.01704979e+00 8.16164553e-01 2.16899261e-01 9.97150421e-01 8.73181760e-01 -4.60412711e-01 -1.24546576e+00 -6.70722842e-01 1.33891845e+00 -4.67204720e-01 7.32277155e-01 -3.25262517e-01 -1.04980886e+00 1.05566502e+00 2.74141580e-01 -3.94613743e-01 4.94444013e-01 4.47081655e-01 -6.32538795e-01 5.58571517e-01 -9.04659867e-01 5.41317225e-01 1.19494712e+00 -8.05261731e-01 -1.67601836e+00 1.92018941e-01 1.00570774e+00 -5.34472942e-01 -4.83584404e-01 4.96788993e-02 2.51713902e-01 -8.23099911e-01 8.34826052e-01 -1.20573092e+00 8.32653344e-01 -2.62144804e-01 -3.30997437e-01 -1.09330034e+00 -1.18463941e-01 -3.85037661e-01 8.19396004e-02 1.02323782e+00 6.50412977e-01 -6.06545150e-01 7.55070210e-01 5.18295228e-01 -4.30047095e-01 -4.52679664e-01 -8.91417801e-01 -6.39916420e-01 6.20974064e-01 -4.02855545e-01 4.68723089e-01 7.47514784e-01 4.39305604e-01 1.02594638e+00 3.82253498e-01 -8.21247101e-02 1.70389771e-01 1.80183113e-01 2.36350045e-01 -1.33656180e+00 -3.55155945e-01 -5.39034605e-01 -3.04721624e-01 -1.08235037e+00 7.60441124e-01 -1.26437032e+00 3.60711783e-01 -1.93953598e+00 1.24950081e-01 -2.63431340e-01 2.31494978e-01 8.12644541e-01 -1.18043415e-01 -2.66515799e-02 3.04078925e-02 -2.39126757e-01 -2.71784753e-01 9.12234187e-02 6.55159414e-01 1.15440123e-01 1.21828042e-01 -6.45628512e-01 -1.18918467e+00 8.67240667e-01 9.57229316e-01 -5.23958564e-01 -2.45088905e-01 -7.66094506e-01 4.11923736e-01 2.47965232e-01 2.38885522e-01 -5.78216136e-01 -2.12872326e-01 -1.47669688e-01 1.55754700e-01 -2.11468041e-01 1.32920042e-01 -5.39932251e-02 -2.05312043e-01 4.87757653e-01 -7.84899533e-01 4.08647239e-01 2.94998229e-01 1.89429268e-01 -2.47147530e-01 -8.60013723e-01 6.79768085e-01 -5.63299358e-01 -1.02191806e+00 -5.51306605e-01 -5.85679829e-01 5.33372045e-01 6.51244342e-01 -2.34515145e-01 -5.37331820e-01 -1.33713469e-01 -7.63654292e-01 -2.39664651e-02 5.39815068e-01 3.55615944e-01 4.50657070e-01 -7.12585270e-01 -5.98163605e-01 3.60156852e-03 -4.42978404e-02 1.52413875e-01 -2.88941205e-01 -3.33424052e-03 -8.93581986e-01 5.75203419e-01 -1.52536690e-01 -1.12143137e-01 -1.21077251e+00 4.02919680e-01 3.05731207e-01 -3.09451014e-01 -8.80883992e-01 8.78294945e-01 1.41424760e-01 -5.59123933e-01 2.51260065e-02 -4.44144338e-01 -2.48713136e-01 -2.82878369e-01 6.16326630e-01 -2.93198615e-01 -1.07381143e-01 -5.92760205e-01 -3.97858530e-01 1.39365911e-01 4.78338227e-02 -3.42207640e-01 1.58618426e+00 2.15281442e-01 -2.74020851e-01 2.73120195e-01 9.22788203e-01 -1.42043546e-01 -7.64466405e-01 7.77145196e-03 7.54724085e-01 1.46903560e-01 -2.88226485e-01 -7.29562819e-01 -4.88150150e-01 8.57188821e-01 -2.45645851e-01 4.29272443e-01 5.99775970e-01 4.01123464e-01 9.61316347e-01 6.73933744e-01 1.85084209e-01 -1.11650062e+00 -1.68469205e-01 1.06485963e+00 4.40850526e-01 -7.44417846e-01 -2.90935010e-01 -5.63059032e-01 -7.84334183e-01 1.25894225e+00 3.00500482e-01 -3.56884778e-01 -1.60097808e-01 4.12860215e-01 -5.47416955e-02 -3.50290984e-01 -7.90501118e-01 -3.34706306e-01 -2.51783758e-01 7.72307754e-01 7.71734953e-01 2.00711921e-01 -6.59503162e-01 7.67734528e-01 -6.62547708e-01 -3.27539146e-01 6.16770267e-01 1.12220550e+00 -7.55334139e-01 -1.37572539e+00 -2.19391987e-01 3.01697314e-01 -7.06536114e-01 -5.27663767e-01 -5.40975749e-01 9.32091653e-01 -3.16504985e-01 9.12453830e-01 2.14068323e-01 9.74690765e-02 2.13282347e-01 7.08956957e-01 6.67903125e-01 -1.18702018e+00 -6.33311152e-01 -3.85631770e-01 1.05475700e+00 -5.19134343e-01 -3.65263253e-01 -6.42199755e-01 -2.27194715e+00 -2.92719752e-01 6.68440014e-02 3.70573521e-01 6.80111229e-01 1.22783422e+00 1.56140372e-01 4.96750265e-01 2.30865423e-02 -3.07303935e-01 -8.33301306e-01 -6.42490089e-01 -3.31343919e-01 4.36633348e-01 1.28613591e-01 -2.21371815e-01 -1.52407050e-01 2.37982094e-01]
[10.461100578308105, 9.262295722961426]
d6490856-d9e5-446c-916a-b3470c21c295
data-augmentation-for-compositional-data
2205.09906
null
https://arxiv.org/abs/2205.09906v1
https://arxiv.org/pdf/2205.09906v1.pdf
Data Augmentation for Compositional Data: Advancing Predictive Models of the Microbiome
Data augmentation plays a key role in modern machine learning pipelines. While numerous augmentation strategies have been studied in the context of computer vision and natural language processing, less is known for other data modalities. Our work extends the success of data augmentation to compositional data, i.e., simplex-valued data, which is of particular interest in the context of the human microbiome. Drawing on key principles from compositional data analysis, such as the Aitchison geometry of the simplex and subcompositions, we define novel augmentation strategies for this data modality. Incorporating our data augmentations into standard supervised learning pipelines results in consistent performance gains across a wide range of standard benchmark datasets. In particular, we set a new state-of-the-art for key disease prediction tasks including colorectal cancer, type 2 diabetes, and Crohn's disease. In addition, our data augmentations enable us to define a novel contrastive learning model, which improves on previous representation learning approaches for microbiome compositional data. Our code is available at https://github.com/cunningham-lab/AugCoDa.
['John P. Cunningham', 'Thomas P. Quinn', 'Elliott Gordon-Rodriguez']
2022-05-20
null
null
null
null
['disease-prediction']
['medical']
[ 8.35003257e-01 -1.42217636e-01 -3.76970500e-01 -1.97301939e-01 -3.02971721e-01 -5.61700463e-01 8.82794499e-01 8.86851013e-01 -2.43126526e-01 3.93374115e-01 6.57815933e-01 -5.50717652e-01 1.89854622e-01 -9.03470814e-01 -8.38591099e-01 -8.19958210e-01 -4.59852368e-01 4.29527134e-01 -4.05763209e-01 -2.82948136e-01 1.53674141e-01 2.33518779e-01 -1.43158960e+00 5.82258344e-01 7.37435997e-01 7.57604957e-01 -1.06214836e-01 6.33430839e-01 -3.38632971e-01 5.31441152e-01 -1.39130699e-03 6.81290627e-02 2.83095002e-01 -2.72147059e-01 -7.17012405e-01 -2.77273804e-01 4.25324559e-01 -2.22005546e-02 -1.79975748e-01 8.14314485e-01 4.11220282e-01 -1.58614114e-01 8.41571748e-01 -1.14142311e+00 -6.00420892e-01 5.12242496e-01 -7.08527088e-01 7.68228546e-02 1.36928111e-01 2.07776800e-01 1.07026803e+00 -6.67795420e-01 5.78291953e-01 1.33139944e+00 1.03531933e+00 4.00886089e-01 -1.48374689e+00 -3.34223568e-01 2.94920921e-01 -1.97976589e-01 -1.01848912e+00 -3.22629333e-01 3.16869289e-01 -8.99410963e-01 8.49401176e-01 2.94687778e-01 7.71251082e-01 1.18016899e+00 9.26423892e-02 9.85749364e-01 1.17790663e+00 -4.63392347e-01 1.55044720e-01 -3.14558446e-01 1.32720601e-02 8.44629705e-01 3.53420168e-01 8.92806128e-02 -4.77044672e-01 -3.98386091e-01 7.14510202e-01 4.12464529e-01 -1.52292103e-01 -5.48931956e-01 -1.66018260e+00 9.13562119e-01 4.41956729e-01 -1.07724167e-01 -4.20968503e-01 3.28889757e-01 6.63778186e-01 -7.92835355e-02 8.32363188e-01 6.74944758e-01 -7.13370502e-01 1.45315602e-01 -2.70772249e-01 1.53217182e-01 6.61408007e-01 6.55655921e-01 4.81554896e-01 -3.72896522e-01 -4.32392843e-02 9.02715206e-01 3.85355771e-01 3.17051977e-01 2.23805919e-01 -5.18950224e-01 1.27667665e-01 8.93832922e-01 -1.62482604e-01 -6.49129093e-01 -5.16149282e-01 -4.86175597e-01 -1.11693728e+00 -1.10992335e-01 4.53938425e-01 1.07588284e-01 -8.83546531e-01 1.82354653e+00 7.64878333e-01 4.34477955e-01 -1.33509994e-01 6.68426633e-01 7.80097008e-01 4.32111353e-01 1.74162909e-01 1.46141723e-02 1.67485404e+00 -8.02794337e-01 -4.16299790e-01 -1.28275812e-01 8.34286571e-01 -8.16385388e-01 1.08212340e+00 6.51344240e-01 -5.16171455e-01 -7.05051348e-02 -8.53645027e-01 -3.37335348e-01 -7.35187232e-01 -3.27870071e-01 1.19000363e+00 4.80122268e-01 -7.52600253e-01 2.70659506e-01 -1.09738541e+00 -6.40185356e-01 5.85268438e-01 1.56645983e-01 -3.23276907e-01 -3.94168913e-01 -8.41193736e-01 6.23355865e-01 6.96076304e-02 -1.40352532e-01 -7.24101007e-01 -1.29639733e+00 -9.13793325e-01 -5.00811636e-01 1.33823186e-01 -9.67478693e-01 1.22531891e+00 -4.33795571e-01 -1.05236804e+00 1.05486333e+00 -1.63499728e-01 -4.00962502e-01 3.69726896e-01 -3.32781106e-01 -2.40686890e-02 -1.33459434e-01 -1.35814726e-01 7.53949940e-01 3.27615082e-01 -1.10009205e+00 -3.69644612e-01 -4.70979422e-01 -8.22877418e-03 1.12716086e-01 -5.10237455e-01 -8.85141194e-02 -1.02380246e-01 -8.82749140e-01 -1.00135289e-01 -8.69609654e-01 -5.23922384e-01 4.30120021e-01 -4.81011093e-01 -1.03444695e-01 2.73310393e-01 -5.75810134e-01 9.05800641e-01 -1.89785457e+00 4.35825169e-01 1.07968755e-01 4.04669046e-01 -2.06602328e-02 -3.82039994e-01 6.17826223e-01 -1.44621059e-01 3.35709304e-01 -6.86675370e-01 -3.72640789e-01 -2.77841799e-02 1.21701658e-01 -1.43385842e-01 7.49184191e-01 5.62863171e-01 8.36807847e-01 -1.29526389e+00 -1.68428525e-01 3.91160548e-01 7.77515650e-01 -8.60092342e-01 8.08866173e-02 -8.46346438e-01 5.85470259e-01 -2.43001804e-01 8.43472898e-01 4.76437658e-01 -4.11478341e-01 2.15719849e-01 -2.63648510e-01 -3.60713303e-01 5.65637946e-01 -6.67729616e-01 1.97970891e+00 -3.68082792e-01 2.89759785e-01 1.49850458e-01 -9.07728195e-01 5.68422019e-01 1.97721124e-02 8.58453214e-01 -4.34212297e-01 9.91222411e-02 6.77306056e-02 1.25568390e-01 -4.40767199e-01 2.72349864e-01 -9.18615907e-02 1.37983352e-01 6.34035349e-01 -3.80547553e-01 -1.13872916e-01 2.46195346e-01 1.83306202e-01 1.17793107e+00 3.92121047e-01 6.00094259e-01 -5.72156668e-01 1.02707967e-01 4.21915919e-01 4.18498665e-01 4.31417853e-01 9.36807767e-02 5.60110629e-01 5.34630477e-01 -4.49582309e-01 -1.25476086e+00 -8.20377827e-01 -4.62591499e-01 1.38587058e+00 -3.04905593e-01 -5.02944469e-01 -2.91042954e-01 -2.67993659e-01 3.63028765e-01 1.52692989e-01 -1.07207036e+00 -3.60541157e-02 -4.32348877e-01 -1.62467003e+00 6.71333432e-01 3.63010436e-01 1.05791986e-02 -5.66127717e-01 -2.18223333e-01 6.16672859e-02 -1.39769642e-02 -7.75034845e-01 -3.82237822e-01 4.23705310e-01 -8.45539749e-01 -1.48514056e+00 -3.47187221e-01 -8.04418206e-01 5.81137419e-01 2.71358311e-01 1.18234730e+00 6.80809096e-02 -8.75913918e-01 4.08100575e-01 -3.38350892e-01 -8.16792250e-01 -5.82449198e-01 -1.00396434e-02 1.35295525e-01 -1.97722197e-01 4.47173923e-01 -3.62765372e-01 -9.55173910e-01 -3.24578099e-02 -1.16948009e+00 3.04950148e-01 4.50390339e-01 7.76003122e-01 9.75337565e-01 -5.02551496e-01 4.77806628e-01 -1.01692140e+00 4.03893173e-01 -1.13129389e+00 -2.91526347e-01 1.03496406e-02 -4.75592017e-01 1.08614892e-01 4.42940474e-01 -4.26311105e-01 -5.53361595e-01 1.89635932e-01 -1.76264346e-01 4.30960394e-02 -6.59984946e-02 8.03542733e-01 8.15753490e-02 9.67245847e-02 8.75121534e-01 -2.80016288e-02 5.87252617e-01 -7.14211166e-01 4.90188688e-01 5.63334048e-01 4.53601927e-01 -7.42578685e-01 3.96038145e-01 6.93441451e-01 2.00713053e-01 -9.78438914e-01 -6.23821914e-01 -7.85827756e-01 -5.68285763e-01 3.01337987e-01 7.51112878e-01 -1.08976316e+00 -7.35284805e-01 6.58988416e-01 -7.94074237e-01 -4.95037019e-01 -1.13657564e-01 3.86076003e-01 -2.11222291e-01 5.12895703e-01 -8.04588199e-01 -5.36853731e-01 -5.63645482e-01 -9.73201096e-01 1.15921330e+00 -3.85864794e-01 -3.96901846e-01 -1.22890079e+00 5.99723637e-01 3.16276640e-01 3.38289112e-01 7.27658331e-01 1.28031552e+00 -7.59032726e-01 -3.50743532e-01 -2.72087636e-03 -1.57324225e-01 2.13737041e-01 5.06322145e-01 1.85777873e-01 -9.41122115e-01 -2.03703061e-01 -4.47415054e-01 -3.30757231e-01 1.25780034e+00 3.15223366e-01 1.29721081e+00 -2.33527541e-01 -4.46754485e-01 7.71092951e-01 1.29736483e+00 -3.14344555e-01 3.10929298e-01 3.22744161e-01 8.26867580e-01 6.42296970e-01 3.39115202e-01 4.96004224e-01 5.78546822e-01 3.32107127e-01 7.61909425e-01 -4.11988348e-01 -1.49339259e-01 -1.27520978e-01 -1.83359440e-02 7.20750690e-01 1.27495214e-01 -1.74822927e-01 -1.28776419e+00 6.95733249e-01 -1.74856722e+00 -4.89285827e-01 -3.67941588e-01 2.14760256e+00 1.20737743e+00 -3.98551494e-01 2.43321672e-01 -6.12637326e-02 3.84187579e-01 5.91759011e-02 -7.28659153e-01 -1.14336073e-01 -2.24748373e-01 3.37163180e-01 3.33297253e-01 3.94819617e-01 -1.39365292e+00 4.78743434e-01 6.23895502e+00 1.81878731e-01 -1.01646471e+00 -9.29364488e-02 6.83976352e-01 -3.41612697e-01 -5.20542204e-01 -3.89919937e-01 -3.94315928e-01 2.69906759e-01 6.14735663e-01 2.99420774e-01 5.49944401e-01 4.36558306e-01 1.59750387e-01 -7.06612179e-03 -1.51049566e+00 6.81494296e-01 -4.01571356e-02 -1.50067222e+00 -1.65760517e-02 3.43682855e-01 8.50486994e-01 5.76989770e-01 2.59439290e-01 1.04732461e-01 5.22219360e-01 -1.30160427e+00 2.26819351e-01 2.57192761e-01 7.80585110e-01 -4.42692339e-01 6.30550086e-01 -4.30985764e-02 -1.02804339e+00 2.76730001e-01 -6.00854084e-02 -2.20929042e-01 -2.59286612e-01 9.42338824e-01 -1.23810959e+00 4.08617616e-01 4.92232323e-01 1.00638938e+00 -4.33137149e-01 1.07392025e+00 1.03795886e-01 4.84760612e-01 -5.35575926e-01 5.28207496e-02 5.17901406e-03 -2.28353545e-01 1.93161845e-01 1.60422540e+00 2.52103284e-02 -6.39646202e-02 2.13079557e-01 6.28134549e-01 -2.59635985e-01 4.02189195e-01 -7.37591565e-01 -4.86627132e-01 5.43650270e-01 1.11454558e+00 -5.02712667e-01 -2.42215484e-01 -4.79111969e-01 4.22222972e-01 1.74765259e-01 8.36158246e-02 -4.63016093e-01 1.89572990e-01 1.24155414e+00 1.56286564e-02 -4.69189174e-02 -1.13654092e-01 -6.38918102e-01 -1.02931070e+00 -2.05327585e-01 -1.25017917e+00 4.46853906e-01 -1.06212258e-01 -1.69170916e+00 -2.90606543e-02 -6.00210838e-02 -1.01916349e+00 -1.22068357e-03 -9.48891342e-01 -4.56912071e-01 7.86000311e-01 -1.61530733e+00 -1.34821045e+00 -4.22543734e-01 1.97507665e-01 2.65367717e-01 3.23904492e-02 1.23807430e+00 2.88184881e-01 -6.84696436e-01 1.71086118e-01 3.56564254e-01 -3.51823750e-03 8.43992770e-01 -1.26931894e+00 8.36721480e-01 5.90240836e-01 -1.73972756e-01 1.03817403e+00 6.19596064e-01 -8.00498605e-01 -2.00442934e+00 -1.37338090e+00 3.69291335e-01 -5.06245553e-01 9.48719323e-01 -5.45202076e-01 -8.81366849e-01 6.31267309e-01 1.14631668e-01 3.67526338e-02 1.28822219e+00 5.34562588e-01 -8.75566423e-01 2.44887531e-01 -1.00087547e+00 5.66441357e-01 1.16641355e+00 -4.38073665e-01 -1.24963284e-01 7.22494423e-01 8.72594476e-01 -2.86494285e-01 -1.16567087e+00 7.46370912e-01 8.09408069e-01 -3.45652461e-01 1.31789792e+00 -1.03861988e+00 8.97653818e-01 -3.66906047e-01 -4.13360596e-01 -1.36627853e+00 -1.99146628e-01 -2.05817804e-01 -3.33121210e-01 8.27086210e-01 3.58222961e-01 -6.88014030e-01 5.04272103e-01 1.44230640e-02 -3.85769725e-01 -1.07839978e+00 -4.42474186e-01 -4.05197710e-01 4.68816131e-01 -2.60623485e-01 6.15331948e-01 1.43416584e+00 2.26727977e-01 2.08193347e-01 -2.30127767e-01 1.37821048e-01 7.66116798e-01 1.53131383e-02 8.02660823e-01 -1.36176300e+00 -2.79462934e-01 -5.66683471e-01 -1.55422360e-01 -6.85445011e-01 -2.42722511e-01 -1.30388105e+00 -1.11447098e-02 -1.80648708e+00 3.73471379e-01 -6.41266525e-01 -4.30729836e-01 6.15864992e-01 -3.54217529e-01 4.20424104e-01 -2.44509652e-01 1.07004210e-01 -2.41599306e-01 3.58127624e-01 1.11719072e+00 -3.58429700e-01 -1.70151606e-01 -3.94738764e-01 -9.38852072e-01 5.70747733e-01 1.22470975e+00 -6.01844303e-02 -2.84903914e-01 -6.20257318e-01 4.57540244e-01 -6.50973022e-01 3.92298639e-01 -4.03961927e-01 -1.44202009e-01 -3.84190053e-01 3.22049528e-01 -4.18103665e-01 2.21034274e-01 -3.89235407e-01 1.34436622e-01 7.56641448e-01 -4.01678026e-01 -4.07152511e-02 4.25257355e-01 6.90150559e-01 3.62550527e-01 2.58400291e-01 6.52661324e-01 -2.87404209e-01 -4.09601092e-01 4.32047129e-01 -3.50519210e-01 7.89502561e-02 6.41936719e-01 1.41997278e-01 -8.44896734e-01 3.10612440e-01 -2.60238707e-01 3.61060649e-01 7.54180908e-01 5.59513092e-01 3.67744952e-01 -1.13347042e+00 -1.08695960e+00 1.92136094e-01 5.12997031e-01 4.21194583e-01 -8.64756927e-02 1.05545747e+00 -7.09713519e-01 3.32848907e-01 -1.64406434e-01 -8.49859238e-01 -1.14354682e+00 6.78931057e-01 1.29687965e-01 -1.13143660e-01 -4.79699850e-01 9.59805250e-01 4.16350096e-01 -9.04555440e-01 3.16136211e-01 -7.43595600e-01 2.07916815e-02 2.09110513e-01 6.05067253e-01 2.76290655e-01 1.05491899e-01 -6.08370602e-02 -4.74714369e-01 1.69563681e-01 -1.58702701e-01 3.05276215e-01 1.63281965e+00 2.95125246e-01 -6.39874458e-01 6.01853013e-01 1.23002112e+00 -8.33826363e-02 -8.65713179e-01 -5.16117692e-01 -1.00736678e-01 -2.68967003e-01 -1.63223848e-01 -7.92644799e-01 -4.90248322e-01 8.53499711e-01 4.81409311e-01 6.04099631e-02 8.28347862e-01 -1.09353237e-01 4.66799915e-01 2.80706108e-01 6.48788586e-02 -5.82942963e-01 1.58344120e-01 4.83491212e-01 8.93093944e-01 -1.47731984e+00 3.37137610e-01 -5.50635159e-01 -2.09314242e-01 6.54821575e-01 3.70372385e-01 2.03482255e-01 5.03126264e-01 5.22213399e-01 -8.27258676e-02 -2.14949325e-01 -9.79145110e-01 -1.64362520e-01 1.56343505e-01 6.37744546e-01 1.12526786e+00 3.74069363e-01 -8.03745538e-02 1.51607364e-01 -2.23358780e-01 -2.16133282e-01 3.17019045e-01 1.01202786e+00 -4.35767859e-01 -1.17117977e+00 -4.75608319e-01 8.39707732e-01 -4.60583687e-01 -6.94292784e-01 -5.88850260e-01 5.57195544e-01 1.27628803e-01 7.51021445e-01 1.69644535e-01 -1.96631238e-01 -6.59727901e-02 1.87963724e-01 6.20396674e-01 -9.11220729e-01 -5.94223797e-01 8.49148184e-02 2.97601104e-01 -2.83766121e-01 -6.21765256e-01 -9.57325757e-01 -1.21004629e+00 -4.42168862e-01 1.35246575e-01 -2.17748746e-01 9.51770365e-01 6.68475151e-01 3.92069876e-01 7.61794090e-01 2.24352285e-01 -7.04499722e-01 -3.08307290e-01 -1.20516002e+00 -3.57172010e-03 3.88740987e-01 6.56482279e-01 -4.98447478e-01 -2.99253352e-02 2.41936579e-01]
[5.696694374084473, 5.682091236114502]
27a4a8c6-0597-410b-9d61-747ea0fb58a1
cyclegan-without-checkerboard-artifacts-for
2012.00287
null
https://arxiv.org/abs/2012.00287v1
https://arxiv.org/pdf/2012.00287v1.pdf
CycleGAN without checkerboard artifacts for counter-forensics of fake-image detection
In this paper, we propose a novel CycleGAN without checkerboard artifacts for counter-forensics of fake-image detection. Recent rapid advances in image manipulation tools and deep image synthesis techniques, such as Generative Adversarial Networks (GANs) have easily generated fake images, so detecting manipulated images has become an urgent issue. Most state-of-the-art forgery detection methods assume that images include checkerboard artifacts which are generated by using DNNs. Accordingly, we propose a novel CycleGAN without any checkerboard artifacts for counter-forensics of fake-mage detection methods for the first time, as an example of GANs without checkerboard artifacts.
['Hitoshi Kiya', 'Yuma Kinoshita', 'Miki Tanaka', 'Takayuki Osakabe']
2020-12-01
null
null
null
null
['fake-image-detection']
['computer-vision']
[ 5.17931223e-01 -1.53134972e-01 2.75127113e-01 3.18777919e-01 -2.81415015e-01 -7.11393952e-01 6.10741556e-01 -1.43792063e-01 -4.42587808e-02 8.79629433e-01 -4.09010559e-01 -5.78025579e-01 5.06874204e-01 -1.02221930e+00 -9.41626132e-01 -7.18987346e-01 2.83868700e-01 -2.22375557e-01 2.65579313e-01 -4.05493855e-01 4.19665009e-01 2.70668715e-01 -1.09413850e+00 4.15079892e-01 8.58168721e-01 6.18666351e-01 -2.20368072e-01 7.57459521e-01 1.47316530e-01 1.19092393e+00 -1.23834622e+00 -8.00052285e-01 5.09705007e-01 -1.02847540e+00 -2.53769487e-01 1.83105797e-01 3.96939546e-01 -8.55175138e-01 -1.03365171e+00 1.85366762e+00 3.01909328e-01 -6.65584266e-01 3.87626320e-01 -1.80960691e+00 -1.17159677e+00 3.77716482e-01 -7.22111583e-01 1.37971684e-01 -6.54216930e-02 9.05888736e-01 -1.40859736e-02 -4.82121706e-01 5.69054961e-01 1.25058365e+00 5.39633393e-01 9.42885041e-01 -8.27340305e-01 -1.13661003e+00 -6.40570521e-01 4.63489026e-01 -1.30141282e+00 -4.33217168e-01 9.47270215e-01 -1.56588688e-01 1.74118415e-01 3.92757088e-01 7.44330227e-01 1.37578750e+00 4.45016116e-01 5.91904640e-01 1.37446535e+00 -3.52762431e-01 -6.43361434e-02 2.33870298e-01 -5.05065501e-01 8.24963570e-01 1.02038598e+00 5.30213952e-01 -2.65221655e-01 -9.94731337e-02 1.23653269e+00 9.70303342e-02 -3.57284725e-01 2.49881104e-01 -8.88958097e-01 8.33392859e-01 3.61549586e-01 2.70861775e-01 -7.16178045e-02 7.40852296e-01 4.75548565e-01 4.32205856e-01 2.28381827e-01 3.68236333e-01 3.42355371e-01 1.44261628e-01 -8.30145061e-01 1.27101481e-01 4.22134101e-01 9.55438554e-01 5.27930617e-01 5.89806139e-01 -1.41566992e-01 8.12805444e-02 6.80368915e-02 7.00040042e-01 6.73350871e-01 -6.54702842e-01 4.22891378e-01 4.38078195e-01 1.74293041e-01 -1.68295622e+00 3.13572973e-01 -2.22529154e-02 -1.03428590e+00 4.59047645e-01 3.94866824e-01 -7.65491873e-02 -9.42998171e-01 1.26332498e+00 1.72816813e-01 4.25099403e-01 1.19270898e-01 8.70154500e-01 5.39594352e-01 5.36388874e-01 -3.82191122e-01 -1.53338417e-01 1.42605567e+00 -9.29523349e-01 -1.02507269e+00 -2.11561546e-01 3.90604407e-01 -6.99203908e-01 8.84203136e-01 5.42595088e-01 -8.04085374e-01 -4.30526525e-01 -1.47901225e+00 1.65397391e-01 -3.40083003e-01 -1.28578305e-01 5.27837932e-01 1.48473871e+00 -7.67637134e-01 3.98914039e-01 -3.58453453e-01 3.48484248e-01 8.41534495e-01 -3.06551829e-02 -5.30572653e-01 -1.26640111e-01 -1.30784976e+00 7.29836643e-01 4.66918319e-01 2.27539375e-01 -1.24013412e+00 -2.44786114e-01 -5.77010989e-01 -8.67426619e-02 3.79320532e-01 -3.99027258e-01 7.93909132e-01 -1.54527152e+00 -1.54133797e+00 1.03328931e+00 6.02466047e-01 -6.13427937e-01 1.09644997e+00 1.44222274e-01 -9.24297988e-01 4.60488826e-01 -1.23295888e-01 1.77402973e-01 1.55315626e+00 -1.42415273e+00 -6.95848241e-02 -1.41693115e-01 -1.74054027e-01 -5.96237302e-01 -4.07732666e-01 2.50167340e-01 9.17096287e-02 -8.48195076e-01 -3.25308591e-01 -7.54458666e-01 -6.77138939e-02 9.22171995e-02 -7.09106207e-01 3.61611634e-01 1.47288430e+00 -8.64971161e-01 1.21256876e+00 -2.13642001e+00 -6.11770093e-01 2.58686766e-03 4.52505916e-01 9.01641309e-01 -1.97835103e-01 3.66743535e-01 -6.25508204e-02 4.13625598e-01 -3.82540822e-01 -9.42486972e-02 -2.57835627e-01 -4.88537885e-02 -6.70766771e-01 9.38840568e-01 3.88285100e-01 1.26141179e+00 -1.25723827e+00 -4.46413577e-01 2.49156088e-01 3.78255576e-01 -1.39580965e-01 1.83813140e-01 -2.43373454e-01 6.80432975e-01 -1.26900390e-01 6.41329706e-01 1.25914860e+00 -5.17155323e-03 2.10415632e-01 -3.72269638e-02 1.48558229e-01 -1.11724615e-01 -6.88435435e-01 8.44140947e-01 2.65187714e-02 9.56254482e-01 -1.77890420e-01 -6.47232175e-01 9.87344205e-01 1.82634786e-01 -2.73183614e-01 -7.71749914e-01 4.62930173e-01 4.28222448e-01 -2.21265201e-02 -6.93360329e-01 6.21871412e-01 -1.92644909e-01 -3.95938754e-02 4.84970987e-01 -2.30966046e-01 -1.04278944e-01 -1.49207622e-01 1.68166861e-01 1.27512062e+00 -1.16732240e-01 8.25184882e-02 1.10914484e-01 4.74029213e-01 -2.44786479e-02 3.32948446e-01 8.55262339e-01 -3.18960726e-01 7.53784537e-01 5.58156133e-01 -4.85996544e-01 -1.39245546e+00 -4.48442101e-01 4.18215454e-01 -1.43668681e-01 5.58115304e-01 -2.15719298e-01 -1.09675169e+00 -8.71556163e-01 -1.03581704e-01 3.13981801e-01 -4.79790241e-01 -4.65997279e-01 -7.50318766e-01 -5.64162552e-01 1.47502887e+00 1.81462198e-01 1.18876970e+00 -1.18811035e+00 -5.64490855e-01 1.66983724e-01 -3.49511951e-02 -1.40057027e+00 -4.39287454e-01 -6.58668935e-01 -5.98662019e-01 -1.46692669e+00 -6.24373078e-01 -6.58946514e-01 1.02655506e+00 5.56613326e-01 7.56383657e-01 7.60325849e-01 -4.36022818e-01 -3.05519551e-01 -3.49173784e-01 -4.07516211e-01 -1.18025255e+00 -7.56138206e-01 -2.58941144e-01 3.32680821e-01 -6.43530637e-02 -3.50845903e-01 -8.80979002e-01 2.64808774e-01 -1.49175596e+00 8.51862282e-02 4.50124204e-01 8.63148928e-01 1.54890850e-01 3.42211783e-01 4.95287806e-01 -1.19045258e+00 6.31021082e-01 -3.04140061e-01 -9.65960324e-01 1.82086542e-01 -4.18201685e-01 -3.07058573e-01 1.00770533e+00 -6.43764019e-01 -7.99464405e-01 -5.10037124e-01 3.06994617e-02 -8.47540379e-01 -4.54351902e-02 -1.23346169e-02 -3.33203614e-01 -6.80048645e-01 6.66548669e-01 6.21575296e-01 8.87090415e-02 -1.02378808e-01 3.33705425e-01 6.83648765e-01 8.43191147e-01 1.29925162e-02 1.23413312e+00 8.34089637e-01 2.61279166e-01 -6.48600876e-01 -2.65847325e-01 3.00503016e-01 1.61954165e-01 -2.81426698e-01 5.59848249e-01 -6.04207814e-01 -9.45880771e-01 1.40903091e+00 -1.54806232e+00 -1.96289107e-01 1.44481048e-01 -6.67154882e-03 -2.07794100e-01 1.01691198e+00 -6.13570690e-01 -1.02118957e+00 -4.43893492e-01 -1.16013825e+00 7.31799603e-01 9.96568203e-02 4.34123248e-01 -5.62564433e-01 -2.12391704e-01 3.42610031e-01 5.59036434e-01 1.01587772e+00 5.18238723e-01 -1.43067136e-01 -1.00232184e+00 -6.13285661e-01 -3.95736903e-01 6.52762413e-01 1.70299962e-01 -7.46878609e-02 -7.73831010e-01 -4.24875587e-01 4.69070882e-01 -4.06683832e-02 7.25051820e-01 -2.99149871e-01 1.55033636e+00 -8.64747345e-01 -1.61743268e-01 7.90848792e-01 1.55603576e+00 3.87314618e-01 1.62162316e+00 4.07840967e-01 9.46791589e-01 1.30990103e-01 2.88572997e-01 2.52712220e-01 -1.42351702e-01 3.94615620e-01 9.43299472e-01 -5.25229797e-02 -2.85225123e-01 -5.57801187e-01 4.87599790e-01 5.05506575e-01 1.68718889e-01 -7.08296657e-01 -3.33358705e-01 3.82207125e-01 -1.56754434e+00 -1.27648520e+00 -6.59510374e-01 2.08182812e+00 6.19790912e-01 1.15087107e-01 -1.12040192e-01 3.41326118e-01 1.22241461e+00 3.48684877e-01 -4.21483576e-01 -2.45888621e-01 -4.47689265e-01 2.83796489e-01 1.07283366e+00 4.66906950e-02 -9.63373005e-01 1.10696030e+00 5.69862318e+00 1.13100708e+00 -1.18261361e+00 5.16991019e-01 6.88874424e-01 4.65821564e-01 -3.76719803e-01 2.14520127e-01 -2.85171211e-01 1.23158634e+00 5.27995646e-01 2.39317507e-01 4.38460737e-01 5.14255047e-01 1.19158752e-01 -3.21629159e-02 -4.15286720e-01 1.24914265e+00 4.32708144e-01 -1.63394606e+00 1.75532863e-01 2.82528132e-01 8.48456800e-01 -7.17500925e-01 2.15939030e-01 -2.16499329e-01 1.39235646e-01 -9.73854542e-01 7.80741334e-01 2.91560348e-02 1.16837823e+00 -8.20176959e-01 6.77049279e-01 1.61125764e-01 -5.87690532e-01 1.69482872e-01 -5.00439644e-01 2.19981939e-01 2.19876423e-01 8.58140826e-01 -6.13757014e-01 4.79931563e-01 2.59005785e-01 3.23906869e-01 -5.04856229e-01 8.73089433e-01 -7.71596730e-01 8.42250645e-01 8.11566114e-02 3.68930213e-02 3.26456934e-01 -3.05734932e-01 6.89558208e-01 9.10736084e-01 5.18929303e-01 -5.89548089e-02 -5.40550172e-01 1.26734459e+00 -5.09342194e-01 -4.44221526e-01 -9.95573163e-01 -3.61882389e-01 4.52361852e-01 9.51090991e-01 -1.02115917e+00 -5.13350129e-01 7.75868446e-02 1.44784486e+00 -3.03359389e-01 -6.82337657e-02 -1.10495508e+00 -7.36290693e-01 3.45366925e-01 1.46415815e-01 3.09867442e-01 -2.66630292e-01 -1.40425712e-01 -1.36815977e+00 -4.43112757e-03 -1.32523048e+00 -1.00261709e-02 -7.92463183e-01 -1.07947624e+00 4.94605124e-01 -5.70995271e-01 -1.43599153e+00 3.91617179e-01 -4.85516846e-01 -9.14046705e-01 3.99130881e-01 -1.58371019e+00 -1.18874276e+00 -5.08674979e-01 5.58614016e-01 2.92725265e-01 -3.06054771e-01 4.91396159e-01 2.24690184e-01 -4.59105313e-01 6.88776493e-01 1.51216373e-01 5.37543535e-01 4.58755702e-01 -5.83863497e-01 9.55456495e-01 1.41832280e+00 -3.08856890e-02 4.19047058e-01 6.18601561e-01 -1.05257654e+00 -1.71387172e+00 -1.04586387e+00 3.41622680e-01 -2.12213740e-01 5.21854579e-01 -3.77501786e-01 -8.68278086e-01 5.58011234e-01 4.59909827e-01 3.05897117e-01 1.70307830e-01 -1.28443205e+00 -6.36627972e-01 1.01527952e-01 -1.72093964e+00 4.73826438e-01 8.34162235e-01 -7.00376272e-01 -4.17864742e-03 4.44168687e-01 7.75468290e-01 -2.79515147e-01 -9.77648720e-02 1.15383945e-01 5.64296842e-01 -1.19678664e+00 8.49967539e-01 -2.83928186e-01 5.64325273e-01 -6.81738019e-01 3.37749004e-01 -1.05824184e+00 1.62339941e-01 -1.30767167e+00 -3.65568340e-01 1.02510166e+00 -4.75757390e-01 -8.10385406e-01 8.02030742e-01 -1.79864928e-01 1.33063257e-01 4.36320975e-02 -7.68885195e-01 -1.10353661e+00 -2.78848171e-01 -6.30987957e-02 8.36793005e-01 1.22767735e+00 -2.73091346e-01 -3.77703130e-01 -1.08398783e+00 2.01022878e-01 9.51520503e-01 -2.77930796e-01 9.61138964e-01 -4.95238155e-01 -3.21641207e-01 -2.06557006e-01 -8.60946894e-01 -7.37693608e-01 -1.08026840e-01 -3.80933374e-01 -1.57254890e-01 -6.91182852e-01 6.30090833e-02 -1.00430809e-01 9.68198255e-02 1.60232574e-01 -2.77267009e-01 9.80375886e-01 2.59055018e-01 2.10584983e-01 -2.00626582e-01 3.44504178e-01 1.43474698e+00 -3.31000447e-01 4.56236333e-01 -3.82941246e-01 -6.15055382e-01 3.99335653e-01 8.39654028e-01 -1.02262151e+00 -1.72397286e-01 -4.07695621e-01 3.42535377e-01 8.98564793e-03 7.85503149e-01 -1.01877010e+00 3.48545983e-02 -1.65317625e-01 1.83150694e-01 -3.05186152e-01 3.68606038e-02 -5.71509063e-01 4.95077252e-01 1.03689933e+00 2.15421796e-01 1.99601725e-01 2.10379079e-01 7.38340616e-01 -3.13452870e-01 -3.05940598e-01 8.33681583e-01 -5.95280766e-01 -4.45972264e-01 1.59168571e-01 -5.95982254e-01 -1.83234930e-01 1.10542345e+00 -4.96094257e-01 -9.95606601e-01 -4.58442479e-01 5.00940122e-02 -3.29708457e-01 7.78929353e-01 1.86048105e-01 1.05705392e+00 -1.36055541e+00 -7.05281496e-01 3.61711055e-01 5.40689984e-03 -2.71571338e-01 3.76972169e-01 4.70924169e-01 -1.22319365e+00 6.15049293e-03 -5.50306976e-01 -8.91985893e-02 -1.36222875e+00 8.96115780e-01 2.16451675e-01 -2.09547147e-01 -4.27956462e-01 5.74171007e-01 1.43975362e-01 2.21065596e-01 -4.47870225e-01 1.99322283e-01 3.34787309e-01 -4.28277344e-01 7.21304655e-01 5.65006256e-01 8.97054654e-03 -6.27606571e-01 -1.44970447e-01 1.31112158e-01 2.07094461e-01 1.11543588e-01 9.50110674e-01 2.48392578e-02 -5.12114167e-01 -2.70492315e-01 1.10054958e+00 4.54346798e-02 -1.11279035e+00 3.17015648e-01 -3.23270619e-01 -1.17117774e+00 1.97175983e-02 -5.46761394e-01 -1.37665415e+00 8.01277459e-01 4.67348754e-01 4.06692177e-01 9.57503736e-01 -5.65073133e-01 1.37579799e+00 -5.73937595e-02 6.42284334e-01 -6.13229752e-01 5.24993420e-01 2.20124163e-02 6.40751362e-01 -1.03847492e+00 -8.80002305e-02 -3.98420841e-01 -3.91562253e-01 1.25808978e+00 5.56505024e-01 -4.97112304e-01 9.31239277e-02 1.11108728e-01 4.66816425e-02 -1.94685549e-01 -1.84234366e-01 3.36372197e-01 -3.62547874e-01 7.12626994e-01 -3.46049398e-01 1.63847849e-01 -4.78327513e-01 1.58998296e-01 9.18796733e-02 2.07166765e-02 1.17306602e+00 1.02261043e+00 -1.43304035e-01 -1.16874087e+00 -8.23460102e-01 7.55953938e-02 -7.07886636e-01 -2.12165490e-01 -5.85938990e-01 6.72730625e-01 2.86754221e-01 9.26414132e-01 -6.84390739e-02 -6.30977988e-01 -1.01559483e-01 -4.71814454e-01 8.32022667e-01 -8.25416520e-02 -7.47269154e-01 -3.53066981e-01 -2.64601707e-01 -3.49110752e-01 -3.10465395e-01 -2.10036293e-01 -7.54965782e-01 -7.44720101e-01 -8.78944814e-01 -1.79989383e-01 8.11969399e-01 6.07640624e-01 4.78693157e-01 2.18667597e-01 9.57195520e-01 -6.15021229e-01 -4.85314906e-01 -7.48477459e-01 -5.95826983e-01 6.35777295e-01 4.18572932e-01 -2.95474261e-01 -5.83760202e-01 1.13260977e-01]
[12.528511047363281, 1.1292022466659546]
2fd5959f-8f81-42fd-bc59-c2393064dd20
poseaug-a-differentiable-pose-augmentation
2105.02465
null
https://arxiv.org/abs/2105.02465v1
https://arxiv.org/pdf/2105.02465v1.pdf
PoseAug: A Differentiable Pose Augmentation Framework for 3D Human Pose Estimation
Existing 3D human pose estimators suffer poor generalization performance to new datasets, largely due to the limited diversity of 2D-3D pose pairs in the training data. To address this problem, we present PoseAug, a new auto-augmentation framework that learns to augment the available training poses towards a greater diversity and thus improve generalization of the trained 2D-to-3D pose estimator. Specifically, PoseAug introduces a novel pose augmentor that learns to adjust various geometry factors (e.g., posture, body size, view point and position) of a pose through differentiable operations. With such differentiable capacity, the augmentor can be jointly optimized with the 3D pose estimator and take the estimation error as feedback to generate more diverse and harder poses in an online manner. Moreover, PoseAug introduces a novel part-aware Kinematic Chain Space for evaluating local joint-angle plausibility and develops a discriminative module accordingly to ensure the plausibility of the augmented poses. These elaborate designs enable PoseAug to generate more diverse yet plausible poses than existing offline augmentation methods, and thus yield better generalization of the pose estimator. PoseAug is generic and easy to be applied to various 3D pose estimators. Extensive experiments demonstrate that PoseAug brings clear improvements on both intra-scenario and cross-scenario datasets. Notably, it achieves 88.6% 3D PCK on MPI-INF-3DHP under cross-dataset evaluation setup, improving upon the previous best data augmentation based method by 9.1%. Code can be found at: https://github.com/jfzhang95/PoseAug.
['Jiashi Feng', 'Jianfeng Zhang', 'Kehong Gong']
2021-05-06
null
http://openaccess.thecvf.com//content/CVPR2021/html/Gong_PoseAug_A_Differentiable_Pose_Augmentation_Framework_for_3D_Human_Pose_CVPR_2021_paper.html
http://openaccess.thecvf.com//content/CVPR2021/papers/Gong_PoseAug_A_Differentiable_Pose_Augmentation_Framework_for_3D_Human_Pose_CVPR_2021_paper.pdf
cvpr-2021-1
['monocular-3d-human-pose-estimation', 'weakly-supervised-3d-human-pose-estimation']
['computer-vision', 'computer-vision']
[-2.43836820e-01 2.70974517e-01 -2.44471520e-01 -3.95386875e-01 -8.32196951e-01 -5.46259582e-01 3.73048842e-01 -2.30429947e-01 -1.85538322e-01 6.77652776e-01 3.80135864e-01 7.74420649e-02 7.03945458e-02 -5.41733921e-01 -1.05216897e+00 -4.92078662e-01 -1.48849770e-01 8.87487173e-01 3.62120429e-03 -4.76710469e-01 -9.19098482e-02 6.10830605e-01 -1.25836611e+00 -3.12448204e-01 7.91677773e-01 7.59813547e-01 -5.75733418e-03 5.41896701e-01 5.26046276e-01 -2.31254905e-01 -5.15571177e-01 -3.20237696e-01 5.95332325e-01 -7.13934153e-02 -5.28873086e-01 2.17309460e-01 6.17690980e-01 -5.08086145e-01 -4.23093915e-01 6.34082198e-01 8.43018413e-01 3.03933740e-01 5.19527078e-01 -1.10590577e+00 -2.62340307e-01 2.19269127e-01 -5.81664085e-01 -8.05025771e-02 6.75381958e-01 2.40756512e-01 6.34007752e-01 -1.11813056e+00 7.06859469e-01 1.35083854e+00 8.80742490e-01 6.83493495e-01 -1.23754823e+00 -7.89647579e-01 2.66135633e-01 -1.65952519e-01 -1.36870098e+00 -9.95904431e-02 8.71372342e-01 -2.53156155e-01 6.95375800e-01 2.97387332e-01 8.89338076e-01 1.57065856e+00 2.58458167e-01 7.26793289e-01 8.66132975e-01 -1.34484947e-01 -4.76661511e-02 -3.21302295e-01 -3.07489753e-01 7.82267451e-01 3.23828489e-01 1.52880609e-01 -6.37756526e-01 -1.09498791e-01 1.15913439e+00 7.26650143e-03 -3.10525507e-01 -1.05893016e+00 -1.42515624e+00 5.94995022e-01 7.31507838e-01 -5.17542124e-01 -2.82058835e-01 2.21280083e-01 4.54371452e-01 -1.34813219e-01 3.81733537e-01 6.48276925e-01 -9.31717575e-01 -3.17729980e-01 -2.50098109e-01 8.42707694e-01 5.75042367e-01 1.16823423e+00 5.65878868e-01 1.47273894e-02 -4.62535284e-02 6.87155306e-01 3.27895522e-01 7.69167602e-01 2.48255983e-01 -1.03099215e+00 9.07274246e-01 6.72001123e-01 -8.63551795e-02 -1.04902315e+00 -7.83919990e-01 -8.71223569e-01 -7.82597840e-01 -9.57884192e-02 2.64972210e-01 -2.52024710e-01 -9.25468564e-01 1.91608083e+00 9.38439190e-01 -3.27846780e-02 -2.84308434e-01 1.13729501e+00 5.72551370e-01 2.94202834e-01 -1.63178518e-01 1.20694831e-01 1.24260461e+00 -9.40498829e-01 -3.26100677e-01 -4.79879886e-01 6.29438877e-01 -6.54741943e-01 1.05111206e+00 3.62613797e-01 -1.08261049e+00 -7.81642318e-01 -1.12497330e+00 -4.77117449e-02 -4.99763153e-02 3.19850355e-01 7.19987929e-01 4.58901018e-01 -4.63217050e-01 4.63835448e-01 -1.02824068e+00 -1.67740077e-01 2.00096086e-01 5.37088871e-01 -5.38406014e-01 1.25659302e-01 -1.05739999e+00 9.51243937e-01 4.87518698e-01 3.09789270e-01 -7.11660028e-01 -9.06483412e-01 -1.14820147e+00 -4.04495746e-01 7.33353674e-01 -1.12962270e+00 1.19897628e+00 -2.18132436e-01 -1.56877160e+00 5.89460373e-01 2.47393638e-01 -1.27573162e-01 8.60217273e-01 -8.38695586e-01 -2.42013466e-02 -1.08094132e-02 1.97249457e-01 8.39667916e-01 7.51337230e-01 -1.32947278e+00 -1.40320629e-01 -6.36061668e-01 -4.52338047e-02 6.49256110e-01 -3.44693512e-02 -7.16817796e-01 -9.61297512e-01 -9.84978855e-01 4.68643993e-01 -1.32181823e+00 -3.05614620e-01 1.77675202e-01 -5.28858542e-01 1.06778555e-01 4.96028095e-01 -7.18755722e-01 9.06139612e-01 -1.87372029e+00 6.86164498e-01 2.81420976e-01 1.53076708e-01 1.24966681e-01 -7.62334988e-02 3.33221823e-01 -7.19723627e-02 -1.48166165e-01 -1.99287206e-01 -5.18389583e-01 3.29933614e-02 4.07660037e-01 4.32270505e-02 5.91606498e-01 3.80459756e-01 1.01865828e+00 -8.23252559e-01 -4.17155504e-01 3.75629395e-01 5.68688810e-01 -1.01331246e+00 2.44680658e-01 -1.23420730e-01 8.13250601e-01 -6.44966006e-01 7.23376870e-01 7.21796691e-01 -7.97257796e-02 1.44301608e-01 -4.13950086e-01 2.88978964e-01 2.73495436e-01 -1.28858995e+00 2.15665078e+00 -3.57300639e-01 6.33473322e-03 -1.91371396e-01 -5.39860547e-01 1.03074777e+00 -5.44373654e-02 5.70373833e-01 -3.06509525e-01 3.62046689e-01 1.60953939e-01 -1.40626878e-01 -2.27083609e-01 5.50173998e-01 1.21123210e-01 -5.38644493e-01 7.76669979e-02 1.65485784e-01 -2.74903327e-01 -2.01060444e-01 -1.94449313e-02 7.33607292e-01 7.71143317e-01 3.80622476e-01 -7.58217797e-02 4.46494401e-01 -2.58524060e-01 6.56692326e-01 2.78768092e-01 -6.36534020e-02 8.31283748e-01 1.24663636e-01 -3.76081914e-01 -1.22595179e+00 -1.50083888e+00 -2.00923875e-01 7.98746824e-01 2.38027707e-01 -4.53041911e-01 -6.31553829e-01 -7.93564379e-01 3.97641689e-01 2.91958570e-01 -6.24245882e-01 -4.21985000e-01 -9.24488783e-01 -5.34278572e-01 4.44697350e-01 8.78691673e-01 5.68611443e-01 -5.00466287e-01 -3.90094370e-01 -1.50952842e-02 -2.44331554e-01 -1.05471969e+00 -6.52675033e-01 8.73687416e-02 -1.08117115e+00 -1.09097528e+00 -5.37833810e-01 -6.04387462e-01 8.07244360e-01 -1.06632777e-01 8.99599731e-01 -1.02582544e-01 -1.62600771e-01 2.91073561e-01 -5.26716232e-01 -1.68561742e-01 -1.82242051e-01 4.65415686e-01 5.36234140e-01 -4.36643660e-01 -1.67248666e-01 -5.77816784e-01 -7.25268483e-01 6.77335978e-01 -4.30208832e-01 1.95170268e-01 5.79439819e-01 9.95256782e-01 8.63905549e-01 -4.85294908e-01 3.88113022e-01 -5.93826234e-01 1.71162918e-01 -3.44495833e-01 -3.17176104e-01 -2.23738045e-01 -3.61955166e-01 1.16455965e-01 4.07140762e-01 -5.71353376e-01 -9.72028732e-01 3.73276144e-01 -3.16548586e-01 -4.75267202e-01 -3.01497597e-02 3.29831302e-01 -4.93509084e-01 -8.33055153e-02 7.49850035e-01 3.92471999e-02 3.43937069e-01 -7.39472508e-01 4.56821620e-01 1.09768949e-01 8.62060010e-01 -1.00772643e+00 1.23520529e+00 1.72424570e-01 2.45616600e-01 -3.79045457e-01 -7.65955329e-01 -4.05781776e-01 -9.37321961e-01 -2.80635148e-01 5.37482858e-01 -1.12109315e+00 -6.80888534e-01 4.35404420e-01 -9.05178308e-01 -2.91486025e-01 -1.63664818e-01 6.01880491e-01 -7.84928322e-01 4.24180925e-01 -3.95114243e-01 -4.83022630e-01 -3.53070617e-01 -1.13239491e+00 1.54836118e+00 -9.57892835e-03 -4.88529533e-01 -6.62079811e-01 -7.86171388e-03 4.84156370e-01 -1.47117570e-01 8.21917892e-01 4.79206353e-01 -3.75425220e-01 -3.16324025e-01 -5.12085915e-01 2.53102958e-01 2.58131981e-01 1.30955935e-01 -2.95279086e-01 -4.80776370e-01 -5.78210533e-01 -3.35699618e-01 -3.73563051e-01 3.10597181e-01 1.83554381e-01 1.22178543e+00 -2.17583790e-01 -4.06152785e-01 8.97356927e-01 9.70099449e-01 -3.72837752e-01 4.63568360e-01 4.37377453e-01 8.84181380e-01 3.92755568e-01 9.93974686e-01 5.87533653e-01 4.94341344e-01 1.11170018e+00 5.53040206e-01 7.06074536e-02 -2.00735509e-01 -6.28769696e-01 2.86463559e-01 9.06666338e-01 -3.35352629e-01 9.23551992e-03 -8.69904339e-01 6.91169277e-02 -1.80967736e+00 -3.73070031e-01 6.88126683e-02 2.27102017e+00 8.66204262e-01 2.77268261e-01 3.35483164e-01 3.10486313e-02 4.31018561e-01 3.88688594e-03 -7.66460061e-01 5.36775254e-02 2.52166122e-01 1.76863596e-01 5.09090185e-01 2.97491670e-01 -1.03606367e+00 8.68205130e-01 5.44100142e+00 6.22359872e-01 -8.47671151e-01 -9.17317793e-02 1.87935293e-01 -3.45843643e-01 -5.10913581e-02 -3.21716189e-01 -9.35943484e-01 3.25302213e-01 4.31181788e-01 1.78264499e-01 1.96883917e-01 9.68747735e-01 4.25155200e-02 1.22153908e-01 -1.13297832e+00 9.33735013e-01 2.67016366e-02 -1.03419602e+00 1.02852292e-01 2.39438564e-01 6.73012614e-01 -1.42131343e-01 6.26681969e-02 4.66276467e-01 -1.72442019e-01 -6.66491926e-01 8.46306205e-01 4.76588696e-01 7.78847277e-01 -9.63541687e-01 7.32079804e-01 3.84830683e-01 -1.28608215e+00 9.36789587e-02 -8.42977017e-02 -1.08421773e-01 3.21585864e-01 1.65612504e-01 -9.02099609e-01 8.29041064e-01 6.52928829e-01 5.88062108e-01 -7.08998203e-01 8.83515537e-01 -3.42213184e-01 1.69680387e-01 -4.48905677e-01 1.03155419e-01 -4.91410494e-02 9.45943296e-02 9.22844231e-01 6.07508361e-01 2.74252236e-01 -1.10844456e-01 4.37517464e-01 6.05641782e-01 2.32976284e-02 1.74111910e-02 -2.97710836e-01 3.99057090e-01 6.39572322e-01 1.08001423e+00 -3.01878184e-01 2.25386452e-02 1.05067857e-01 1.11521912e+00 3.52741599e-01 1.27861068e-01 -1.12420404e+00 -2.22893115e-02 8.83751035e-01 3.40555340e-01 1.58652112e-01 -5.86077631e-01 -2.66464323e-01 -1.13621497e+00 4.31449056e-01 -1.05940855e+00 2.53894240e-01 -7.13989317e-01 -1.08400810e+00 3.99781495e-01 5.35959959e-01 -1.38824654e+00 -4.62164223e-01 -6.13403976e-01 -1.62562400e-01 6.62836909e-01 -7.77277827e-01 -1.31005335e+00 -5.40182948e-01 2.91326791e-01 4.99676734e-01 -6.86750636e-02 7.22879648e-01 1.77816838e-01 -5.29278338e-01 8.92607331e-01 -3.61187369e-01 2.71365475e-02 9.14212465e-01 -1.22037458e+00 8.00156415e-01 5.89733958e-01 -1.41184822e-01 7.84971356e-01 7.19570220e-01 -8.64783406e-01 -1.66879201e+00 -1.14011812e+00 2.77627766e-01 -8.87416482e-01 3.46372128e-01 -5.93443990e-01 -6.33893967e-01 7.68539727e-01 -5.87150156e-01 7.57332221e-02 4.00056154e-01 2.21586719e-01 -3.96908194e-01 -7.64387175e-02 -1.03800845e+00 7.68751681e-01 1.56802928e+00 -1.49253458e-01 -6.22232437e-01 2.40263849e-01 9.31401372e-01 -1.29013419e+00 -1.23097134e+00 8.64492893e-01 7.56765127e-01 -5.23999512e-01 1.38312674e+00 -5.57700515e-01 2.88258642e-01 -4.64357078e-01 -1.61160752e-01 -1.26820660e+00 -2.68660307e-01 -6.29555643e-01 -5.06821036e-01 9.27339733e-01 2.40233615e-01 -6.56292915e-01 1.01747108e+00 2.35612869e-01 -4.57134485e-01 -1.25326622e+00 -9.83582675e-01 -9.08092260e-01 9.38862190e-02 -4.97473270e-01 8.62433732e-01 5.38739145e-01 -2.53588349e-01 1.97472543e-01 -5.55806875e-01 4.60464001e-01 6.33753955e-01 -1.56300351e-01 1.44816613e+00 -1.03381300e+00 -4.92498130e-01 -1.35605067e-01 -6.43182099e-01 -1.45754790e+00 -3.37014124e-02 -7.51173258e-01 1.06643558e-01 -1.12720573e+00 2.73613334e-02 -4.28863019e-01 2.34311298e-01 4.87171322e-01 -3.74240696e-01 3.33248049e-01 2.14007407e-01 5.69548681e-02 -3.45897734e-01 9.07921135e-01 1.68134618e+00 1.74370125e-01 -2.99654156e-01 7.14415237e-02 -4.72150832e-01 6.42266572e-01 8.25972378e-01 -1.22336820e-01 -3.61861169e-01 -2.94905096e-01 1.61955804e-01 -3.66201438e-02 6.17105544e-01 -1.16152561e+00 -2.15912566e-01 2.87003163e-02 8.24623585e-01 -8.16451192e-01 5.87723494e-01 -6.25927567e-01 3.40551883e-01 5.15900612e-01 -5.43410741e-02 1.94668785e-01 4.03865248e-01 7.49029279e-01 1.79923326e-01 2.70858467e-01 4.83751476e-01 7.79286847e-02 -5.59892535e-01 5.61557710e-01 2.36249328e-01 1.79788053e-01 9.80687261e-01 -4.27228391e-01 2.34024879e-02 -1.81993797e-01 -7.74263620e-01 4.09757555e-01 5.96537113e-01 7.56353378e-01 5.38445473e-01 -1.66739869e+00 -4.38911766e-01 3.30834389e-01 3.01560998e-01 6.74070597e-01 3.22218448e-01 8.80830169e-01 -5.71035743e-01 1.07756346e-01 -3.30382407e-01 -8.24091792e-01 -1.08364975e+00 2.70559251e-01 2.98039854e-01 -1.70705482e-01 -6.90380156e-01 8.77658129e-01 7.43467435e-02 -1.00313580e+00 1.64099962e-01 -4.25921500e-01 2.99325645e-01 -2.44166195e-01 1.24533162e-01 5.97490788e-01 2.69369986e-02 -7.53074050e-01 -5.60218930e-01 7.81371891e-01 7.15490207e-02 1.45106301e-01 1.28349638e+00 -1.19148463e-01 3.09984893e-01 1.29449353e-01 1.30067110e+00 5.65134399e-02 -1.64802814e+00 -3.05592548e-02 -5.78305900e-01 -5.55622637e-01 -3.38265032e-01 -7.49783933e-01 -9.14419651e-01 4.42232430e-01 4.56386298e-01 -5.92480183e-01 8.68560016e-01 1.35036260e-01 8.78575802e-01 2.83664256e-01 5.58333039e-01 -1.09815955e+00 4.67531741e-01 5.14841616e-01 1.37730014e+00 -1.12336636e+00 3.88726354e-01 -7.45917976e-01 -5.47707915e-01 1.06741881e+00 1.03504872e+00 -1.90973252e-01 3.98008943e-01 -1.42710973e-02 -9.60523263e-02 -2.13344321e-01 -2.63471574e-01 7.39096031e-02 8.65363955e-01 6.75997615e-01 2.98253685e-01 2.09369943e-01 -1.62309647e-01 6.04195237e-01 -7.12561607e-01 -4.15054590e-01 1.10205542e-02 9.31248486e-01 -7.29238391e-02 -1.20652163e+00 -5.57762027e-01 2.19165936e-01 1.29621625e-01 3.65333259e-01 -2.44020537e-01 1.20935225e+00 2.21562520e-01 3.27698708e-01 -1.41693249e-01 -7.35659957e-01 6.67704761e-01 -1.14531815e-01 8.28079104e-01 -7.36953020e-01 -4.32179958e-01 1.99321538e-01 2.07712591e-01 -8.82780671e-01 -6.90002069e-02 -8.07968795e-01 -1.33918691e+00 -1.87914163e-01 -4.37239140e-01 -2.46577233e-01 5.06799459e-01 7.85370648e-01 4.42933977e-01 5.98546743e-01 4.53107297e-01 -1.42330420e+00 -9.52114701e-01 -9.68034923e-01 -1.70390949e-01 4.66125876e-01 1.17643967e-01 -1.30248547e+00 -1.02687694e-01 -2.82401890e-01]
[6.977750778198242, -0.9864155650138855]
d3bf5fa4-d04a-4ea4-ae9a-672d9197fde5
informative-bayesian-model-selection-for-rr
2105.11531
null
https://arxiv.org/abs/2105.11531v1
https://arxiv.org/pdf/2105.11531v1.pdf
Informative Bayesian model selection for RR Lyrae star classifiers
Machine learning has achieved an important role in the automatic classification of variable stars, and several classifiers have been proposed over the last decade. These classifiers have achieved impressive performance in several astronomical catalogues. However, some scientific articles have also shown that the training data therein contain multiple sources of bias. Hence, the performance of those classifiers on objects not belonging to the training data is uncertain, potentially resulting in the selection of incorrect models. Besides, it gives rise to the deployment of misleading classifiers. An example of the latter is the creation of open-source labelled catalogues with biased predictions. In this paper, we develop a method based on an informative marginal likelihood to evaluate variable star classifiers. We collect deterministic rules that are based on physical descriptors of RR Lyrae stars, and then, to mitigate the biases, we introduce those rules into the marginal likelihood estimation. We perform experiments with a set of Bayesian Logistic Regressions, which are trained to classify RR Lyraes, and we found that our method outperforms traditional non-informative cross-validation strategies, even when penalized models are assessed. Our methodology provides a more rigorous alternative to assess machine learning models using astronomical knowledge. From this approach, applications to other classes of variable stars and algorithmic improvements can be developed.
['D. Mery', 'M. Catelan', 'P. Huijse', 'K. Pichara', 'F. Pérez-Galarce']
2021-05-24
null
null
null
null
['classification-of-variable-stars']
['miscellaneous']
[ 4.78828419e-03 -2.82777455e-02 -2.55205750e-01 -4.44285184e-01 -6.88962162e-01 -6.58643901e-01 9.62682605e-01 -3.08741387e-02 -2.79740751e-01 1.10309863e+00 -3.46359640e-01 -5.04914045e-01 -4.20867920e-01 -6.91190243e-01 -4.09489185e-01 -1.13725638e+00 2.37151742e-01 7.51853645e-01 5.40050566e-01 8.04822668e-02 4.70530897e-01 7.07149148e-01 -1.80740309e+00 -2.69496381e-01 7.70138204e-01 1.04911256e+00 1.15609176e-01 4.43134069e-01 -1.20680630e-01 7.82145083e-01 -8.02384198e-01 -4.67751771e-01 7.66483545e-02 -3.44071805e-01 -5.55697083e-01 8.54904950e-02 2.13657185e-01 2.96121866e-01 1.21034019e-01 9.85917032e-01 3.60167980e-01 -1.38011247e-01 1.14130998e+00 -1.23432934e+00 -2.96337843e-01 4.21974421e-01 -3.06368172e-01 -4.05430607e-02 -1.25926554e-01 2.05955729e-01 1.05473948e+00 -7.75378048e-01 5.65842800e-02 1.13502216e+00 8.14534664e-01 2.11523324e-01 -1.20694280e+00 -6.62961006e-01 -2.45084748e-01 3.23895156e-01 -1.33473778e+00 -1.47199944e-01 7.41054356e-01 -8.10678899e-01 4.80735689e-01 4.20637220e-01 6.98702276e-01 7.54412472e-01 1.13023356e-01 3.24942201e-01 1.60163045e+00 -6.63753211e-01 4.43696469e-01 5.60985148e-01 1.71376973e-01 3.76170844e-01 6.40044212e-01 5.27297556e-01 -4.36799824e-01 -5.31744123e-01 3.11463118e-01 -4.26571667e-01 -1.84139401e-01 -4.85623926e-01 -7.92455792e-01 1.11075258e+00 -1.88562355e-03 2.73299396e-01 -2.24029720e-01 -3.86798903e-02 8.80044401e-02 -3.24329175e-02 5.07403493e-01 7.25310981e-01 -6.61469400e-01 5.37649207e-02 -8.61888230e-01 3.40442896e-01 8.81622076e-01 4.81679201e-01 6.37224853e-01 -5.31902164e-02 1.05434284e-01 9.45939243e-01 6.27658248e-01 5.92230380e-01 6.05323672e-01 -5.26530683e-01 -4.04528826e-02 4.97807384e-01 2.32925892e-01 -6.45054162e-01 -2.68891454e-01 -7.41469145e-01 -4.38667625e-01 6.24427199e-01 5.82809746e-01 2.62145877e-01 -7.98608363e-01 1.37146890e+00 2.59553790e-01 -3.62531208e-02 9.49757099e-02 8.19244623e-01 7.58214593e-01 2.55107939e-01 1.28439635e-01 -3.54215026e-01 1.04606795e+00 -5.58863044e-01 -5.39935648e-01 -5.04961573e-02 4.27411586e-01 -6.84668303e-01 7.03601539e-01 8.90051007e-01 -6.70429528e-01 -4.18760478e-01 -9.92688894e-01 5.51256537e-01 -2.47220010e-01 5.09260893e-02 7.62869239e-01 8.61757278e-01 -6.09519720e-01 6.96203232e-01 -5.09809971e-01 -4.54210024e-03 9.83211845e-02 2.48475030e-01 -3.73172723e-02 4.91819561e-01 -1.00969219e+00 1.17285371e+00 4.40550089e-01 5.28836027e-02 -8.69545341e-01 -2.26896852e-01 -3.30737442e-01 -1.36757106e-01 3.50689948e-01 -2.09701866e-01 1.39155066e+00 -1.08829391e+00 -1.24440324e+00 8.07283103e-01 -1.31928157e-02 -4.06104743e-01 3.70831460e-01 1.42733112e-01 -4.52446908e-01 -1.26900852e-01 -1.54634401e-01 2.74227019e-02 9.83767152e-01 -1.37218142e+00 -7.18805552e-01 -3.47880810e-01 -2.52222240e-01 -2.64179438e-01 -1.02759860e-01 3.63188148e-01 -1.38682485e-01 -6.42655432e-01 4.07341152e-01 -1.03497648e+00 -2.37901360e-01 -3.81530374e-01 -2.59279218e-02 -7.56176293e-01 3.42155963e-01 -4.10546690e-01 1.16178048e+00 -1.95261800e+00 1.31828964e-01 3.40797633e-01 1.94878086e-01 2.88534909e-01 5.83607376e-01 -8.91195890e-03 -1.67977765e-01 1.28622204e-01 -6.65299296e-01 -3.52850906e-03 -1.78925440e-01 2.55519003e-01 -3.74188304e-01 5.07598877e-01 1.23291105e-01 2.39137843e-01 -6.71549976e-01 -5.53332150e-01 2.13956073e-01 3.15006465e-01 -3.26905549e-02 2.29179785e-01 -3.76854241e-01 3.63476992e-01 -4.56029743e-01 6.24529898e-01 5.83577454e-01 -4.05147634e-02 6.56090677e-02 1.36856452e-01 -2.90630877e-01 4.56549615e-01 -1.08135068e+00 8.60123336e-01 -2.39338294e-01 6.19601071e-01 -4.29559320e-01 -1.10504007e+00 1.38757420e+00 1.98762298e-01 2.19819397e-01 -4.58981991e-02 1.58946812e-01 5.77518463e-01 6.98217303e-02 -4.48807240e-01 4.06903446e-01 -4.52886015e-01 1.08532026e-01 7.51901865e-02 7.88878053e-02 -6.44137502e-01 -2.14379150e-02 -2.67968774e-01 7.46496499e-01 3.58289421e-01 4.72564757e-01 -1.24075361e-01 7.27240086e-01 1.40724525e-01 7.23496974e-01 7.97330379e-01 -5.64427227e-02 5.40027976e-01 3.97903860e-01 -4.45496172e-01 -8.77444506e-01 -9.61895287e-01 -8.34750533e-01 6.88445568e-01 -1.30084530e-01 -3.14964294e-01 -3.33012372e-01 -7.30115116e-01 1.72091529e-01 9.64305758e-01 -2.32475922e-01 -1.92535490e-01 -2.95604672e-03 -1.29700255e+00 5.36617637e-01 2.72242039e-01 4.16594781e-02 -9.50380325e-01 -6.49979234e-01 3.45238857e-02 9.06861424e-02 -5.32774508e-01 6.26894593e-01 6.25940681e-01 -9.55777168e-01 -1.33372009e+00 -4.06366795e-01 -3.31635296e-01 4.20267195e-01 1.14277333e-01 1.30717444e+00 2.53738821e-01 -6.72744066e-02 1.69493467e-01 -6.26985073e-01 -7.19757020e-01 -5.95506787e-01 -1.30003124e-01 2.10879639e-01 -7.42467120e-02 7.99612701e-01 -2.07147092e-01 3.47853228e-02 7.05381215e-01 -7.42361486e-01 -3.27347517e-01 6.33119106e-01 1.04633069e+00 4.79255885e-01 2.80941963e-01 5.80664992e-01 -6.89817429e-01 1.59040228e-01 -5.96786916e-01 -1.19114125e+00 3.21169287e-01 -1.06453919e+00 5.00791788e-01 4.32887256e-01 -5.68147600e-01 -1.23663425e+00 9.98265594e-02 -1.67115062e-01 -1.83993429e-01 -2.56993979e-01 4.72440302e-01 -6.75059780e-02 -3.26826811e-01 8.67318571e-01 -7.49815330e-02 -5.83461672e-02 -8.78345311e-01 -7.00968355e-02 1.17098391e+00 5.27699471e-01 -6.19325101e-01 1.10233557e+00 2.09905103e-01 3.68015885e-01 -7.04737008e-01 -1.02823818e+00 -5.41642666e-01 -5.46234727e-01 -5.73114991e-01 5.04756451e-01 -6.70058668e-01 -2.76129693e-01 4.10922348e-01 -7.92301714e-01 8.90382379e-02 -1.45065218e-01 1.00475705e+00 -6.22329056e-01 4.51378226e-01 1.84002414e-01 -1.46954477e+00 1.38500527e-01 -1.19805527e+00 6.91020310e-01 3.80005717e-01 -2.09504008e-01 -9.31401610e-01 2.33613715e-01 3.64928842e-01 1.29221916e-01 -5.44643924e-02 7.06871629e-01 -9.40125465e-01 -4.81009215e-01 -2.98316151e-01 1.24581702e-01 4.68108892e-01 -2.66481806e-02 3.39276373e-01 -1.22534764e+00 -1.93825483e-01 3.87547165e-01 -4.57653135e-01 9.86869097e-01 3.11565429e-01 1.19876444e+00 1.35720491e-01 -3.00835222e-01 4.41728890e-01 1.33940208e+00 2.23294303e-01 4.41955209e-01 5.42530715e-01 2.96475738e-01 9.00147200e-01 9.60337102e-01 3.79342884e-01 -1.18802330e-02 7.86608875e-01 6.54573917e-01 1.90465495e-01 1.82945043e-01 -8.27355161e-02 2.63321698e-01 5.83595037e-01 -3.49503338e-01 3.72979231e-02 -9.20833886e-01 4.39780265e-01 -1.71008790e+00 -8.76328528e-01 -7.62424707e-01 2.72809529e+00 8.16678166e-01 3.01163256e-01 7.40280896e-02 4.63201910e-01 6.62567019e-01 -7.72081688e-02 -4.31338847e-01 -1.42309293e-01 -3.29785347e-01 1.72795966e-01 8.11120391e-01 3.69991362e-01 -1.02916551e+00 5.77247024e-01 7.07569695e+00 8.20903480e-01 -9.31740344e-01 8.31050873e-02 3.98941338e-01 1.38442442e-01 -4.24727380e-01 5.13994515e-01 -1.04000175e+00 5.12822449e-01 8.99679601e-01 1.55731723e-01 1.17540555e-02 1.07974160e+00 2.54766047e-02 -5.01702726e-01 -7.32373893e-01 8.70765328e-01 2.71201491e-01 -7.54626691e-01 -3.35923135e-01 1.61157370e-01 5.13469040e-01 -7.64307985e-03 -3.08386415e-01 2.51382649e-01 4.12657797e-01 -9.01437461e-01 7.38472521e-01 8.58831525e-01 4.24115241e-01 -7.30705798e-01 9.61553454e-01 4.41655874e-01 -6.47970974e-01 -7.15138316e-02 -6.76912427e-01 -2.07524672e-01 -1.27704054e-01 9.76051033e-01 -1.05636060e+00 5.66052616e-01 7.96614587e-01 2.89379627e-01 -9.84654725e-01 1.51814580e+00 -5.02815843e-01 1.00448287e+00 -4.15531486e-01 -3.34237725e-01 -2.20095232e-01 -3.73980194e-01 6.07814789e-01 7.39290893e-01 2.82508999e-01 -3.01070809e-01 -2.25563645e-01 7.71634579e-01 5.08609712e-01 1.04087226e-01 -5.44208705e-01 1.32939383e-01 2.93642431e-01 1.33096230e+00 -6.11661673e-01 -3.00131202e-01 -6.60545588e-01 1.83544517e-01 1.53641611e-01 5.12537844e-02 -7.76067734e-01 -9.54132378e-02 3.79238963e-01 -6.70664161e-02 3.13468158e-01 -4.91602719e-02 -4.46129203e-01 -9.94619966e-01 -2.59257972e-01 -8.29384148e-01 4.83113199e-01 -8.13348889e-01 -1.41562688e+00 4.87157136e-01 2.92768419e-01 -1.21116662e+00 -4.45173264e-01 -9.81381238e-01 -2.47353762e-01 9.90812838e-01 -1.35666406e+00 -6.31352067e-01 -2.21843973e-01 -5.08846119e-02 2.46992916e-01 -6.62908316e-01 6.08878314e-01 3.36011616e-03 -3.80303532e-01 2.16155604e-01 4.17769909e-01 -2.56503582e-01 8.47902536e-01 -1.29000509e+00 -1.56238720e-01 7.33996928e-01 2.61974931e-01 4.82734442e-02 1.18086672e+00 -6.91780388e-01 -8.04821551e-01 -7.21431851e-01 8.61686766e-01 -6.48781061e-01 7.05120981e-01 6.12638183e-02 -9.96778011e-01 4.12719309e-01 -3.89199227e-01 -1.79911867e-01 6.68687224e-01 3.33661914e-01 -3.04109484e-01 -5.27070314e-02 -1.02342200e+00 1.10288374e-01 5.99341035e-01 -3.43900263e-01 -1.07916880e+00 2.82878369e-01 2.33655617e-01 1.66538909e-01 -8.64061117e-01 6.96307242e-01 5.55896819e-01 -1.05254579e+00 8.28942418e-01 -6.39275491e-01 -1.93875488e-02 -6.72854066e-01 -5.61459996e-02 -1.29615259e+00 -2.71625906e-01 -1.52964458e-01 -1.51906740e-02 1.26134968e+00 4.74867642e-01 -6.78066969e-01 5.51258147e-01 4.51541781e-01 -4.65891771e-02 -4.31374997e-01 -8.86195421e-01 -9.61326182e-01 1.25625491e-01 -5.39409161e-01 4.43645656e-01 7.50643492e-01 -4.16841060e-01 6.94540069e-02 -3.37521851e-01 1.47401005e-01 7.82659531e-01 1.55866981e-01 7.96975076e-01 -1.90592158e+00 -6.61907494e-01 -4.79350030e-01 -5.06990552e-01 -5.88897288e-01 3.33405554e-01 -6.80597782e-01 3.48438501e-01 -1.01115394e+00 2.15820432e-01 -8.57295513e-01 -1.88037902e-01 2.49255285e-01 -2.27394983e-01 3.30703020e-01 -1.60307258e-01 5.20467818e-01 -1.38275951e-01 4.04168636e-01 7.16996431e-01 -2.27351952e-02 1.22307509e-01 4.16743100e-01 -2.80956089e-01 1.07840216e+00 1.02340305e+00 -8.24016988e-01 1.92996196e-03 3.46932650e-01 4.02898520e-01 -3.94902498e-01 4.40451115e-01 -1.06824446e+00 -1.84100330e-01 -2.71694928e-01 3.18903118e-01 -8.23322237e-01 2.81687707e-01 -7.42391586e-01 4.01914477e-01 4.45970684e-01 -2.73471791e-02 -2.75091022e-01 -3.13662469e-01 3.32567781e-01 -2.75558501e-01 -1.16753340e+00 9.48791504e-01 -9.32555124e-02 -4.52420533e-01 -1.70579389e-01 -4.47662413e-01 -1.48110718e-01 9.74291503e-01 -9.63436381e-04 3.43849100e-02 -1.99089527e-01 -4.35412198e-01 -1.08221196e-01 6.07568741e-01 2.00266719e-01 2.00888082e-01 -1.00661099e+00 -6.59308791e-01 5.24972491e-02 3.85099709e-01 -5.39140888e-02 -4.35354203e-01 7.17646360e-01 -5.06415963e-01 5.49654543e-01 3.67236435e-02 -7.54925847e-01 -1.28493059e+00 6.66787148e-01 5.41452587e-01 -7.88329765e-02 -2.06343621e-01 6.22024834e-01 2.02431381e-02 -4.23285663e-01 3.18351626e-01 -6.67740479e-02 -5.46688557e-01 1.91819191e-01 2.51697719e-01 3.53094369e-01 2.22806960e-01 -7.28565276e-01 -1.51735157e-01 4.94895041e-01 2.96975434e-01 -1.13243967e-01 1.18532574e+00 1.29036993e-01 -2.28774160e-01 8.95466030e-01 3.13002557e-01 2.54350543e-01 -1.05572128e+00 -2.57628530e-01 4.54854637e-01 -7.12208688e-01 1.20132439e-01 -8.63668144e-01 -6.40103400e-01 5.78807056e-01 5.88105321e-01 5.35486341e-01 1.03895450e+00 4.10263240e-01 -1.69538483e-01 4.01836127e-01 5.18823981e-01 -1.08214498e+00 -4.59571034e-01 2.40464926e-01 7.94927299e-01 -1.51245368e+00 2.04447284e-01 -3.87722552e-01 -4.72673327e-01 1.21503174e+00 4.25618410e-01 2.00866371e-01 5.33897698e-01 4.57412750e-02 3.91707122e-02 -4.74020950e-02 -8.21523309e-01 -4.34225708e-01 2.96557635e-01 5.79645753e-01 5.30673385e-01 2.16397330e-01 -9.55885112e-01 5.20644724e-01 -3.25105816e-01 -1.75491184e-01 5.53594112e-01 5.95299006e-01 -5.96152484e-01 -1.37523937e+00 -1.22968745e+00 6.45237505e-01 -5.19295871e-01 7.81145468e-02 -6.73793256e-01 7.26521850e-01 2.21558884e-01 1.09262168e+00 -4.07106755e-03 -2.49733269e-01 -8.12831894e-02 4.87532318e-01 4.00731862e-01 -3.20101976e-01 -4.14991379e-01 -1.32630453e-01 4.33899283e-01 1.40824392e-01 -7.87347972e-01 -9.54225659e-01 -7.72481918e-01 3.30226161e-02 -8.81774724e-01 6.95535719e-01 1.02079058e+00 1.08374631e+00 -4.29481298e-01 3.44104916e-01 8.77048433e-01 -6.79617405e-01 -1.00027251e+00 -1.27940798e+00 -7.03233421e-01 3.56786877e-01 5.09887077e-02 -1.31176615e+00 -1.02877867e+00 -2.20985726e-01]
[7.642273426055908, 3.4706814289093018]
54efc717-2a99-43d3-b47e-1472baa52a42
disfluency-detection-for-vietnamese
null
null
https://aclanthology.org/2022.wnut-1.21
https://aclanthology.org/2022.wnut-1.21.pdf
Disfluency Detection for Vietnamese
In this paper, we present the first empirical study for Vietnamese disfluency detection. To conduct this study, we first create a disfluency detection dataset for Vietnamese, with manual annotations over two disfluency types. We then empirically perform experiments using strong baseline models, and find that: automatic Vietnamese word segmentation improves the disfluency detection performances of the baselines, and the highest performance results are obtained by fine-tuning pre-trained language models in which the monolingual model PhoBERT for Vietnamese does better than the multilingual model XLM-R.
['Dat Quoc Nguyen', 'Thinh Hung Truong', 'Mai Dao']
null
null
null
null
coling-wnut-2022-10
['vietnamese-word-segmentation', 'xlm-r']
['natural-language-processing', 'natural-language-processing']
[-6.30410194e-01 -2.81007260e-01 -4.02025491e-01 -7.30870664e-02 -8.38910639e-01 -1.15150642e+00 3.84722799e-01 -2.04185247e-01 -8.22571516e-01 8.43751609e-01 5.40424585e-01 -5.62661052e-01 7.15794861e-01 -1.76589131e-01 -2.55776793e-01 -3.33541363e-01 -1.37047321e-01 5.62272429e-01 1.96685806e-01 -4.39298421e-01 3.22083384e-02 2.48779565e-01 -8.44120681e-01 8.69394690e-02 1.17480981e+00 9.64933634e-02 2.88431466e-01 5.20139575e-01 3.94440979e-01 6.84046805e-01 -1.09936130e+00 -3.63966376e-01 2.24886835e-01 -3.83132011e-01 -1.02810490e+00 -3.76355559e-01 1.58168212e-01 -2.83050537e-01 -2.78713882e-01 1.09720850e+00 3.52026433e-01 -5.28088724e-03 5.69186807e-01 -7.19311476e-01 -8.70452166e-01 1.02105701e+00 -1.02811202e-01 1.05396521e+00 6.03852987e-01 4.23042059e-01 9.51887608e-01 -1.13543701e+00 1.09428346e+00 1.29297388e+00 3.06806922e-01 5.73747635e-01 -1.02516234e+00 -8.24661851e-01 2.95847714e-01 8.96254331e-02 -1.81836474e+00 -4.81514305e-01 4.76182491e-01 -4.63053197e-01 2.14531970e+00 1.54796213e-01 4.54996079e-01 1.26902294e+00 3.38167280e-01 5.17336786e-01 1.09024858e+00 -6.88990891e-01 -3.32514912e-01 1.35799006e-01 5.09705365e-01 3.98518622e-01 2.72226363e-01 3.35233301e-01 -2.45835841e-01 3.10255498e-01 6.23216152e-01 -7.14089274e-01 -4.14798051e-01 1.01068902e+00 -1.29173636e+00 7.89988399e-01 9.37895700e-02 1.15142167e+00 -6.66219220e-02 -4.64511067e-01 5.76769531e-01 6.93532348e-01 9.03276801e-01 6.37512445e-01 -6.33662760e-01 -3.01179111e-01 -9.07056093e-01 2.49120250e-01 6.67461872e-01 1.10103023e+00 5.00171900e-01 3.72675508e-01 -3.23428363e-01 8.01429868e-01 1.00047246e-03 5.41182578e-01 5.69266081e-01 -6.68761849e-01 8.53168607e-01 3.71159352e-02 1.20608613e-01 -5.46524823e-01 -4.41421390e-01 -1.65077823e-03 -2.29051068e-01 -3.75740021e-01 4.10205156e-01 -4.69442487e-01 -1.25145173e+00 1.66992450e+00 -2.22799718e-01 -4.76650000e-01 1.75823733e-01 8.44834805e-01 6.12845838e-01 1.08054602e+00 6.33860588e-01 -8.35147619e-01 1.02320755e+00 -1.03932571e+00 -1.21671820e+00 -3.70529562e-01 1.16064787e+00 -1.01311970e+00 1.69447625e+00 5.13797402e-01 -1.08703661e+00 -5.96552849e-01 -1.08606470e+00 -2.07458004e-01 -4.36318785e-01 -7.19471946e-02 5.64084530e-01 8.90813231e-01 -1.11899507e+00 1.09474093e-01 -9.10776973e-01 -2.72605509e-01 -3.31578881e-01 1.46733299e-01 -2.51007199e-01 -4.34814058e-02 -1.82834184e+00 1.25394940e+00 9.33782876e-01 -1.06438741e-01 -7.87432730e-01 -5.66749215e-01 -9.25696611e-01 -1.90861195e-01 1.40400767e-01 1.12326749e-01 1.11656594e+00 -9.77232158e-01 -1.21063137e+00 1.26443839e+00 -3.83855641e-01 -4.19405177e-02 1.65601030e-01 -4.06058043e-01 -1.10762048e+00 -2.28105143e-01 3.32923502e-01 5.30254245e-01 1.46230280e-01 -1.14290440e+00 -8.10286462e-01 -1.18933432e-01 -1.46716371e-01 2.32381254e-01 -3.04470539e-01 9.65062380e-01 -5.50858676e-01 -1.09641254e+00 -4.60182697e-01 -9.13663328e-01 9.38646030e-03 -1.67340028e+00 -4.07497585e-01 -7.81313956e-01 6.16492748e-01 -1.18052971e+00 2.24311352e+00 -2.14634681e+00 5.77012252e-04 1.13763049e-01 7.06861913e-02 7.04360425e-01 -7.17441887e-02 1.73146769e-01 -2.03692734e-01 6.22642219e-01 4.60140072e-02 -2.71709085e-01 -1.71131045e-01 4.61456567e-01 -4.94824871e-02 4.21371192e-01 3.34571451e-01 1.01183522e+00 -8.11186969e-01 -6.58130050e-01 -8.42467770e-02 2.00451702e-01 -5.87337315e-01 3.27464640e-01 -2.27622643e-01 4.89345789e-01 3.02574068e-01 8.89950156e-01 6.63072228e-01 4.14287716e-01 5.93695462e-01 2.08462313e-01 -7.16237545e-01 8.05240870e-01 -5.45027792e-01 1.64082706e+00 -4.40695614e-01 4.39354718e-01 -3.22742276e-02 -3.41890603e-01 4.50889587e-01 6.70278132e-01 -8.86132941e-02 -9.03834939e-01 1.75033659e-01 5.25714338e-01 4.26458478e-01 -7.38103092e-01 6.65867567e-01 -1.73835024e-01 -5.06821454e-01 1.54111341e-01 4.05972451e-01 5.64276660e-03 7.33183384e-01 1.43357128e-01 8.44263196e-01 -3.05810124e-01 4.99746799e-01 -7.64252186e-01 4.33184594e-01 4.98953044e-01 8.67418051e-01 4.51122612e-01 -5.01718163e-01 3.30369234e-01 2.94884831e-01 -2.37064421e-01 -6.73985779e-01 -1.34240925e+00 -1.93769693e-01 1.31868780e+00 -1.33482784e-01 -6.88028634e-01 -9.74315584e-01 -1.04773211e+00 -5.10471582e-01 1.05985630e+00 -1.78477213e-01 1.27033949e-01 -1.54022503e+00 -8.87468100e-01 9.18103516e-01 6.77177131e-01 3.14473242e-01 -1.53049064e+00 1.69265680e-02 2.82762378e-01 -7.13883221e-01 -1.20614958e+00 -1.09922576e+00 3.15577358e-01 -2.52564728e-01 -6.25405490e-01 -3.97387147e-01 -1.60115576e+00 -3.99709977e-02 -1.46157578e-01 1.55744123e+00 3.50940645e-01 1.35769425e-02 -2.68604100e-01 -5.88618636e-01 -7.36989500e-03 -6.80362999e-01 3.09118778e-01 2.83865035e-01 -9.12632644e-01 9.57377374e-01 -1.67482883e-01 -7.77783990e-02 -6.09270260e-02 -6.94003284e-01 -6.63604021e-01 2.44388372e-01 5.20185113e-01 5.11177540e-01 -1.02871321e-01 4.63089913e-01 -1.04695129e+00 6.94365978e-01 -6.54426515e-01 -4.77647036e-01 2.21778214e-01 -6.91956878e-01 -2.61418700e-01 7.66437113e-01 -5.87763965e-01 -1.29578888e+00 -6.27188385e-02 -7.15299904e-01 -1.61169887e-01 -2.70892411e-01 6.64164364e-01 -7.75135338e-01 3.51254821e-01 6.48104072e-01 -1.65716246e-01 -8.82776201e-01 -5.01025558e-01 3.55798990e-01 7.58430302e-01 8.93836498e-01 -3.80205929e-01 3.86300147e-01 -2.97274500e-01 -1.03346276e+00 -1.04744005e+00 -5.68141878e-01 -5.26804328e-01 -1.04291797e+00 1.16956361e-01 1.21128058e+00 -1.05058897e+00 -1.63986430e-01 4.69060868e-01 -1.55032408e+00 -3.75123858e-01 1.47670969e-01 4.52392042e-01 -1.29122347e-01 2.65522897e-01 -1.25558805e+00 -3.63465250e-01 -2.47664973e-01 -1.12945735e+00 5.50849199e-01 -2.00037763e-01 -5.95184982e-01 -1.38416731e+00 6.12818182e-01 -2.34984457e-02 1.02889836e-01 -6.83564544e-02 1.11577141e+00 -6.15037322e-01 2.26249956e-02 2.67703414e-01 8.58764574e-02 4.10950363e-01 1.75916910e-01 -5.61626405e-02 -7.35925853e-01 -3.44754487e-01 1.77010372e-02 -3.53700161e-01 9.28441107e-01 4.38535750e-01 2.98140585e-01 -5.33092320e-01 -2.41628468e-01 7.35172689e-01 1.32962048e+00 5.71156025e-01 6.37147605e-01 2.00867400e-01 9.07347500e-01 4.01744395e-01 7.89294541e-01 6.34100139e-02 7.63439834e-01 5.61750054e-01 -1.78183109e-01 -3.00057709e-01 -3.63004625e-01 -2.44906306e-01 7.98045695e-01 1.31960475e+00 -1.69799149e-01 -8.99131179e-01 -1.10831332e+00 7.73815036e-01 -1.35804594e+00 -7.91958988e-01 -3.00244063e-01 2.01293492e+00 1.18546963e+00 1.78934857e-01 4.49741781e-01 5.78143336e-02 5.80629349e-01 2.28949085e-01 1.98834777e-01 -7.16867626e-01 -4.98173326e-01 4.60147709e-01 2.78697282e-01 1.15982389e+00 -1.34082139e+00 2.30376744e+00 8.10474205e+00 8.04099023e-01 -1.09997201e+00 6.14792228e-01 2.75655150e-01 -1.58994342e-03 -4.45178390e-01 -1.02493264e-01 -1.16204786e+00 3.52567405e-01 1.31254518e+00 -4.60303463e-02 5.63124299e-01 3.12488019e-01 3.10761213e-01 -1.04114212e-01 -9.65677559e-01 7.56927490e-01 1.92501292e-01 -8.81096542e-01 -1.46667466e-01 -1.23518236e-01 7.70681918e-01 1.92173153e-01 -1.12682953e-01 5.01854002e-01 5.02157807e-01 -1.22849226e+00 9.04760659e-01 -2.07662657e-01 1.06729174e+00 -1.05506241e+00 6.07965410e-01 1.73052654e-01 -1.07707846e+00 2.19888836e-01 -2.83083707e-01 -1.99874267e-01 4.63555545e-01 4.49663699e-01 -6.52135968e-01 2.09191978e-01 6.61629915e-01 6.88300312e-01 -5.82639396e-01 3.58279705e-01 -5.16919971e-01 1.47234821e+00 -1.76146388e-01 2.23995760e-01 1.33024067e-01 -6.67276233e-02 5.83179355e-01 2.04330492e+00 -5.99500388e-02 1.69843972e-01 6.41413033e-01 7.56525695e-01 1.12833239e-01 3.72790247e-01 -5.27736723e-01 -1.74326032e-01 4.95935738e-01 8.80199075e-01 -8.39867771e-01 -4.12156731e-01 -5.93973875e-01 1.49548113e+00 6.71300292e-01 5.41895509e-01 -8.14619064e-01 -5.14003396e-01 8.16501379e-01 -6.53807372e-02 -3.60326767e-02 -8.27456176e-01 -3.40414673e-01 -1.34728205e+00 -1.92724466e-01 -1.02460194e+00 6.88818276e-01 6.29682690e-02 -1.07340729e+00 1.06838548e+00 1.20958872e-01 -7.01145709e-01 -7.45960116e-01 -5.96256793e-01 -7.09791124e-01 1.13320398e+00 -1.28291202e+00 -1.13881969e+00 4.34944212e-01 7.75211215e-01 8.05934727e-01 5.32915667e-02 9.07800853e-01 6.15191996e-01 -8.68743837e-01 9.58081782e-01 -4.06953901e-01 3.74763221e-01 8.15081537e-01 -1.41487420e+00 7.46925950e-01 1.51653159e+00 3.45962495e-01 7.06018865e-01 5.85297167e-01 -1.30356240e+00 -6.13962591e-01 -1.01424956e+00 1.94018257e+00 -5.87825358e-01 6.84708416e-01 -6.30906820e-01 -9.01744962e-01 1.23131669e+00 5.89424670e-01 -3.62386405e-01 5.48472762e-01 4.05151635e-01 -2.97745645e-01 5.71994007e-01 -1.02504611e+00 3.70444268e-01 1.40673649e+00 -3.94651532e-01 -6.67710900e-01 4.73499686e-01 1.12888014e+00 -2.55678236e-01 -8.48235369e-01 2.78110802e-01 2.07572684e-01 -8.57545137e-01 5.94884694e-01 -6.13892198e-01 -1.03970505e-01 -6.41928688e-02 -1.65494859e-01 -1.60130680e+00 -7.15649307e-01 -5.50175905e-01 3.75721753e-01 1.53690469e+00 7.45427668e-01 -5.79567790e-01 1.32729128e-01 -1.65871810e-02 -5.00992298e-01 1.16901502e-01 -9.34263587e-01 -1.30036283e+00 1.02685940e+00 -3.77070040e-01 2.23950669e-01 1.11092198e+00 4.39860553e-01 5.38054883e-01 -1.79317534e-01 1.68456197e-01 3.16056237e-03 -3.04102719e-01 -1.19052805e-01 -7.44934320e-01 -1.11453965e-01 -3.18715811e-01 3.55444439e-02 -9.05252934e-01 8.15830112e-01 -1.00389206e+00 3.09151500e-01 -8.44011843e-01 -5.63330809e-03 -3.65418255e-01 -1.00327201e-01 5.71126163e-01 -5.97394228e-01 5.16231835e-01 1.49457082e-01 3.19276959e-01 -3.67924094e-01 6.15603104e-02 8.95179570e-01 1.32499054e-01 -8.12446058e-01 -3.48530054e-01 -4.79728252e-01 5.60232997e-01 9.00791824e-01 -4.88878965e-01 -1.39583930e-01 -8.03075969e-01 -9.58981365e-02 -1.30918678e-02 -3.27968627e-01 -7.00609028e-01 -2.82211483e-01 -2.22339958e-01 2.60237128e-01 -3.32730591e-01 -3.45042646e-01 -1.88035797e-02 -2.65664876e-01 4.69471991e-01 9.61003751e-02 9.26942945e-01 3.61858219e-01 -4.92169023e-01 -3.46087158e-01 -6.71309307e-02 7.50346243e-01 -2.72430927e-01 -8.97893310e-01 4.35008444e-02 -1.14308441e+00 6.98551714e-01 7.95206010e-01 1.18107289e-01 -3.26881975e-01 1.89061370e-02 -6.14075780e-01 9.95591432e-02 3.77849042e-01 6.89777792e-01 1.63093597e-01 -1.03696406e+00 -8.01870763e-01 4.97043312e-01 -2.58302595e-03 -6.63407266e-01 -2.71143138e-01 4.39948261e-01 -7.07294345e-01 9.85882938e-01 -9.03246105e-02 -2.19056457e-02 -9.48182225e-01 9.98004675e-01 8.36712942e-02 -5.98720253e-01 -2.82117754e-01 8.77647340e-01 1.32898167e-01 -2.81644791e-01 1.32298172e-01 -5.13887882e-01 -4.33444470e-01 -1.53796792e-01 4.52798456e-01 5.02040744e-01 3.43401879e-01 -1.31883097e+00 -7.92190433e-01 2.40377471e-01 -2.72675842e-01 -4.66553658e-01 5.78899205e-01 -2.47910693e-01 -1.53035492e-01 4.14274901e-01 1.05885756e+00 7.00748563e-01 -5.20454109e-01 3.09257627e-01 3.69108468e-01 -2.50321031e-01 6.00902066e-02 -1.17033601e+00 -9.12512779e-01 4.61617678e-01 3.64868104e-01 2.30634063e-02 1.19005227e+00 2.74901956e-01 9.53227341e-01 -4.79105636e-02 2.11763680e-01 -1.30345511e+00 -2.41051748e-01 1.18889463e+00 7.32282162e-01 -1.11534274e+00 -8.27311039e-01 -8.16574037e-01 -8.56403768e-01 6.96131885e-01 1.01737928e+00 -2.58616030e-01 9.22565222e-01 6.07719183e-01 7.16886103e-01 -1.14011303e-01 -4.50822711e-01 -4.91650373e-01 2.29511887e-01 7.38271892e-01 1.03412008e+00 4.91738111e-01 -1.22957969e+00 9.37253535e-01 -6.59083962e-01 -4.91573572e-01 2.59874135e-01 6.91768885e-01 -3.09914052e-01 -1.40667987e+00 -2.98534721e-01 1.34236980e-02 -8.72968793e-01 -7.09293008e-01 -6.51969969e-01 1.11533260e+00 4.51288491e-01 1.40113592e+00 3.22021872e-01 -5.83712816e-01 4.04250294e-01 1.22856811e-01 5.56795955e-01 -8.35332513e-01 -1.01339543e+00 4.60061222e-01 2.77130961e-01 -6.63765907e-01 -6.70697093e-02 -6.40009284e-01 -1.46887541e+00 -2.86148399e-01 -2.25982696e-01 2.04877511e-01 -1.89764455e-01 1.18372357e+00 -1.58243552e-01 -7.19162151e-02 4.93667364e-01 -6.43880129e-01 3.53120118e-02 -1.25405192e+00 -6.92324579e-01 2.94408143e-01 2.72004545e-01 -4.08859968e-01 -6.65961385e-01 1.58446476e-01]
[10.541409492492676, 9.964916229248047]
5a013e93-87ae-40a8-9583-fcbc9f464535
ganglionnet-objectively-assess-the-density
2007.02367
null
https://arxiv.org/abs/2007.02367v1
https://arxiv.org/pdf/2007.02367v1.pdf
GanglionNet: Objectively Assess the Density and Distribution of Ganglion Cells With NABLA-N Network
Hirschsprungs disease (HD) is a birth defect which is diagnosed and managed by multiple medical specialties such as pediatric gastroenterology, surgery, radiology, and pathology. HD is characterized by absence of ganglion cells in the distal intestinal tract with a gradual normalization of ganglion cell numbers in adjacent upstream bowel, termed as the transition zone (TZ). Definitive surgical management to remove the abnormal bowel requires accurate assessment of ganglion cell density in histological sections from the TZ, which is difficult, time-consuming and prone to operator error. We present an automated method to detect and count immunostained ganglion cells using a new NABLA_N network based deep learning (DL) approach, called GanglionNet. The morphological image analysis methods are applied for refinement of the regions for counting of the cells and define ganglia regions (a set of ganglion cells) from the predicted masks. The proposed model is trained with single point annotated samples by the expert pathologist. The GanglionNet is tested on ten completely new High Power Field (HPF) images with dimension of 2560x1920 pixels and the outputs are compared against the manual counting results by the expert pathologist. The proposed method shows a robust 97.49% detection accuracy for ganglion cells, when compared to counts by the expert pathologist, which demonstrates the robustness of GanglionNet. The proposed DL based ganglion cell detection and counting method will simplify and standardize TZ diagnosis for HD patients.
['Vijayan K. Asari', 'TJ Browen', 'Raj P. Kapur', 'Md Zahangir Alom']
2020-07-05
null
null
null
null
['cell-detection']
['computer-vision']
[ 1.20104201e-01 3.83700252e-01 1.21031947e-01 2.84328848e-01 -1.18829221e-01 -5.92373729e-01 3.71741690e-02 7.71760643e-01 -7.18861759e-01 7.03853428e-01 -2.39397570e-01 -2.38573804e-01 2.18157202e-01 -9.42196965e-01 -5.93353987e-01 -9.42205429e-01 -3.37336779e-01 5.74040174e-01 3.24658126e-01 1.99264005e-01 1.43952489e-01 5.16311765e-01 -1.25145328e+00 1.13051295e-01 9.00538385e-01 8.63520980e-01 4.50955600e-01 9.69197094e-01 6.00363538e-02 7.00436115e-01 -4.93124098e-01 2.24751666e-01 5.39567769e-01 -5.59321463e-01 -5.42944729e-01 -3.54190432e-02 2.20179990e-01 -4.74946558e-01 2.55000412e-01 1.42162406e+00 4.44345772e-01 -2.05098704e-01 9.46363211e-01 -4.19711053e-01 -2.47113183e-01 3.65426570e-01 -7.66862392e-01 4.14652854e-01 -1.13897145e-01 1.07607968e-01 2.97957003e-01 -3.76639456e-01 7.39448786e-01 7.20239341e-01 8.33998084e-01 5.03565371e-01 -1.23475826e+00 -6.35074496e-01 -4.56296086e-01 -1.04302526e-01 -1.49579644e+00 2.57189423e-01 -4.84423861e-02 -8.05548966e-01 9.69900191e-01 2.02064931e-01 1.34691572e+00 3.22585963e-02 3.18759441e-01 5.14020145e-01 1.16068530e+00 -6.87195837e-01 2.26337478e-01 -1.23576038e-01 -1.34473592e-01 1.19948280e+00 7.22803056e-01 8.58758688e-02 2.67853111e-01 1.61848769e-01 1.52933156e+00 1.25885874e-01 -2.68600494e-01 -2.08861977e-01 -1.12424290e+00 8.07291925e-01 5.81276536e-01 5.43253243e-01 -5.57853937e-01 -6.12031072e-02 5.97270370e-01 7.45860413e-02 2.64485657e-01 4.08202797e-01 -1.54622883e-01 5.89671612e-01 -8.70183170e-01 4.72810445e-03 6.35650516e-01 7.27725327e-01 4.36481446e-01 -1.26035675e-01 -9.75490361e-02 7.80461490e-01 1.16573013e-01 1.37138814e-01 6.32584393e-01 -8.58646989e-01 -2.32207879e-01 1.22430170e+00 -1.85314536e-01 -8.03575635e-01 -6.35708690e-01 -5.24367154e-01 -1.32815242e+00 7.20204294e-01 5.19841909e-01 -1.86033487e-01 -1.19503641e+00 1.11767006e+00 5.88777483e-01 -6.79778755e-02 -1.32783771e-01 1.09706020e+00 8.18216503e-01 4.78975296e-01 1.16670072e-01 -1.57170132e-01 1.33434188e+00 -9.44615781e-01 -3.94242465e-01 1.50906751e-02 1.06921136e+00 -5.83676159e-01 5.07495165e-01 5.53769767e-01 -9.11648870e-01 -6.01863787e-02 -1.19887197e+00 -1.79941371e-01 -1.79318026e-01 6.11387789e-01 5.23887038e-01 1.97269693e-01 -1.17265487e+00 3.11569780e-01 -1.09636116e+00 -9.37122166e-01 4.73569036e-01 7.84981191e-01 -5.92218816e-01 3.09742898e-01 -4.22648638e-01 8.58388603e-01 5.98976552e-01 -5.01674153e-02 -7.48441935e-01 -5.82016706e-01 -7.24534512e-01 7.76670054e-02 -2.59795070e-01 -1.01428390e+00 1.03852189e+00 -8.93494725e-01 -1.17249143e+00 1.60983396e+00 1.24289624e-01 -9.80270624e-01 5.49937904e-01 3.88413101e-01 2.36204177e-01 4.83297378e-01 2.31529653e-01 8.58378768e-01 1.92651898e-01 -9.99278963e-01 -1.41255903e+00 -4.33197767e-01 -1.93729952e-01 2.42447823e-01 3.44049841e-01 -2.93246537e-01 -6.26982003e-02 -6.20666385e-01 4.04843390e-01 -6.18341804e-01 -4.29887384e-01 2.50501901e-01 -3.01219434e-01 1.72605202e-01 2.43736446e-01 -9.03272152e-01 9.34726179e-01 -2.05515766e+00 -3.51482421e-01 3.74805540e-01 4.84147489e-01 3.92778218e-01 3.67186278e-01 -6.36173636e-02 6.57454282e-02 -1.19581588e-01 -5.20661026e-02 1.05532646e-01 -5.26066840e-01 4.61647622e-02 4.64761287e-01 8.99357557e-01 -3.17076564e-01 4.09072727e-01 -9.19479847e-01 -7.95569599e-01 5.57468355e-01 2.88023204e-01 -5.85693240e-01 3.91981490e-02 2.13688239e-02 6.38890982e-01 2.64984835e-02 8.60265076e-01 5.98537028e-01 -3.79757166e-01 2.93951452e-01 -8.46041515e-02 -3.35413158e-01 -2.65966147e-01 -8.56845021e-01 1.29699242e+00 -1.79731756e-01 3.56972307e-01 5.65440059e-01 -1.00041080e+00 9.58129168e-01 4.06414509e-01 2.72143364e-01 -5.67434669e-01 5.78703105e-01 5.00081122e-01 3.21185559e-01 -6.32582188e-01 -1.94892839e-01 -4.36114252e-01 2.53083169e-01 7.94216245e-02 4.11947131e-01 5.96811771e-02 6.63754225e-01 -3.90069008e-01 9.72204387e-01 -4.15717721e-01 1.27317750e+00 -5.72309375e-01 5.79852879e-01 6.25987411e-01 6.89930201e-01 5.64121604e-01 -3.35010558e-01 4.88977253e-01 5.32686830e-01 -8.30897570e-01 -1.27379251e+00 -8.99610877e-01 -3.59356612e-01 5.21827281e-01 2.05402330e-01 4.96393800e-01 -9.76372123e-01 -3.82989377e-01 1.68581940e-02 8.93276185e-02 -7.88361132e-01 3.65601361e-01 -6.94732666e-01 -8.12939703e-01 4.89974439e-01 2.12698072e-01 6.58935428e-01 -1.03863740e+00 -8.89588058e-01 3.99588972e-01 5.97417019e-02 -5.37923217e-01 -9.43414308e-03 3.89258891e-01 -9.24964368e-01 -1.27357602e+00 -1.01052213e+00 -1.58920515e+00 1.26490736e+00 -1.72851875e-01 5.56482017e-01 5.46782494e-01 -8.45023096e-01 -3.13301444e-01 -2.46947378e-01 -6.04770839e-01 -6.89537585e-01 2.75695766e-03 -2.92262197e-01 -6.37478054e-01 3.85051042e-01 -5.98694980e-01 -1.07452071e+00 -1.07282497e-01 -6.08224154e-01 2.93465853e-01 8.93896282e-01 1.03426468e+00 1.10025823e+00 -1.42351165e-01 2.08662525e-01 -9.19280767e-01 3.52292776e-01 -4.23882365e-01 -9.88061607e-01 -6.14814973e-03 -5.04073143e-01 -2.53187537e-01 6.52888298e-01 -3.69539082e-01 -5.16123295e-01 7.49391764e-02 -5.79311401e-02 -1.33896396e-01 -4.25645888e-01 3.39126229e-01 6.96040392e-01 -4.06532675e-01 7.94001937e-01 8.45920369e-02 3.26080561e-01 -1.45157024e-01 -4.01736528e-01 3.50257754e-01 9.61082399e-01 5.82542829e-02 -2.78759990e-02 6.42332315e-01 2.97624946e-01 -4.53669995e-01 -3.82349283e-01 -8.28796148e-01 -4.21275347e-01 -9.74607468e-02 9.94689941e-01 -8.03101659e-01 -6.20258808e-01 4.28540260e-01 -1.02645922e+00 -4.44780499e-01 -1.26165763e-01 8.96624386e-01 -4.87535089e-01 9.94469151e-02 -1.18073308e+00 -5.43445885e-01 -9.81336355e-01 -9.00829077e-01 5.74327528e-01 4.92196798e-01 -2.86982745e-01 -1.15107489e+00 3.00029606e-01 8.43701477e-04 5.83159551e-02 7.84770906e-01 1.37939084e+00 -5.33878505e-01 -4.12501246e-01 -4.36063021e-01 -2.38611802e-01 3.08769792e-01 -3.47757130e-03 1.21788077e-01 -6.64359570e-01 -1.17077388e-01 -2.65248045e-02 1.14985839e-01 8.68056118e-01 1.17322135e+00 6.11995816e-01 -4.52082813e-01 -5.46939909e-01 1.01032209e+00 2.06384325e+00 5.05664885e-01 2.79526711e-01 5.13883173e-01 2.89636672e-01 5.30195832e-01 4.93311852e-01 3.45640719e-01 3.96234840e-02 -1.29150733e-01 6.64640427e-01 -8.85209024e-01 -2.36713126e-01 -2.54348479e-02 -5.14609993e-01 6.52968466e-01 -4.68523502e-01 -9.36995074e-02 -9.27765608e-01 1.16149211e+00 -1.40894270e+00 -7.17135549e-01 -1.30427957e-01 2.14367223e+00 7.90809572e-01 -1.33589566e-01 1.97880939e-01 2.22501069e-01 1.01298118e+00 -6.49269521e-01 -3.72266322e-01 -5.15301824e-01 5.17690703e-02 4.81863171e-01 1.07955337e+00 4.78800625e-01 -1.07901347e+00 5.61615407e-01 6.18369532e+00 6.04988873e-01 -1.35331762e+00 -5.54256588e-02 7.81117380e-01 -3.96412313e-02 5.32013834e-01 -3.68455946e-01 -5.97600341e-01 2.16420710e-01 -2.10237759e-03 -1.24684766e-01 4.41929430e-01 8.13528180e-01 2.72968821e-02 -5.64364552e-01 -7.90851951e-01 7.80644894e-01 -2.12300330e-01 -1.50015736e+00 -3.62183736e-03 -3.39935012e-02 8.80844533e-01 1.35396749e-01 -2.56804049e-01 1.76907763e-01 2.08788782e-01 -1.09723854e+00 3.76163244e-01 3.70844007e-01 1.07488656e+00 -6.03801012e-01 1.28936732e+00 5.82363605e-01 -9.57046986e-01 -7.57630318e-02 -2.86025017e-01 -1.13695338e-01 -3.03635776e-01 3.49578202e-01 -1.71626389e+00 -2.61759281e-01 7.47536421e-01 -4.55455780e-02 -1.42220467e-01 1.55067933e+00 8.50743651e-02 3.87623370e-01 -5.00158370e-01 -1.69401973e-01 2.40181103e-01 -2.19508186e-01 3.44790071e-01 1.20653427e+00 6.98174953e-01 -1.13663645e-02 -3.29148948e-01 1.09907568e+00 3.78918238e-02 5.84984720e-01 -4.20945436e-01 2.55937159e-01 2.62334913e-01 1.50505030e+00 -1.50226784e+00 -3.28423947e-01 -2.78878599e-01 5.47283709e-01 1.49973720e-01 1.47327945e-01 -2.89812118e-01 -2.07808584e-01 2.52907127e-02 4.95005190e-01 1.14275657e-01 5.25179625e-01 -6.22945666e-01 -4.83186603e-01 -3.84218782e-01 -4.68203008e-01 4.55479920e-01 -2.16194049e-01 -8.36151481e-01 2.06611574e-01 -2.89082557e-01 -1.30541265e+00 -2.43642047e-01 -7.88059235e-01 -7.84162521e-01 8.05859327e-01 -9.74373579e-01 -9.28436816e-01 -4.87787277e-01 1.06279857e-01 1.80672660e-01 5.27307093e-02 9.92709577e-01 3.32003534e-01 -2.08683640e-01 4.42306250e-01 6.38336837e-02 2.99137652e-01 2.79090077e-01 -1.64092326e+00 -1.43293589e-01 5.97267449e-01 -5.47202170e-01 5.81127524e-01 7.18303025e-01 -7.24681854e-01 -3.58514428e-01 -1.26908135e+00 8.13974023e-01 4.58927661e-01 1.29282013e-01 -9.54611301e-02 -5.99328458e-01 6.25825465e-01 6.27181381e-02 3.32079291e-01 6.76764607e-01 -5.70810199e-01 5.08223414e-01 1.15548566e-01 -1.80294847e+00 4.62253869e-01 2.68920004e-01 1.06108144e-01 -3.01915377e-01 3.54791701e-01 1.47166193e-01 -8.26653123e-01 -9.05491412e-01 6.67118192e-01 5.82395792e-01 -1.11362934e+00 5.59411585e-01 3.00067037e-01 2.05842629e-01 -3.91142845e-01 4.12346929e-01 -9.99651968e-01 -3.26928586e-01 -1.16649903e-01 4.73773360e-01 5.56853294e-01 4.05140609e-01 -5.47625482e-01 9.87251043e-01 -1.21928006e-01 -2.75261372e-01 -9.63350177e-01 -8.50438356e-01 -2.32221663e-01 7.32678175e-02 5.03151119e-01 2.87004352e-01 9.27718878e-01 2.51793265e-01 -9.77333188e-02 3.87194782e-01 2.51053870e-01 5.62746704e-01 -4.26409058e-02 5.18875957e-01 -1.20315790e+00 -9.52893570e-02 -4.54298913e-01 -9.16606009e-01 -5.35733819e-01 -6.66336894e-01 -9.17546868e-01 -6.73065260e-02 -1.94851208e+00 8.26324001e-02 -1.80140391e-01 -1.20479323e-01 3.69681746e-01 1.96233705e-01 2.99581885e-01 -3.36440027e-01 9.56092775e-02 -1.90853640e-01 -3.01266402e-01 1.59681678e+00 1.36976108e-01 -3.18611741e-01 -1.56716574e-02 -2.77219117e-01 1.21482408e+00 8.72380435e-01 -5.52126884e-01 -1.02013173e-02 -9.92917493e-02 1.33822456e-01 3.73601168e-01 5.10683596e-01 -1.39577413e+00 3.76555175e-01 1.87964708e-01 7.35701084e-01 -8.17185640e-01 -3.09107572e-01 -5.85438430e-01 1.87329277e-01 1.12524986e+00 -1.18798554e-01 -2.35582352e-01 1.07440203e-01 8.18710476e-02 -3.89488190e-01 -3.18483770e-01 1.19410598e+00 -7.82978535e-01 -3.51433665e-01 -9.32561059e-04 -7.10827887e-01 -3.87288541e-01 1.32249391e+00 -6.67162538e-01 -3.18182081e-01 1.38137996e-01 -9.64128137e-01 1.14747547e-01 7.36621737e-01 -7.64132619e-01 3.87909174e-01 -9.83250678e-01 -7.55976260e-01 4.11289573e-01 1.71617244e-03 7.49408305e-01 3.07612896e-01 1.46257067e+00 -1.79157209e+00 5.15894711e-01 -5.19790828e-01 -6.56560183e-01 -1.30683315e+00 3.95678848e-01 6.55884624e-01 -5.00233591e-01 -7.38484085e-01 1.20831072e+00 5.49180925e-01 -3.92207742e-01 2.14956403e-01 -1.16808176e+00 -4.56900954e-01 -2.70422637e-01 2.94689834e-01 4.61821675e-01 -4.99970838e-02 -4.01707798e-01 5.20434417e-02 5.05307972e-01 6.16592094e-02 3.00005257e-01 9.67007637e-01 -1.52727515e-01 -3.43558729e-01 1.56437799e-01 8.73872101e-01 -2.78286159e-01 -9.01378572e-01 4.96070907e-02 -3.67955208e-01 -2.11745854e-02 3.01650744e-02 -8.88268888e-01 -1.08171618e+00 6.04638100e-01 1.01236844e+00 2.90323049e-01 1.08418989e+00 -2.20193297e-01 7.21458197e-01 -1.23560727e-01 5.61403990e-01 -1.00241065e+00 -6.81494057e-01 2.00401813e-01 6.36723101e-01 -1.13161659e+00 -8.76325592e-02 -4.42750335e-01 -1.64312631e-01 1.18646455e+00 7.73768783e-01 -9.03298974e-01 3.21488529e-01 5.47629654e-01 3.49319100e-01 -2.67677873e-01 -5.26710033e-01 -7.00720847e-02 4.74308282e-02 6.81841552e-01 5.11517227e-01 1.90327570e-01 -7.06438839e-01 5.99593401e-01 -2.67604947e-01 2.90917784e-01 3.33498508e-01 7.35001266e-01 -7.38491476e-01 -2.93473661e-01 -5.02017140e-01 8.53287876e-01 -9.21861887e-01 -1.67727008e-01 -1.62023857e-01 1.15350389e+00 8.70521069e-01 2.76558787e-01 4.12641913e-01 1.64898559e-01 -1.63856432e-01 -4.24715459e-01 6.37716234e-01 -7.49078453e-01 -9.30736065e-01 5.56882322e-01 -1.44785315e-01 3.56003270e-02 -2.34265700e-01 -3.99353981e-01 -1.89874268e+00 -1.40803009e-01 -2.67849207e-01 -1.22040585e-01 4.53216612e-01 7.61436939e-01 -3.44403416e-01 3.53018999e-01 2.78878920e-02 -3.20641816e-01 1.07195251e-01 -1.18079340e+00 -1.20753205e+00 2.53789686e-02 6.40067399e-01 -3.91885787e-01 -5.93660831e-01 3.54262859e-01]
[14.669384956359863, -2.9980196952819824]
94328455-5c2d-4e14-9cb5-714eba08f5b9
compliance-challenges-in-forensic-image
2203.00469
null
https://arxiv.org/abs/2203.00469v1
https://arxiv.org/pdf/2203.00469v1.pdf
Compliance Challenges in Forensic Image Analysis Under the Artificial Intelligence Act
In many applications of forensic image analysis, state-of-the-art results are nowadays achieved with machine learning methods. However, concerns about their reliability and opaqueness raise the question whether such methods can be used in criminal investigations. So far, this question of legal compliance has hardly been discussed, also because legal regulations for machine learning methods were not defined explicitly. To this end, the European Commission recently proposed the artificial intelligence (AI) act, a regulatory framework for the trustworthy use of AI. Under the draft AI act, high-risk AI systems for use in law enforcement are permitted but subject to compliance with mandatory requirements. In this paper, we review why the use of machine learning in forensic image analysis is classified as high-risk. We then summarize the mandatory requirements for high-risk AI systems and discuss these requirements in light of two forensic applications, license plate recognition and deep fake detection. The goal of this paper is to raise awareness of the upcoming legal requirements and to point out avenues for future research.
['Christian Riess', 'Nicole Scheler', 'Benedikt Lorch']
2022-03-01
null
null
null
null
['license-plate-recognition']
['computer-vision']
[ 4.03557479e-01 2.63227820e-01 6.93680942e-02 -3.57100636e-01 -6.35630071e-01 -5.31514227e-01 6.26644433e-01 7.55656511e-02 -6.05723739e-01 5.80541074e-01 -4.14713085e-01 -6.94831312e-01 -1.55317649e-01 -6.67770028e-01 -3.00436139e-01 -6.07274532e-01 3.10930431e-01 3.66100758e-01 1.54214963e-01 1.60598725e-01 6.53682590e-01 1.11168277e+00 -1.35500014e+00 7.32427061e-01 8.13893557e-01 7.16544032e-01 -5.30668497e-01 3.36174697e-01 -2.07754932e-02 9.80674326e-01 -9.04149413e-01 -1.28469169e+00 6.08553350e-01 -4.31454599e-01 -8.94666791e-01 2.00031459e-01 3.94353807e-01 -7.63654113e-01 -3.09322178e-01 1.32036090e+00 1.80313841e-01 -3.28879684e-01 7.64138877e-01 -1.17209482e+00 -7.78379500e-01 2.36881271e-01 -5.78118622e-01 4.66936529e-01 1.52277440e-01 4.71727163e-01 6.99934602e-01 -4.25019056e-01 6.79453790e-01 8.24249566e-01 3.66168827e-01 7.12072790e-01 -7.71136522e-01 -7.03079641e-01 -1.51663974e-01 7.42400885e-01 -1.12801492e+00 -5.29612362e-01 9.61921096e-01 -7.24239886e-01 8.20397019e-01 2.07356587e-01 3.87947887e-01 1.01451898e+00 2.64808297e-01 8.26137543e-01 1.55303466e+00 -6.40539289e-01 4.08107162e-01 4.45072114e-01 4.26156193e-01 5.84385812e-01 9.31547046e-01 1.18821986e-01 -4.26217109e-01 -5.06033242e-01 5.35596013e-01 -4.34947163e-01 1.87038392e-01 -6.09970558e-03 -6.05402708e-01 1.12607396e+00 -1.03882603e-01 8.09160233e-01 -3.83990228e-01 -1.64539322e-01 6.94651067e-01 8.61516148e-02 4.72786248e-01 6.61400557e-01 1.35057032e-01 -1.97125971e-01 -1.12051845e+00 1.58832997e-01 4.29063946e-01 2.52633363e-01 3.76696765e-01 3.19049537e-01 5.16457796e-01 6.51131094e-01 4.83645082e-01 4.21436965e-01 -5.79446740e-02 -1.02732313e+00 2.23001271e-01 3.97851348e-01 -4.95758839e-02 -1.00402093e+00 -1.61753614e-02 1.81390509e-01 -4.43113357e-01 7.93280065e-01 5.65509439e-01 1.53635904e-01 -6.46368623e-01 6.83514237e-01 2.74590566e-03 1.11932740e-01 1.98435318e-02 8.65024567e-01 6.76205933e-01 2.84374982e-01 5.00956774e-02 -1.12648532e-01 1.37811136e+00 -2.09521621e-01 -1.07648981e+00 -2.56127328e-01 4.86630768e-01 -6.25979960e-01 6.10434771e-01 1.05183232e+00 -9.47972059e-01 -1.50654957e-01 -1.14009047e+00 4.81710173e-02 -3.97481471e-01 -6.39237789e-03 7.83896387e-01 1.65077126e+00 -3.21696162e-01 4.13727105e-01 -7.29434431e-01 -1.55182436e-01 8.93791318e-01 3.29019994e-01 -6.17219210e-01 -1.16063654e-02 -1.09491718e+00 1.33281815e+00 1.40138790e-01 4.16281909e-01 -4.19403881e-01 -2.37844095e-01 -8.16143334e-01 -4.67463672e-01 1.82723686e-01 1.04759254e-01 1.00562751e+00 -8.79118383e-01 -1.25545478e+00 1.72682047e+00 2.43921787e-01 -7.49872088e-01 7.31129587e-01 -1.35627434e-01 -8.47825110e-01 7.57694304e-01 8.04717019e-02 1.01123385e-01 8.97656739e-01 -1.26927161e+00 -5.51728845e-01 -4.83702928e-01 7.03717396e-02 -6.67674184e-01 -1.91973820e-01 7.32550085e-01 -5.99894188e-02 -3.98946702e-01 -4.55221474e-01 -8.99640799e-01 7.50192553e-02 1.51097877e-02 -2.17640951e-01 -1.61311164e-01 1.03962243e+00 -8.49065065e-01 1.13139963e+00 -2.27615523e+00 -7.81307936e-01 2.89978564e-01 1.40492812e-01 9.06247973e-01 2.18976468e-01 1.55531377e-01 2.19546142e-03 2.21047267e-01 -5.44219434e-01 -6.55004606e-02 1.58332214e-01 1.64301947e-01 -4.69874233e-01 1.18660867e+00 3.73766840e-01 7.54233420e-01 -7.57064700e-01 -7.02590644e-01 6.79457903e-01 3.32146198e-01 -7.49437883e-02 -1.52515378e-02 3.41034949e-01 1.52260527e-01 -5.35806537e-01 9.86582160e-01 9.53404725e-01 3.04265559e-01 1.96156397e-01 -1.16061911e-01 -5.71885370e-02 1.23501979e-01 -8.83167565e-01 9.03653979e-01 8.81021768e-02 8.82110298e-01 1.67718902e-01 -1.00372314e+00 1.07920182e+00 2.70927638e-01 3.45788956e-01 -7.31273353e-01 1.75514609e-01 6.13336563e-01 2.41302788e-01 -9.46770430e-01 4.82111871e-01 -5.04317164e-01 9.69187543e-02 6.21058047e-01 -2.09062681e-01 -1.95312664e-01 1.12823091e-01 -1.00539513e-01 9.83599722e-01 1.34493366e-01 3.43170315e-01 5.83235174e-02 6.30033076e-01 -1.51461009e-02 4.96582985e-01 5.87135136e-01 -7.58319974e-01 4.32023495e-01 5.62889218e-01 -7.55241394e-01 -1.14431262e+00 -7.77467191e-01 -4.25484151e-01 2.65160859e-01 -1.37144715e-01 2.50354577e-02 -1.17620838e+00 -9.77555811e-01 1.13244364e-02 8.78090799e-01 -6.85152769e-01 -1.68444738e-01 -7.79053152e-01 -7.23468363e-01 1.02564120e+00 2.01347083e-01 6.16781592e-01 -1.25936294e+00 -1.20001483e+00 -7.67695829e-02 3.36553305e-02 -1.43001616e+00 4.60087985e-01 -3.50650787e-01 -7.15228319e-01 -1.46141231e+00 -5.05355418e-01 2.33423710e-02 5.08231282e-01 1.39823696e-02 5.20460248e-01 5.12774885e-01 -4.42129880e-01 4.20259446e-01 -4.25111979e-01 -6.55359626e-01 -8.73807371e-01 -2.80672759e-01 -1.00312538e-01 3.09202999e-01 1.07300639e+00 -2.25256816e-01 -3.08659494e-01 1.45531148e-01 -1.44256794e+00 -6.70870543e-01 5.44908166e-01 5.14745951e-01 1.14625655e-01 -8.39801654e-02 6.39027297e-01 -1.09639180e+00 7.09243178e-01 -1.71580568e-01 -6.22588813e-01 3.59818608e-01 -5.95332503e-01 -3.50416929e-01 2.32901573e-01 -6.72760457e-02 -1.23401403e+00 -2.32210740e-01 -1.88933074e-01 -3.69539738e-01 -6.50647402e-01 2.42820486e-01 -1.26490489e-01 -3.95833313e-01 8.97010088e-01 -9.35371816e-02 8.09254199e-02 -3.11432332e-01 -3.37133110e-02 1.20410359e+00 5.29526234e-01 -6.30666435e-01 9.18625534e-01 9.26662087e-01 4.15423885e-02 -1.33821273e+00 -5.47292769e-01 -3.93511772e-01 -7.49139667e-01 -5.89928567e-01 9.43997443e-01 -2.71876454e-01 -7.39939332e-01 5.45671523e-01 -1.41727650e+00 1.25587791e-01 -1.82958335e-01 4.52340394e-01 -4.21552718e-01 1.23763669e+00 -5.49780488e-01 -1.49421442e+00 -2.87935227e-01 -1.23469102e+00 8.45990121e-01 -1.66079834e-01 -4.78260398e-01 -8.34028840e-01 -1.01708218e-01 1.23597443e+00 3.81582491e-02 4.47826743e-01 5.82562804e-01 -7.90354431e-01 -2.90364772e-01 -7.06966996e-01 -1.71676263e-01 7.48748660e-01 -1.22454083e-02 2.01990142e-01 -1.60934854e+00 1.32638246e-01 3.74657035e-01 -2.70317584e-01 7.17676222e-01 1.29364654e-01 8.52715552e-01 -2.47737151e-02 -6.69062063e-02 2.40885481e-01 1.08766067e+00 6.16982102e-01 1.26111341e+00 8.69861305e-01 2.72976816e-01 1.24139965e+00 8.00628364e-01 3.36729705e-01 -2.52004802e-01 6.78127944e-01 4.22286689e-01 2.27977574e-01 1.49806529e-01 2.90545881e-01 4.64604020e-01 1.99957997e-01 -4.38728780e-01 3.93741280e-02 -1.10642898e+00 3.60176772e-01 -1.61073554e+00 -1.27833939e+00 -5.61784446e-01 2.22543430e+00 1.87725440e-01 4.28826123e-01 1.95931464e-01 4.82842088e-01 8.63641620e-01 -1.10615836e-02 -3.26269567e-01 -8.24291766e-01 -2.35710993e-01 1.41257897e-01 2.46997684e-01 2.60854155e-01 -1.11487389e+00 8.20407927e-01 6.46760845e+00 9.45630431e-01 -9.45888162e-01 3.26564342e-01 6.95569932e-01 2.20496148e-01 -1.11668758e-01 -5.18654585e-02 -4.58764374e-01 5.36049008e-01 1.02838647e+00 1.56993404e-01 -5.30901402e-02 6.71269119e-01 2.27477923e-01 -3.25221121e-01 -6.60897195e-01 8.90921235e-01 3.87808353e-01 -1.48647141e+00 -2.01876894e-01 6.14168823e-01 1.39112055e-01 -3.63170475e-01 2.06207648e-01 -1.54713333e-01 -2.28099927e-01 -9.16692197e-01 8.87638330e-01 1.87727854e-01 7.36979544e-01 -8.84266734e-01 1.03701687e+00 -6.87256753e-02 -4.40611839e-01 -9.91950836e-03 -6.28007591e-01 2.38221511e-02 4.57386315e-01 7.44266093e-01 -6.92463338e-01 4.51898783e-01 6.26851439e-01 3.78302306e-01 -4.69913185e-01 9.72983778e-01 -5.31095445e-01 8.49882424e-01 -1.69744268e-01 4.34201993e-02 4.45981771e-01 -4.02022362e-01 5.40704846e-01 1.52769363e+00 -4.42848653e-02 8.96646231e-02 -5.75007975e-01 1.02894843e+00 3.98162872e-01 -1.30133241e-01 -9.11335409e-01 -3.48190606e-01 1.39656529e-01 9.99980807e-01 -1.03080165e+00 -6.05432615e-02 -5.86763144e-01 9.25670862e-01 -1.42385632e-01 -6.63272385e-03 -6.56315863e-01 -2.63883710e-01 3.74715835e-01 4.71606612e-01 -7.89003447e-02 -6.54711127e-02 -5.42776406e-01 -1.08111739e+00 2.96030372e-01 -1.03699875e+00 5.18640697e-01 -6.39292836e-01 -1.31956935e+00 6.28099024e-01 1.58706442e-01 -1.32555056e+00 -1.63438872e-01 -1.17937338e+00 -4.62852120e-01 5.60273290e-01 -1.50403774e+00 -1.44622684e+00 2.74664223e-01 3.34297836e-01 1.36112392e-01 -6.51422381e-01 7.78239369e-01 1.38596207e-01 -3.69769931e-01 2.82697558e-01 -1.16772108e-01 4.60942417e-01 6.82157397e-01 -8.51876318e-01 2.68686652e-01 8.54953170e-01 1.69466645e-01 6.59493208e-01 7.73313642e-01 -9.22872245e-01 -1.06213641e+00 -4.87454236e-01 9.19772685e-01 -5.55710673e-01 8.50150645e-01 -2.08888084e-01 -1.11153328e+00 5.07307410e-01 1.91079214e-01 4.79238033e-02 1.22717834e+00 -2.81406462e-01 -6.88559234e-01 1.95487842e-01 -1.69406152e+00 2.94131488e-01 2.43923008e-01 -7.56785035e-01 -1.01976919e+00 4.26232755e-01 -2.29617760e-01 2.66258061e-01 -6.53610289e-01 5.04248664e-02 5.99860787e-01 -1.21085632e+00 7.75334418e-01 -6.52784586e-01 3.64640087e-01 -2.37296000e-01 1.63637564e-01 -4.14950103e-01 4.39586677e-02 -5.86376429e-01 6.85930774e-02 1.06596100e+00 1.48077711e-01 -8.12453330e-01 9.79202986e-01 1.01581144e+00 -7.25499839e-02 -2.69377053e-01 -1.44337642e+00 -1.04273462e+00 3.83636624e-01 -8.70418489e-01 3.36071372e-01 1.20634830e+00 3.75886917e-01 -2.56892413e-01 -5.79479516e-01 -4.11267811e-03 9.56305265e-01 -1.82638496e-01 5.61711192e-01 -1.16019654e+00 -4.67279963e-02 -4.17046189e-01 -1.03927338e+00 -4.73469421e-02 4.92795795e-01 -5.03469169e-01 -5.40210843e-01 -1.19737923e+00 1.92641810e-01 1.99711025e-01 6.01850897e-02 3.11236531e-01 7.74205029e-02 5.07594645e-01 5.62207103e-01 4.16894257e-01 -2.02792645e-01 -1.69562191e-01 7.21398294e-01 -2.04599857e-01 3.72407258e-01 -4.25892211e-02 -6.63370252e-01 9.99307752e-01 1.00937831e+00 -5.40475607e-01 8.24884996e-02 -1.44022554e-01 1.80560037e-01 -4.83679771e-01 5.03984869e-01 -8.27063501e-01 9.32589844e-02 -3.16468984e-01 1.14990249e-01 -3.34928155e-01 4.74450469e-01 -1.14388108e+00 1.71639130e-01 5.18034458e-01 6.15735799e-02 -3.39522362e-01 1.84036106e-01 2.04786628e-01 -2.97648609e-01 -1.00946021e+00 1.11400878e+00 -1.27401248e-01 -5.55728555e-01 -1.50201470e-01 -8.71077538e-01 -4.96217906e-01 1.35421538e+00 -9.22858953e-01 -5.26432097e-01 -2.71771163e-01 -4.80737537e-01 -4.20570016e-01 5.33095658e-01 1.17715746e-01 8.38336527e-01 -9.94973779e-01 -8.09243560e-01 -1.15113921e-01 2.93942034e-01 -6.82787657e-01 4.93520379e-01 7.07764983e-01 -9.20697510e-01 4.04424727e-01 -4.38093930e-01 -9.00714472e-02 -1.38839805e+00 8.11162829e-01 2.11216316e-01 -2.09215641e-01 -7.41386533e-01 2.76660949e-01 -1.26684919e-01 -6.96315616e-02 -1.51861101e-01 4.04970169e-01 -2.73353130e-01 -1.42476652e-02 9.79574263e-01 7.09565222e-01 1.44743264e-01 -9.72580969e-01 -3.81734222e-01 4.47574615e-01 -1.66030347e-01 -2.59671092e-01 1.44875538e+00 3.52840424e-01 -2.80419320e-01 1.72501668e-01 8.21410120e-01 1.66737929e-01 -6.95140839e-01 3.51716816e-01 3.40933353e-01 -8.23002577e-01 -1.06906421e-01 -7.62103677e-01 -1.16138434e+00 1.26728380e+00 3.86736512e-01 3.50315511e-01 8.40106428e-01 -1.98606730e-01 5.05722582e-01 5.33414245e-01 5.45170665e-01 -1.49695289e+00 -2.28455141e-01 7.08143637e-02 8.59785795e-01 -1.07457745e+00 3.80743176e-01 -6.29975855e-01 -8.74347806e-01 1.52747333e+00 2.22951800e-01 -7.11173564e-02 3.47905517e-01 3.66136074e-01 2.66205907e-01 -4.35012221e-01 9.14014056e-02 1.08236313e-01 1.88696131e-01 1.16744816e+00 4.13911015e-01 -4.94244620e-02 -7.00786293e-01 4.27342772e-01 8.23215172e-02 2.52088964e-01 7.57934034e-01 1.13701558e+00 -4.77732062e-01 -1.44534767e+00 -1.09405494e+00 4.74861354e-01 -9.89054024e-01 2.47764379e-01 -1.05625737e+00 9.25651968e-01 3.15465927e-01 1.05818963e+00 -1.98056951e-01 -1.71859771e-01 2.48433426e-01 1.45079177e-02 4.45328355e-01 -3.94239783e-01 -8.50925863e-01 -1.07038394e-01 4.52652752e-01 -2.30138466e-01 -5.63668489e-01 -9.93241549e-01 -9.10977960e-01 -4.56034452e-01 -3.69455934e-01 2.22045407e-01 7.60714173e-01 1.10833538e+00 -6.92298412e-02 -9.62031037e-02 -1.89250137e-03 -4.07569885e-01 -5.11685014e-01 -6.76581502e-01 -8.17031145e-01 3.50623727e-01 -1.27304509e-01 -6.03244245e-01 -4.80711818e-01 3.46305639e-01]
[12.510102272033691, 1.0451914072036743]
82a0e584-3373-4a14-9d19-aeb6c6b71a1e
sentiment-after-translation-a-case-study-on
null
null
https://aclanthology.info/papers/N15-1078/n15-1078
https://www.aclweb.org/anthology/N15-1078
Sentiment after Translation: A Case-Study on Arabic Social Media Posts
null
['Saif Mohammad', 'Svetlana Kiritchenko', 'Mohammad Salameh']
2015-05-01
null
null
null
hlt-2015-5
['arabic-sentiment-analysis']
['natural-language-processing']
[-2.44508207e-01 3.89024585e-01 -2.65282035e-01 -2.15905145e-01 -8.60921741e-02 -7.76765764e-01 4.48510379e-01 -7.23253429e-01 -5.48377395e-01 1.31954515e+00 3.66348401e-02 -9.49533224e-01 -2.40340635e-01 -1.05564880e+00 -8.44053447e-01 -8.75781775e-01 -7.42435038e-01 6.86515033e-01 1.44298598e-01 -6.52004302e-01 8.47113907e-01 6.29996777e-01 -1.62287033e+00 5.77558100e-01 6.81926727e-01 5.49597681e-01 5.66466339e-02 1.04480565e+00 6.13576770e-02 1.59847176e+00 -6.05450153e-01 -4.56729174e-01 1.43710867e-01 -1.88022718e-01 -5.35770595e-01 -3.36825520e-01 7.91536197e-02 -5.88986814e-01 -3.10633808e-01 6.07356608e-01 1.14270830e+00 -5.53557873e-02 1.07025254e+00 -1.51067197e+00 -7.97060013e-01 5.44234335e-01 2.30399013e-01 1.20632313e-01 6.10446513e-01 -1.79472014e-01 3.51212233e-01 -1.36621380e+00 6.32680655e-01 8.64870071e-01 9.47046995e-01 5.99273622e-01 -1.20647264e+00 -5.60552299e-01 -7.17137158e-01 -2.57835120e-01 -1.59567785e+00 -6.30042374e-01 6.23617955e-02 -3.34735900e-01 1.69142139e+00 6.72587395e-01 1.35647905e+00 1.47341585e+00 1.36649036e+00 4.72517729e-01 1.13904178e+00 -3.16340059e-01 3.65351558e-01 6.43754780e-01 -2.29148138e-02 5.51889002e-01 1.19686353e+00 6.76312149e-01 -5.60917914e-01 -7.92779744e-01 1.11365557e+00 -3.12949389e-01 2.60771513e-01 -7.06989706e-01 -1.06566191e+00 7.88466871e-01 -1.33808568e-01 2.20882341e-01 -3.46203089e-01 -9.92989019e-02 2.24548295e-01 4.12961632e-01 -2.94082850e-01 4.77862865e-01 -8.89795244e-01 -1.87686309e-01 -1.01472771e+00 1.91863477e-01 1.30615628e+00 1.45811689e+00 4.72722203e-03 3.70100170e-01 1.66493103e-01 2.89523482e-01 5.29429734e-01 1.05927479e+00 4.36951250e-01 -1.33365822e+00 -9.55119133e-02 4.20878716e-02 6.12505674e-01 -1.13473701e+00 -7.12930143e-01 -2.41300568e-01 -9.54724610e-01 5.18771529e-01 -1.59270123e-01 -4.59180593e-01 -7.59686172e-01 4.79171664e-01 -1.81984559e-01 -3.19489211e-01 5.45619786e-01 2.69257158e-01 7.28577793e-01 1.45816103e-01 9.25893057e-03 -4.73371416e-01 8.38296890e-01 -1.26899445e+00 -1.11426139e+00 2.33304992e-01 6.82289064e-01 -1.11704338e+00 4.06794518e-01 4.29345578e-01 -1.87200487e+00 -9.69799384e-02 -1.08317137e+00 1.19022481e-01 -7.97166348e-01 -2.57194549e-01 6.66624963e-01 1.49922979e+00 -1.63972652e+00 6.47176981e-01 -6.05021298e-01 5.40335439e-02 -4.73218322e-01 9.07614470e-01 -3.47142309e-01 2.90655226e-01 -1.28751957e+00 9.52449858e-01 -1.04219764e-01 3.03568155e-01 -3.30324143e-01 -1.84501112e-01 -1.03385210e+00 -6.24127567e-01 -2.59417385e-01 -1.07312346e+00 1.36302292e+00 1.09815992e-01 -1.30379498e+00 1.00962746e+00 -4.14315462e-01 -1.14109404e-01 6.61359608e-01 -1.52096570e-01 -8.18174660e-01 1.58321753e-01 2.87593510e-02 3.47163439e-01 5.30214727e-01 -9.69232559e-01 -6.31655395e-01 -1.00153096e-01 -2.54188716e-01 2.48816490e-01 1.43904146e-03 2.05205917e-01 6.38460100e-01 -9.53604504e-02 1.91062525e-01 -8.61961663e-01 -3.21419090e-01 -7.61404037e-01 -1.77353188e-01 -3.84736449e-01 5.59762537e-01 -2.92163283e-01 1.33848476e+00 -1.84365213e+00 -4.54683900e-01 2.94456817e-03 2.53527164e-01 3.89016122e-02 2.33101144e-01 1.23831403e+00 -2.43849114e-01 1.06667292e+00 3.98064762e-01 9.52541176e-03 3.13382268e-01 5.07273376e-01 2.86674555e-02 2.51206756e-01 9.43330750e-02 1.15280068e+00 -1.09214199e+00 -5.20940363e-01 5.57267785e-01 4.60968643e-01 -3.84588838e-01 6.24593735e-01 9.30012882e-01 2.99172718e-02 -8.94021094e-02 1.33241999e+00 1.05377018e+00 -2.90812284e-01 -2.46628653e-02 3.44544262e-01 -6.44541264e-01 6.00222833e-02 -7.41804183e-01 9.49429095e-01 1.31349964e-02 1.99757457e-01 1.74998537e-01 -6.80728197e-01 3.94654661e-01 1.10545933e+00 4.45659161e-01 -9.24297392e-01 -1.69569388e-01 9.00310338e-01 1.03920102e-01 -9.62983370e-01 6.40755475e-01 -1.72503099e-01 7.69825354e-02 1.40852645e-01 -3.61913115e-01 -3.84120375e-01 5.52458428e-02 4.19574350e-01 1.00224769e+00 -1.33707494e-01 5.48685491e-01 -8.72328460e-01 6.88946426e-01 -2.01589778e-01 3.00597936e-01 1.23978651e+00 -2.90862411e-01 9.21478033e-01 3.19022417e-01 -5.37574828e-01 -5.72780192e-01 -1.20098352e+00 -4.80556756e-01 9.86326516e-01 -1.58388391e-01 -3.87671381e-01 -7.10155964e-01 -2.63155133e-01 -1.85071185e-01 4.93483543e-01 -5.93591571e-01 5.23159266e-01 -5.14705896e-01 -6.85349941e-01 7.65277922e-01 3.17972392e-01 6.49624616e-02 -1.49228525e+00 -5.70813596e-01 3.34754556e-01 -1.07132711e-01 -5.39811492e-01 -1.57590687e-01 6.12205923e-01 -1.19366693e+00 -1.01526415e+00 -8.75530541e-01 -1.03603637e+00 7.18889177e-01 2.44572356e-01 1.18463767e+00 2.84042627e-01 -2.87709564e-01 6.49904788e-01 -1.31030381e-02 -2.67939389e-01 -3.16364132e-02 -5.78750819e-02 5.01899779e-01 -1.07851994e+00 6.46393359e-01 -3.92214775e-01 -1.00622165e+00 5.35368681e-01 -3.85821253e-01 -2.04611585e-01 7.85597622e-01 1.04281425e+00 1.42742008e-01 -1.33182362e-01 2.41971418e-01 -1.16735983e+00 1.11498392e+00 -4.85333532e-01 1.30341025e-02 2.62960047e-01 -8.53362501e-01 -4.53584343e-01 2.27881029e-01 -1.33037493e-02 -6.33760452e-01 -2.51423597e-01 -4.14545417e-01 3.22794706e-01 -4.06866640e-01 3.34393084e-02 2.24873364e-01 -5.26000679e-01 5.21005154e-01 1.50258854e-01 -4.07556817e-02 1.21735156e-01 6.30975887e-02 3.75316888e-01 3.67786407e-01 -7.01224983e-01 5.19809723e-01 1.93325520e-01 -3.00995037e-02 -1.28791058e+00 2.66786069e-01 -1.64763972e-01 -6.60750508e-01 -2.69126564e-01 7.75121748e-01 -9.50198710e-01 -6.35705650e-01 5.21708906e-01 -9.54387486e-01 -1.36340424e-01 -5.25335252e-01 6.30965471e-01 -7.02474535e-01 2.94757709e-02 -4.17852998e-01 -1.17976546e+00 -4.78847146e-01 -1.38869667e+00 9.59160328e-01 3.85633826e-01 -2.05811903e-01 -1.19790471e+00 3.38975668e-01 8.99848714e-02 4.29175526e-01 7.39965662e-02 5.03531873e-01 -2.80004859e-01 -2.98304528e-01 -2.85470933e-01 1.59573574e-02 -1.65876746e-01 -3.09365243e-02 4.80770379e-01 -5.48391879e-01 -4.37962860e-01 2.44738594e-01 -4.86500829e-01 -7.51581416e-02 7.11018145e-01 4.09077644e-01 -6.02920949e-01 -7.12544680e-01 5.56552589e-01 1.27709007e+00 4.72497940e-01 5.88340998e-01 6.77552521e-01 9.28869769e-02 5.17871320e-01 4.48484391e-01 2.08113909e-01 1.07506551e-01 1.08483166e-01 1.14217520e-01 -2.17137560e-01 9.45900455e-02 -1.75644696e-01 4.44304377e-01 1.09980536e+00 -8.23837519e-01 -2.15223029e-01 -4.88283783e-01 6.68056428e-01 -1.69156313e+00 -1.31570101e+00 -6.70152605e-01 9.76623476e-01 3.78074080e-01 3.38884830e-01 1.58041969e-01 1.21232189e-01 5.47049284e-01 -4.76693302e-01 -1.27305150e-01 -9.16352808e-01 -3.82233024e-01 4.21136379e-01 7.14838982e-01 6.51479006e-01 -6.10602558e-01 5.08361042e-01 1.25505590e+01 5.49717486e-01 -5.10860123e-02 1.27314463e-01 3.14815581e-01 1.99914098e-01 -2.24020526e-01 1.62632689e-01 -1.04165065e+00 9.26906094e-02 1.64754760e+00 -4.36735392e-01 3.93212438e-01 6.44444764e-01 3.94769877e-01 -1.96429595e-01 -9.78620768e-01 5.21999180e-01 1.12709433e-01 -1.50397491e+00 -3.76861334e-01 9.64888573e-01 5.80093563e-01 -1.81895152e-01 6.81941509e-01 4.33209062e-01 7.39425182e-01 -1.14519978e+00 7.88427055e-01 3.18626672e-01 8.87266219e-01 -8.18826377e-01 1.07315242e+00 2.76328743e-01 -1.04355454e+00 1.15487054e-02 -5.90443075e-01 -9.22581375e-01 1.62244976e-01 2.02188566e-01 -5.61583102e-01 4.06869203e-01 6.19966745e-01 2.15250537e-01 -8.53338003e-01 1.45581341e+00 -4.29231405e-01 4.22687940e-02 -3.00976187e-01 -4.39922899e-01 4.83290881e-01 5.28384298e-02 3.26606274e-01 1.02295005e+00 2.89785862e-01 5.45581102e-01 -2.62208551e-01 2.92286038e-01 6.63503349e-01 5.84383309e-03 -1.38496268e+00 -8.68683681e-02 3.54620427e-01 9.69217539e-01 -3.99494410e-01 -3.56748998e-02 -6.11456513e-01 6.62115753e-01 -1.66942239e-01 5.75146377e-01 -2.92761147e-01 -8.89406264e-01 4.33969021e-01 -2.30117049e-02 -2.88469017e-01 -1.35338321e-01 -3.23372036e-01 -1.02459288e+00 -6.06416404e-01 -3.90122980e-01 7.32179061e-02 -8.63867998e-01 -1.79313409e+00 6.85844719e-01 -3.50327820e-01 -1.03178740e+00 -9.29034173e-01 -9.72984850e-01 -2.90242791e-01 6.67755604e-01 -6.59234166e-01 -1.30476952e+00 1.77773401e-01 4.65253919e-01 1.82401046e-01 -5.88519633e-01 1.09945750e+00 -1.42670095e-01 -3.18711132e-01 1.08882773e+00 3.51253241e-01 -3.05974871e-01 8.32965672e-01 -1.16271269e+00 4.23785180e-01 1.09189451e-02 -6.59259140e-01 1.28460062e+00 6.81860268e-01 -6.81785107e-01 -1.16639841e+00 -1.52893141e-01 1.38110602e+00 -9.15031374e-01 5.27874053e-01 -1.02810413e-01 1.46300867e-01 6.57190859e-01 9.07351732e-01 -5.09461224e-01 1.08935249e+00 -1.13616258e-01 4.93123859e-01 8.23448122e-01 -1.38805819e+00 5.99272728e-01 1.21222687e+00 -4.81854081e-01 -8.27020168e-01 3.28937650e-01 9.94227231e-01 -5.60786903e-01 -1.34066856e+00 6.99696779e-01 1.14132881e+00 -1.04883564e+00 1.54237401e+00 -1.16912270e+00 4.08816934e-01 1.70770600e-01 -2.18455836e-01 -6.52935803e-01 -5.24122715e-01 -8.07741940e-01 -3.97186369e-01 3.40213925e-01 8.64144087e-01 -1.16800272e+00 5.53630650e-01 1.39166510e+00 -3.58625412e-01 -4.51715857e-01 -8.63314152e-01 -1.06149745e+00 -4.66870656e-03 1.81816339e-01 3.85523647e-01 8.18638563e-01 7.09820271e-01 2.02331603e-01 -2.05201268e-01 -4.37674336e-02 6.59316301e-01 -4.87512946e-01 5.82342803e-01 -8.74011815e-01 2.49149442e-01 -3.85976404e-01 -2.94489831e-01 -2.90531307e-01 -3.26795012e-01 -4.87107635e-01 -8.59293997e-01 -1.53084815e+00 8.49664882e-02 1.97367743e-01 -2.33256146e-01 3.44385266e-01 6.77823246e-01 2.45789096e-01 -2.27382466e-01 1.17054820e-01 -1.86927155e-01 -9.06788632e-02 9.18564737e-01 1.10557920e-03 -2.17724726e-01 3.92412305e-01 -1.01716614e+00 9.60534811e-01 7.85404518e-02 -6.09580100e-01 -5.51487744e-01 1.64939865e-01 5.71885824e-01 3.57775450e-01 4.49405432e-01 -9.70509052e-01 9.87445056e-01 -4.32846308e-01 6.12257063e-01 -1.43385470e+00 -2.65485227e-01 -9.00021493e-01 3.61826897e-01 6.29400074e-01 4.08042461e-01 3.53796333e-01 2.26811334e-01 -9.13986787e-02 -2.01378226e-01 -6.44658387e-01 4.43001747e-01 -7.45609999e-01 -4.55086857e-01 -3.90635371e-01 -1.02313221e+00 -9.67486948e-03 7.88939476e-01 -4.61360693e-01 -6.47767961e-01 -1.95199717e-02 -1.44608676e+00 7.30553791e-02 8.33678424e-01 1.34732248e-02 6.98022604e-01 -1.44444084e+00 -1.11748077e-01 6.18802965e-01 -5.42432606e-01 -4.98194307e-01 2.20674574e-01 1.06156552e+00 -1.31528914e+00 1.26915324e+00 -5.21896303e-01 -2.05424115e-01 -8.96585643e-01 7.19095051e-01 3.15837234e-01 1.75098896e-01 -3.93049836e-01 7.00151443e-01 1.12239346e-02 -6.58594191e-01 1.24011397e-01 2.97559589e-01 -4.87763733e-01 -2.52128989e-01 9.22277749e-01 1.17918742e+00 1.32526740e-01 -3.91715407e-01 -5.58191121e-01 4.39446121e-01 2.08604142e-01 -6.12734079e-01 1.38662541e+00 -2.20989168e-01 -7.44191706e-01 5.00318348e-01 7.50003338e-01 -7.33495504e-02 -3.28966863e-02 1.27011871e+00 4.80256975e-03 -5.47239244e-01 -3.11336160e-01 -7.82959402e-01 -3.42039734e-01 5.93214929e-01 6.17733955e-01 4.59080935e-01 9.69997942e-01 -6.47576571e-01 7.70057857e-01 7.64313757e-01 7.83133745e-01 -1.22997761e+00 -6.36951804e-01 3.69483382e-01 9.39836800e-01 -9.30032432e-01 1.30400375e-01 -3.14221263e-01 -5.55534840e-01 1.08689797e+00 5.38173854e-01 -1.62142932e-01 1.17091513e+00 6.96656525e-01 -1.75611794e-01 -4.53250527e-01 -1.04549897e+00 9.83817875e-02 2.10727617e-01 1.32608259e+00 8.26061428e-01 1.41275764e-01 -1.40426433e+00 8.46834600e-01 -6.34404242e-01 4.76996809e-01 9.47268784e-01 1.55018985e+00 -5.12773991e-01 -1.41692734e+00 -4.41162825e-01 2.31142953e-01 -2.82854587e-01 -3.49930972e-01 -8.62966061e-01 1.09929097e+00 1.97083559e-02 1.37330377e+00 -4.48758155e-01 -4.63731140e-01 8.27405751e-01 3.19600284e-01 4.41484600e-01 -1.25142381e-01 -8.94688129e-01 6.48750722e-01 3.84649158e-01 -1.05672014e+00 -7.94483483e-01 -1.27049911e+00 -1.08188879e+00 -1.16740942e+00 -3.07217389e-01 4.65777248e-01 2.31674016e-01 3.19533348e-01 -2.09303293e-02 1.56313553e-03 6.67447627e-01 -1.00111675e+00 -1.97397545e-01 -6.07395887e-01 -1.20775104e+00 -5.56946039e-01 3.94425273e-01 -5.85869551e-01 -1.15223050e+00 -6.24734610e-02]
[-1.539193034172058, 15.869195938110352]
1434d894-7425-43cb-8197-36edead25ee5
full-resolution-residual-networks-for
1611.08323
null
http://arxiv.org/abs/1611.08323v2
http://arxiv.org/pdf/1611.08323v2.pdf
Full-Resolution Residual Networks for Semantic Segmentation in Street Scenes
Semantic image segmentation is an essential component of modern autonomous driving systems, as an accurate understanding of the surrounding scene is crucial to navigation and action planning. Current state-of-the-art approaches in semantic image segmentation rely on pre-trained networks that were initially developed for classifying images as a whole. While these networks exhibit outstanding recognition performance (i.e., what is visible?), they lack localization accuracy (i.e., where precisely is something located?). Therefore, additional processing steps have to be performed in order to obtain pixel-accurate segmentation masks at the full image resolution. To alleviate this problem we propose a novel ResNet-like architecture that exhibits strong localization and recognition performance. We combine multi-scale context with pixel-level accuracy by using two processing streams within our network: One stream carries information at the full image resolution, enabling precise adherence to segment boundaries. The other stream undergoes a sequence of pooling operations to obtain robust features for recognition. The two streams are coupled at the full image resolution using residuals. Without additional processing steps and without pre-training, our approach achieves an intersection-over-union score of 71.8% on the Cityscapes dataset.
['Alexander Hermans', 'Tobias Pohlen', 'Bastian Leibe', 'Markus Mathias']
2016-11-24
full-resolution-residual-networks-for-1
http://openaccess.thecvf.com/content_cvpr_2017/html/Pohlen_Full-Resolution_Residual_Networks_CVPR_2017_paper.html
http://openaccess.thecvf.com/content_cvpr_2017/papers/Pohlen_Full-Resolution_Residual_Networks_CVPR_2017_paper.pdf
cvpr-2017-7
['thermal-image-segmentation']
['computer-vision']
[ 7.05902100e-01 -3.16667333e-02 -2.21892167e-02 -6.28984749e-01 -7.42126226e-01 -5.05069137e-01 5.28828502e-01 1.47363096e-01 -8.93574715e-01 4.42220271e-01 -2.95658410e-01 -2.61600733e-01 1.48499921e-01 -1.02564013e+00 -8.63965452e-01 -6.71955943e-01 2.68120229e-01 2.64459819e-01 9.05905604e-01 -3.81319076e-01 3.06603789e-01 6.54742718e-01 -1.94079971e+00 2.24359602e-01 1.01710594e+00 1.25392830e+00 4.39583540e-01 6.95612311e-01 -3.15239102e-01 5.60822785e-01 -4.58996534e-01 -2.72514727e-02 3.66541237e-01 -1.88344106e-01 -8.45676899e-01 2.69071311e-01 6.16651595e-01 -2.58787870e-01 -1.98847339e-01 1.22388935e+00 5.79653531e-02 2.88703620e-01 4.36869353e-01 -8.97018373e-01 -2.93210745e-01 1.65297702e-01 -4.37049985e-01 3.66483629e-01 9.51751471e-02 2.02167451e-01 9.35827255e-01 -6.32777989e-01 4.97459948e-01 7.37421453e-01 5.64063370e-01 2.20952198e-01 -1.25258398e+00 -4.92580444e-01 3.14850569e-01 2.98314512e-01 -1.46418703e+00 -3.93616021e-01 5.91991425e-01 -2.73230344e-01 9.45323288e-01 1.72713563e-01 5.88386893e-01 5.30726016e-01 2.77577322e-02 5.69733918e-01 1.05388248e+00 -2.72457451e-01 2.91533947e-01 3.94597985e-02 8.66086781e-02 5.32355189e-01 2.58371413e-01 -1.22440979e-01 -3.12840998e-01 6.32713437e-01 8.88064921e-01 1.62246212e-01 -8.31934810e-02 -3.11010689e-01 -9.76178825e-01 7.18245268e-01 8.35595608e-01 4.95410591e-01 -5.65334201e-01 1.61389992e-01 1.71275400e-02 -4.40857336e-02 1.81467295e-01 4.00747210e-01 -2.41662830e-01 1.04536317e-01 -1.25785863e+00 4.67993543e-02 3.19826126e-01 7.10557401e-01 1.20291603e+00 -7.23881051e-02 1.27750844e-01 8.20009291e-01 -9.90415365e-02 4.04927552e-01 4.22968745e-01 -1.16340172e+00 4.28508043e-01 8.42652500e-01 6.44708648e-02 -9.30105507e-01 -6.81075335e-01 -5.82821071e-01 -7.92551756e-01 5.29295981e-01 6.79780424e-01 1.26593813e-01 -1.31097853e+00 1.41472864e+00 2.88301140e-01 1.47426933e-01 1.65633231e-01 1.01164210e+00 6.38041973e-01 6.08699262e-01 1.10775471e-01 4.28065121e-01 1.57588947e+00 -9.50804055e-01 -3.13013852e-01 -8.34243298e-01 5.54890871e-01 -5.51917970e-01 9.13213193e-01 1.93043217e-01 -9.31674838e-01 -7.25894570e-01 -1.33706820e+00 -3.13340485e-01 -7.95571864e-01 -4.50829752e-02 4.70902890e-01 6.46098673e-01 -1.27292776e+00 3.61442208e-01 -6.85848773e-01 -3.67576957e-01 5.31450510e-01 4.54272330e-01 -6.32015824e-01 -1.55248865e-01 -1.08563232e+00 8.73108387e-01 5.53180218e-01 2.27057651e-01 -4.43397403e-01 -3.20229590e-01 -1.04740071e+00 9.53173861e-02 4.88841951e-01 -4.98175770e-01 1.09452498e+00 -1.02539325e+00 -1.18894935e+00 8.15240264e-01 -3.85264635e-01 -7.55104780e-01 5.76866388e-01 -1.05640598e-01 -2.91475743e-01 5.13415873e-01 3.40680808e-01 1.04144442e+00 7.95428276e-01 -1.34143138e+00 -1.38978767e+00 -3.73767465e-01 2.59022981e-01 3.56606752e-01 1.20377541e-02 -3.70959640e-01 -8.59524310e-01 -2.82861739e-01 5.94389498e-01 -8.18829358e-01 -4.91363078e-01 -1.69213325e-01 -1.87925711e-01 2.40002312e-02 8.28925014e-01 -7.04442561e-01 9.01200652e-01 -2.06022835e+00 -2.45299533e-01 4.04204816e-01 1.20610967e-01 3.58697981e-01 -1.09821938e-01 -1.01480018e-02 2.92280484e-02 2.80134287e-02 -5.15944123e-01 -1.68475285e-01 -2.56660700e-01 1.96302637e-01 -1.13086052e-01 3.81873906e-01 3.66166770e-01 9.95436788e-01 -7.42973626e-01 -4.64641303e-01 6.96978867e-01 4.89177525e-01 -4.88162190e-01 -8.43100771e-02 -1.87676862e-01 4.44157720e-01 -3.97591323e-01 5.21868169e-01 5.64475358e-01 -2.34959558e-01 -2.07707927e-01 -7.08593652e-02 -4.23585057e-01 1.06571577e-01 -1.24354112e+00 1.67948461e+00 -5.21304607e-01 7.31006324e-01 2.06903040e-01 -1.28430307e+00 9.54770744e-01 2.80197430e-03 4.69924539e-01 -1.31487048e+00 1.97461620e-01 3.93663853e-01 -9.99296531e-02 -3.09664130e-01 8.73038709e-01 4.94547412e-02 -1.79179445e-01 1.66352838e-01 -2.84050524e-01 -2.57873625e-01 3.81885976e-01 -1.00966841e-02 1.03260446e+00 -6.20926060e-02 2.18356792e-02 -2.15772185e-02 8.08482051e-01 4.21702951e-01 3.87285709e-01 9.34760809e-01 -2.66260684e-01 1.03200710e+00 2.41572544e-01 -5.85397482e-01 -1.05485690e+00 -7.28405178e-01 -1.01636328e-01 9.10898149e-01 6.15334928e-01 1.43344134e-01 -1.15760469e+00 -5.56250870e-01 -1.62352085e-01 6.66848183e-01 -6.93253875e-01 1.27290590e-02 -7.07896650e-01 -4.49089587e-01 3.17169189e-01 7.03364193e-01 1.05991185e+00 -1.08735013e+00 -1.09636509e+00 3.80376786e-01 -3.53797942e-01 -1.54363167e+00 -1.73787117e-01 3.85491908e-01 -6.47563994e-01 -1.19907558e+00 -6.11337364e-01 -8.53917599e-01 7.79621720e-01 5.88151872e-01 8.93170714e-01 1.75448820e-01 -3.41212153e-01 4.01562005e-02 -2.20174387e-01 -8.94557834e-02 -1.22852378e-01 2.78274059e-01 -4.34570551e-01 1.06974289e-01 3.06095392e-01 -3.56011152e-01 -8.54677260e-01 5.85815370e-01 -9.92158830e-01 2.13636562e-01 8.43434215e-01 6.20521009e-01 8.85325670e-01 2.99614936e-01 3.01167220e-01 -7.34213829e-01 1.31477237e-01 -2.71538664e-02 -7.98083067e-01 1.92642920e-02 -3.14943433e-01 9.20401327e-03 6.54424310e-01 4.74030338e-03 -8.31311166e-01 4.30411696e-01 -5.63391864e-01 -4.76448648e-02 -6.21298611e-01 2.33755678e-01 -1.01193853e-01 -1.19247429e-01 6.11045361e-01 4.27924424e-01 -6.57048598e-02 -1.07925579e-01 4.15183008e-01 6.51631355e-01 9.23151493e-01 -5.32993563e-02 6.74481928e-01 8.01080108e-01 -1.27111897e-01 -1.00055897e+00 -9.06029940e-01 -7.98110902e-01 -9.22216237e-01 -2.16042414e-01 1.25402403e+00 -8.54469001e-01 -4.47850436e-01 5.43067753e-01 -8.52545023e-01 -4.91011739e-01 -2.78482437e-01 2.57147163e-01 -5.38685083e-01 -1.64550133e-02 -3.80586535e-02 -5.14552534e-01 -1.42763451e-01 -1.27659011e+00 1.14596415e+00 4.98804361e-01 -4.39957231e-02 -6.67014718e-01 -4.84416395e-01 7.14321971e-01 5.07928967e-01 1.98663041e-01 4.57982570e-01 -4.99041349e-01 -8.60704660e-01 -3.88866812e-01 -6.95290446e-01 3.74916226e-01 -6.85269991e-03 -3.60040128e-01 -1.01916242e+00 2.71768928e-01 -1.65619716e-01 8.40297863e-02 1.28806508e+00 5.45335293e-01 1.17770290e+00 1.98157653e-01 -2.56801128e-01 6.32356763e-01 1.37323654e+00 3.55448991e-01 7.70406663e-01 7.38446653e-01 6.83577418e-01 6.47609890e-01 5.41868865e-01 -1.10644266e-01 6.30834401e-01 5.51810861e-01 5.50554693e-01 -4.74291414e-01 -3.96906435e-01 -3.53510901e-02 -2.67133005e-02 6.31946772e-02 -1.29551115e-02 -6.88009113e-02 -9.35028553e-01 7.33482540e-01 -1.81473505e+00 -8.37838054e-01 -2.15296388e-01 2.08534980e+00 4.12855953e-01 5.11273146e-01 -4.78823893e-02 2.83868879e-01 6.37236536e-01 2.71329105e-01 -8.04722726e-01 -3.19098592e-01 -1.86103120e-01 2.36836657e-01 1.03769922e+00 5.44498146e-01 -1.36208403e+00 1.21681237e+00 5.75767422e+00 8.54454815e-01 -1.35061324e+00 -4.30888459e-02 7.74879277e-01 1.93445191e-01 -4.21135081e-03 -1.85149491e-01 -8.03223670e-01 2.12323546e-01 7.08022356e-01 5.40161133e-01 3.21016818e-01 7.55068779e-01 2.28649810e-01 -6.57489300e-01 -5.55295348e-01 9.10847127e-01 4.22254130e-02 -1.43198717e+00 -3.03298891e-01 -4.24428582e-02 7.77488112e-01 2.93323815e-01 2.51755808e-02 3.75293791e-02 2.16210991e-01 -1.16176414e+00 9.38339055e-01 3.33625138e-01 7.26620555e-01 -7.23490834e-01 7.81923175e-01 4.33875740e-01 -1.38735950e+00 -1.92150384e-01 -1.87261119e-01 -3.49503458e-02 2.95076340e-01 5.56534111e-01 -6.28610313e-01 4.00327116e-01 8.13778937e-01 5.83906353e-01 -6.61297441e-01 9.45593238e-01 -4.96380746e-01 2.33701080e-01 -4.54951078e-01 1.90452635e-01 6.36038303e-01 -2.63277173e-01 1.52413458e-01 1.16274357e+00 2.38563538e-01 4.82775755e-02 4.31995094e-02 7.85655558e-01 -4.91419947e-03 -1.50410131e-01 -3.48877639e-01 3.28849763e-01 2.26969928e-01 1.32459140e+00 -1.35415852e+00 -4.88547683e-01 -3.87324154e-01 9.91908431e-01 2.70585567e-01 5.46540558e-01 -6.76631987e-01 -5.42254925e-01 7.78797626e-01 2.31217027e-01 6.56257749e-01 -4.57266659e-01 -7.49821126e-01 -7.11386919e-01 5.55682294e-02 -4.92341965e-01 1.88511506e-01 -5.13341606e-01 -4.87983406e-01 6.20704591e-01 -3.58093590e-01 -9.66419578e-01 -1.38195276e-01 -5.81786633e-01 -4.61971998e-01 8.02866817e-01 -2.00575590e+00 -1.09839320e+00 -6.77264512e-01 4.61071610e-01 7.15514064e-01 3.26175421e-01 5.98252654e-01 3.13834786e-01 -6.29129887e-01 1.82990953e-01 -3.42569649e-02 2.88971245e-01 3.49093705e-01 -1.20413220e+00 6.37956917e-01 1.13505185e+00 1.77544326e-01 3.11946332e-01 5.12208402e-01 -4.36335862e-01 -9.36544716e-01 -1.26156855e+00 7.49236643e-01 -2.51264483e-01 3.05764616e-01 -3.05909902e-01 -9.48714912e-01 3.32072586e-01 -1.99512810e-01 1.39773488e-01 1.02001004e-01 -2.05836490e-01 -2.15721473e-01 -2.00468481e-01 -1.02788579e+00 5.64394832e-01 9.10959959e-01 -4.92736429e-01 -5.08211732e-01 -9.15495381e-02 4.50518787e-01 -4.80249137e-01 -4.58888322e-01 4.32607323e-01 5.52512288e-01 -1.34755492e+00 1.10792863e+00 1.16085179e-01 3.28042507e-01 -5.98685384e-01 -2.45766882e-02 -9.23206925e-01 -1.40369236e-01 -6.45358488e-02 5.26464045e-01 7.50330150e-01 5.63087225e-01 -9.13971722e-01 1.02104115e+00 6.49467349e-01 -2.77166694e-01 -6.18586481e-01 -9.09951687e-01 -6.13726735e-01 -2.74332732e-01 -8.15669656e-01 5.88447392e-01 5.40856063e-01 -4.81781095e-01 7.28494525e-02 1.21460393e-01 3.75186950e-01 3.77527952e-01 2.42816150e-01 7.45365739e-01 -1.11105978e+00 3.27610344e-01 -7.62650669e-01 -5.67181349e-01 -1.18573129e+00 3.03860027e-02 -6.57341301e-01 3.64221364e-01 -2.17499948e+00 -2.51134783e-01 -5.27072370e-01 -2.27230400e-01 4.91679519e-01 -8.33901986e-02 8.59498441e-01 -3.36447246e-02 2.27919295e-01 -7.20797479e-01 1.99957773e-01 1.03069019e+00 -1.00656755e-01 -3.46710712e-01 -7.48272836e-02 -7.04052567e-01 8.44734788e-01 1.04447579e+00 -2.27836385e-01 -1.58784747e-01 -4.07951385e-01 -8.36167857e-02 -3.26362133e-01 5.33253670e-01 -1.42368281e+00 4.94596809e-01 -1.52808698e-02 5.24304986e-01 -6.99210286e-01 5.41274905e-01 -8.89407277e-01 -3.77721116e-02 4.27232921e-01 -2.13721529e-01 -2.69276798e-01 2.31772780e-01 4.07951266e-01 -4.13222045e-01 -1.62367165e-01 9.44061935e-01 -1.45047784e-01 -1.33860636e+00 1.87436968e-01 -3.94061774e-01 1.80640593e-02 1.14571345e+00 -7.50950754e-01 -1.85031831e-01 -2.26283729e-01 -5.70195973e-01 2.97824085e-01 7.05352366e-01 4.75103647e-01 6.28231823e-01 -8.45621049e-01 -4.53553885e-01 5.05445004e-01 1.44469529e-01 4.83302981e-01 2.53702343e-01 8.36619675e-01 -8.36780131e-01 5.87134004e-01 -1.51584640e-01 -7.18545377e-01 -1.08153725e+00 1.29134506e-01 4.41206127e-01 -7.55354688e-02 -8.00156355e-01 8.33553731e-01 2.99912393e-01 -2.17063844e-01 8.73937458e-02 -5.29157817e-01 -3.73807102e-01 1.41780749e-01 6.17038727e-01 9.11089703e-02 2.73999870e-01 -1.00943649e+00 -3.33425939e-01 8.82310987e-01 -6.35001063e-03 -2.63890803e-01 1.12733698e+00 -3.15238744e-01 3.04603055e-02 1.41349643e-01 1.10288489e+00 -3.31314594e-01 -1.52504957e+00 -1.84662849e-01 -5.42131299e-03 -3.90543520e-01 3.15799505e-01 -6.99165702e-01 -1.16011930e+00 9.27961826e-01 5.79765022e-01 2.87319392e-01 1.08690524e+00 -6.92155287e-02 9.64033961e-01 2.04638094e-01 2.23371387e-01 -1.38381922e+00 -2.12042987e-01 6.65913939e-01 4.46577132e-01 -1.48494935e+00 -2.22162992e-01 -4.13353890e-01 -5.23514867e-01 9.95014071e-01 4.13159877e-01 -8.45670551e-02 3.41813654e-01 4.90356609e-02 3.14584434e-01 -2.05338880e-01 -6.98683113e-02 -8.53648007e-01 4.72039580e-01 4.12802666e-01 6.05722470e-03 -3.46778780e-02 6.20088950e-02 2.76743025e-01 -3.43939513e-01 -2.63279289e-01 1.79851502e-01 8.70337784e-01 -9.47105944e-01 -7.46919751e-01 -5.07053077e-01 4.45853323e-01 -2.39800677e-01 -2.69675944e-02 -2.35216349e-01 7.67094076e-01 3.16225559e-01 1.21649611e+00 3.39785308e-01 -3.18734765e-01 5.28315842e-01 -6.02056365e-03 5.40437512e-02 -4.84277040e-01 -4.42081332e-01 -1.06634900e-01 -8.18596929e-02 -9.10422921e-01 -3.91158670e-01 -5.38950026e-01 -1.68995082e+00 1.52175978e-01 -1.51681781e-01 -5.89503534e-02 1.14386189e+00 1.08181214e+00 2.78638482e-01 7.47765422e-01 3.67303103e-01 -1.15591586e+00 1.24198519e-01 -5.71894884e-01 -3.74927491e-01 3.05400163e-01 5.17210841e-01 -4.36107397e-01 -2.57642657e-01 9.71412957e-02]
[9.012113571166992, -1.1556973457336426]
2306c7aa-a994-4548-aa51-57dfdf2f7697
vehicle-and-license-plate-recognition-with
2202.05631
null
https://arxiv.org/abs/2202.05631v2
https://arxiv.org/pdf/2202.05631v2.pdf
Vehicle and License Plate Recognition with Novel Dataset for Toll Collection
We propose an automatic framework for toll collection, consisting of three steps: vehicle type recognition, license plate localization, and reading. However, each of the three steps becomes non-trivial due to image variations caused by several factors. The traditional vehicle decorations on the front cause variations among vehicles of the same type. These decorations make license plate localization and recognition difficult due to severe background clutter and partial occlusions. Likewise, on most vehicles, specifically trucks, the position of the license plate is not consistent. Lastly, for license plate reading, the variations are induced by non-uniform font styles, sizes, and partially occluded letters and numbers. Our proposed framework takes advantage of both data availability and performance evaluation of the backbone deep learning architectures. We gather a novel dataset, \emph{Diverse Vehicle and License Plates Dataset (DVLPD)}, consisting of 10k images belonging to six vehicle types. Each image is then manually annotated for vehicle type, license plate, and its characters and digits. For each of the three tasks, we evaluate You Only Look Once (YOLO)v2, YOLOv3, YOLOv4, and FasterRCNN. For real-time implementation on a Raspberry Pi, we evaluate the lighter versions of YOLO named Tiny YOLOv3 and Tiny YOLOv4. The best Mean Average Precision (mAP@0.5) of 98.8% for vehicle type recognition, 98.5% for license plate detection, and 98.3% for license plate reading is achieved by YOLOv4, while its lighter version, i.e., Tiny YOLOv4 obtained a mAP of 97.1%, 97.4%, and 93.7% on vehicle type recognition, license plate detection, and license plate reading, respectively. The dataset and the training codes are available at https://github.com/usama-x930/VT-LPR
['Saeed Anwar', 'Abbas Anwar', 'Hafeez Anwar', 'Muhammad Usama']
2022-02-11
null
null
null
null
['license-plate-recognition', 'license-plate-detection']
['computer-vision', 'computer-vision']
[-3.18413675e-01 -7.43082345e-01 -1.23061180e-01 -3.45229298e-01 -8.32745016e-01 -1.05458450e+00 4.13550615e-01 -5.05096674e-01 -1.88552663e-01 5.70304096e-01 -5.83029568e-01 -4.08907950e-01 4.26301450e-01 -7.89619207e-01 -1.08000839e+00 -7.29678154e-01 5.14828444e-01 3.68502945e-01 3.60553086e-01 2.79652067e-02 2.42424116e-01 5.62692285e-01 -1.39207649e+00 3.03382158e-01 7.87336111e-01 1.33497393e+00 1.92384273e-01 4.89842772e-01 -1.85112134e-01 5.78302681e-01 -6.43029034e-01 -6.15394413e-01 4.67189640e-01 2.52916068e-01 -5.17193340e-02 2.39087418e-01 6.09043419e-01 -6.50677621e-01 -5.79310656e-01 1.30309665e+00 9.54573601e-02 -8.33771676e-02 8.92520249e-01 -1.60717773e+00 -1.03091931e+00 -5.55208512e-02 -7.87773609e-01 2.26141997e-02 -1.24149665e-01 5.27982295e-01 5.48766851e-01 -1.29445767e+00 3.82554770e-01 1.01378858e+00 8.55123878e-01 4.81128216e-01 -8.90452862e-01 -1.05409133e+00 -6.14241771e-02 4.39360201e-01 -1.80013335e+00 -5.73001921e-01 5.47082722e-01 -7.07250237e-01 5.78007817e-01 3.54502141e-01 2.34174013e-01 9.39102829e-01 1.42862186e-01 9.88618255e-01 1.11072302e+00 -4.82009761e-02 8.51589739e-02 5.88293612e-01 4.96461511e-01 5.33297241e-01 4.77080703e-01 -1.10021435e-01 3.52589004e-02 4.10096914e-01 7.38648713e-01 3.07442635e-01 -1.03751689e-01 -1.12248920e-02 -7.43427932e-01 4.62052524e-01 7.16429725e-02 -2.76170522e-01 -1.16370544e-01 -1.13685634e-02 2.97262579e-01 5.49328923e-02 -2.03640051e-02 -2.66176343e-01 -4.01262909e-01 3.55097465e-02 -7.95153737e-01 1.52404070e-01 5.30421078e-01 1.52372420e+00 9.52068150e-01 4.39737260e-01 -1.32651389e-01 1.09637666e+00 5.14370263e-01 1.39692974e+00 2.76634008e-01 -5.68622768e-01 9.06595111e-01 3.03008258e-01 3.81140411e-01 -1.11396563e+00 -2.38126859e-01 -9.67080891e-02 -7.37987101e-01 4.38033700e-01 6.14668548e-01 -1.74467593e-01 -1.20384049e+00 1.14542925e+00 -2.00757951e-01 8.72341096e-02 6.47513345e-02 8.98025334e-01 1.20000672e+00 1.01167214e+00 -1.58574134e-01 2.27689981e-01 1.48388326e+00 -1.10658431e+00 -5.53724229e-01 -4.53946412e-01 2.57869005e-01 -8.03454518e-01 1.10845745e+00 1.35950834e-01 -6.97898090e-01 -7.36518383e-01 -1.15060508e+00 1.16235785e-01 -6.46237910e-01 7.24594593e-01 4.30476069e-02 7.76127696e-01 -8.13975334e-01 -1.33431971e-01 -3.58651280e-01 2.53204703e-02 3.85240793e-01 3.23507369e-01 -2.58743823e-01 -3.33031654e-01 -1.15309787e+00 7.86779642e-01 -1.67443171e-01 4.13290977e-01 -6.72178209e-01 -4.38257813e-01 -7.84079850e-01 -9.20366198e-02 2.80413181e-01 -1.27320420e-02 1.06098425e+00 -9.19005871e-01 -1.36824501e+00 8.67075980e-01 -1.64337903e-01 -1.93599239e-01 5.81250250e-01 4.50166836e-02 -1.00566125e+00 -1.40144467e-01 1.41921073e-01 5.23095965e-01 9.26687777e-01 -1.33239901e+00 -7.39504278e-01 -2.98282623e-01 -1.40782923e-01 -1.30165339e-01 1.69104695e-01 -8.14839900e-02 -9.86941576e-01 -3.80311072e-01 -3.39061320e-01 -1.05785072e+00 4.32619601e-01 -1.86609596e-01 -5.69232345e-01 -1.91701218e-01 8.60112548e-01 -9.75822091e-01 8.54843616e-01 -2.58856320e+00 -8.76638412e-01 3.32888037e-01 3.44083793e-02 6.39439285e-01 -1.70354486e-01 -6.26616180e-02 1.04648560e-01 2.20615074e-01 1.10566854e-01 -5.57434931e-02 3.53581786e-01 -2.56468579e-02 -2.86708325e-01 5.76043546e-01 2.47645169e-01 9.89840269e-01 -2.45365858e-01 -2.99750239e-01 4.25793350e-01 4.49243277e-01 -8.50532204e-02 -1.99402615e-01 1.68526798e-01 -2.01370612e-01 -4.20233160e-01 1.32120049e+00 1.41133559e+00 8.71339366e-02 -1.91041619e-01 -3.23114038e-01 -3.81071299e-01 -1.83532253e-01 -1.34316885e+00 5.42839885e-01 -3.34789664e-01 1.03828216e+00 6.75907359e-02 -4.76099521e-01 1.14417946e+00 2.53444910e-02 2.23853756e-02 -1.15422332e+00 1.62937120e-01 3.53468567e-01 -3.84135574e-01 -4.97622490e-01 6.28565371e-01 3.61522883e-01 -2.06156954e-01 -1.44339055e-01 -3.77842367e-01 2.92237282e-01 3.87505531e-01 -1.89774796e-01 6.41717196e-01 -3.76057744e-01 -1.62015080e-01 -1.87490657e-02 5.06802559e-01 1.38001457e-01 6.53583527e-01 8.13222766e-01 -4.39892143e-01 6.89388037e-01 3.36567581e-01 -4.84784961e-01 -1.18951893e+00 -1.27219081e+00 -3.34003478e-01 6.94803536e-01 5.04251182e-01 2.07486063e-01 -6.63858414e-01 -2.58085877e-01 2.73754001e-01 6.20546162e-01 -2.53052264e-01 6.06738925e-02 -7.45158494e-01 -5.47235310e-01 8.70541334e-01 7.22636819e-01 9.30728257e-01 -8.56730163e-01 1.55673251e-01 -1.43053785e-01 -4.74787466e-02 -1.30649173e+00 -6.87924147e-01 -2.60574698e-01 -3.29332739e-01 -1.04933310e+00 -8.23727071e-01 -1.06259668e+00 6.39738917e-01 5.51330388e-01 8.84591818e-01 -1.36835948e-01 1.29781395e-01 3.98217887e-02 -4.66408394e-02 -3.20247173e-01 -2.71574855e-01 -4.26858515e-01 1.62490830e-01 3.11169207e-01 6.09580636e-01 1.01496950e-01 -4.94657546e-01 6.99806094e-01 -6.23242736e-01 -2.32424363e-01 7.57644296e-01 5.25454283e-01 6.25744641e-01 9.87905413e-02 1.72076881e-01 -5.68468094e-01 2.36799672e-01 -3.56220633e-01 -1.06489921e+00 2.50754863e-01 -4.56319094e-01 -6.00938797e-01 8.32762420e-01 -3.08413684e-01 -7.45973825e-01 1.86702013e-01 -3.26803505e-01 -7.47247577e-01 -4.45069909e-01 -4.88363951e-02 -5.70253074e-01 -6.37776628e-02 1.65722609e-01 6.37290061e-01 -2.28374936e-02 -3.77738386e-01 7.91642666e-02 1.06122279e+00 8.13225687e-01 -1.69953071e-02 1.05921984e+00 2.24298418e-01 -5.00687361e-01 -1.03183484e+00 -1.42316110e-02 -3.11083764e-01 -2.64535010e-01 -4.31437135e-01 7.85865009e-01 -1.15602803e+00 -9.68715549e-01 1.12756610e+00 -1.11431158e+00 -1.09035164e-01 2.99276233e-01 3.50777596e-01 -8.54982361e-02 2.94078261e-01 -6.19328737e-01 -6.91328466e-01 -1.67544261e-01 -1.47374761e+00 9.60874498e-01 4.37335223e-01 3.38345915e-01 -5.45748234e-01 -4.77448374e-01 5.25118887e-01 3.00298542e-01 -1.31808281e-01 7.32330859e-01 -7.21099913e-01 -8.45901132e-01 -6.43521488e-01 -6.34084702e-01 5.83132625e-01 -1.42938957e-01 3.06430250e-01 -1.05532646e+00 -1.86883751e-02 -4.93210733e-01 -9.06496681e-03 8.51447642e-01 4.96824414e-01 9.07700717e-01 -2.51395375e-01 -2.46946305e-01 7.74422109e-01 1.42551410e+00 7.23753810e-01 8.75606179e-01 6.88902259e-01 8.83783340e-01 1.89611807e-01 6.04581714e-01 2.49633208e-01 5.67698956e-01 8.25320959e-01 3.46694589e-01 -8.72736424e-02 -7.28267655e-02 -1.02721140e-01 6.33855224e-01 5.07943273e-01 2.21737958e-02 -5.11507452e-01 -1.05094945e+00 2.30176538e-01 -1.26945901e+00 -1.09413052e+00 -5.30177057e-01 2.13310814e+00 3.38201165e-01 2.02347547e-01 1.96825221e-01 3.43251266e-02 1.09734786e+00 4.29345518e-02 -6.76184297e-01 -2.81462133e-01 -2.50501066e-01 -4.74231839e-01 1.14240730e+00 2.69747525e-01 -1.33531427e+00 9.15661991e-01 5.35127687e+00 9.43414807e-01 -1.30207133e+00 -2.81472236e-01 7.56440818e-01 2.16410190e-01 -6.03333972e-02 -5.05014300e-01 -1.25720870e+00 1.00615370e+00 5.68073213e-01 4.92099375e-02 4.89430666e-01 9.71493185e-01 2.17655927e-01 9.01611373e-02 -7.79850543e-01 1.35603225e+00 1.81707680e-01 -1.32871890e+00 -1.45968840e-01 -1.16311766e-01 5.68700492e-01 1.42714247e-01 4.93855059e-01 6.63605154e-01 2.61703804e-02 -1.01263213e+00 1.05861974e+00 4.39001799e-01 1.05928361e+00 -7.96402097e-01 9.49149191e-01 2.00022191e-01 -1.21845639e+00 -2.48428397e-02 -6.05092824e-01 4.90581185e-01 -9.53170285e-02 2.75684536e-01 -6.91922486e-01 3.49431746e-02 7.36898482e-01 5.93212008e-01 -5.33714950e-01 9.63536084e-01 -4.68436675e-03 5.34017861e-01 -2.73759961e-01 -1.15985684e-01 3.37909579e-01 -2.11603329e-01 3.36745024e-01 1.35578692e+00 3.47674340e-01 -1.03427239e-01 1.01204775e-01 9.24583375e-01 -3.11019540e-01 -1.51961595e-01 -5.46698868e-01 2.04720885e-01 8.51656020e-01 1.23390830e+00 -5.79018354e-01 -3.70878875e-01 -7.08289444e-01 6.72179222e-01 -3.26040655e-01 6.31084442e-01 -1.48351085e+00 -5.10686100e-01 7.66176462e-01 2.81986943e-03 6.84571505e-01 -1.23113573e-01 -4.50648099e-01 -1.02498329e+00 3.81997675e-01 -8.49131525e-01 5.86698167e-02 -7.78181911e-01 -1.07855892e+00 5.18527687e-01 -4.02195126e-01 -1.66875124e+00 -3.63782188e-03 -1.08084226e+00 -4.97838497e-01 9.12656844e-01 -1.56097806e+00 -1.01956522e+00 -4.38874036e-01 5.85905612e-01 8.59984279e-01 -5.26727259e-01 2.68862396e-01 7.87203670e-01 -1.05806375e+00 1.14108336e+00 8.60084414e-01 7.07839131e-01 6.08650625e-01 -7.89473414e-01 4.97004390e-01 8.18335295e-01 -2.03783140e-01 2.60626823e-01 3.29161584e-01 -4.35585171e-01 -1.81244135e+00 -1.50301898e+00 6.32856905e-01 -3.67101848e-01 4.69804168e-01 -4.19583201e-01 -8.51027191e-01 7.06664801e-01 -2.34461695e-01 9.12772343e-02 3.02932978e-01 -4.19223428e-01 -4.94214773e-01 -5.56294858e-01 -1.20783448e+00 4.91690397e-01 3.37579936e-01 -3.85313362e-01 -1.80250019e-01 2.65032232e-01 2.60964930e-01 -5.38074255e-01 -4.09270585e-01 5.34325419e-03 8.56325984e-01 -7.48276234e-01 1.11888242e+00 3.35270092e-02 3.02796155e-01 -6.32961869e-01 -4.47437316e-01 -8.82820904e-01 -3.99017811e-01 1.50660783e-01 3.81329477e-01 1.57253814e+00 5.87304056e-01 -7.56885946e-01 5.99172115e-01 8.56481016e-01 -4.16691899e-01 -3.08966905e-01 -7.74387658e-01 -1.06419706e+00 -6.19975431e-03 -6.43072307e-01 7.35821366e-01 6.37255251e-01 -5.77773511e-01 -2.94520669e-02 -6.14615560e-01 6.91017985e-01 5.41902363e-01 2.73930132e-01 8.95688534e-01 -6.99634969e-01 -1.42798014e-02 -5.42590499e-01 -6.53885543e-01 -1.20018566e+00 5.56675233e-02 -5.87676167e-01 2.32035160e-01 -1.27607012e+00 7.12774321e-02 -3.82063448e-01 -2.27632403e-01 4.24999595e-01 1.26812413e-01 6.52293444e-01 5.11651814e-01 4.98559624e-01 -5.92057347e-01 6.02086112e-02 8.90299618e-01 -7.25408912e-01 -7.15223476e-02 2.89780676e-01 -3.07461172e-01 8.86216104e-01 1.02034199e+00 -2.42806941e-01 2.78985854e-02 -8.00239205e-01 -1.66895255e-01 2.13869885e-02 5.92561781e-01 -9.25786674e-01 2.68088669e-01 -7.67727792e-02 7.78564394e-01 -9.86263990e-01 3.61979008e-01 -7.43304312e-01 7.46214613e-02 2.26262644e-01 2.25570172e-01 1.38782546e-01 3.56209606e-01 3.37216049e-01 -2.50689179e-01 -1.53910995e-01 8.15004766e-01 9.98270735e-02 -1.49210763e+00 3.14785779e-01 -7.60505795e-01 -7.83675015e-02 1.06249297e+00 -6.54561222e-01 -6.03430450e-01 -1.70293748e-01 -2.29056090e-01 2.94293731e-01 5.05765796e-01 5.74832141e-01 7.23794639e-01 -1.47450721e+00 -7.24062085e-01 3.80299836e-01 1.73942000e-01 -2.99862683e-01 5.09024858e-01 6.58207297e-01 -7.82897353e-01 6.70777142e-01 -2.95620084e-01 -6.65269196e-01 -1.29841995e+00 4.23887461e-01 3.78395885e-01 2.00759694e-01 -4.52104092e-01 6.38049006e-01 3.91705394e-01 -3.15870196e-01 1.81000441e-01 -2.55135238e-01 -2.93901980e-01 -1.69734154e-02 6.96280301e-01 6.91282034e-01 1.40325606e-01 -9.82260346e-01 -7.41037130e-01 9.03956473e-01 -1.16219364e-01 3.21681470e-01 9.34022546e-01 -1.18577115e-01 3.33965957e-01 2.38708109e-01 1.13477290e+00 1.54283434e-01 -1.30502343e+00 7.29790404e-02 -3.09942931e-01 -5.37834942e-01 -2.76661754e-01 -8.95836890e-01 -1.25933743e+00 7.64272690e-01 7.86507785e-01 -1.94739713e-03 8.67054701e-01 -2.26095304e-01 9.34518337e-01 4.87009734e-01 3.15918446e-01 -1.11205804e+00 -5.30521512e-01 8.09374630e-01 6.88579679e-01 -1.36328363e+00 -3.16068619e-01 -2.34883174e-01 -9.00120020e-01 1.03962862e+00 7.32877612e-01 -4.74636182e-02 2.20008969e-01 4.78703767e-01 3.87323350e-01 1.45618364e-01 -3.14444333e-01 -2.21314058e-02 2.89541900e-01 6.86367512e-01 -4.29382594e-03 3.49459440e-01 2.63083160e-01 7.58831918e-01 -4.75628749e-02 -1.74714923e-01 4.46127385e-01 5.00233591e-01 -4.42053109e-01 -4.91909593e-01 -8.19988847e-01 4.61477309e-01 -2.33879909e-01 -9.15729329e-02 -2.46864021e-01 1.18294263e+00 2.59027690e-01 1.01497352e+00 4.17442501e-01 -6.52812481e-01 6.04471624e-01 -4.10426334e-02 -8.29021856e-02 3.23520526e-02 -5.27149849e-02 -4.87088365e-03 4.68089283e-02 -2.51895010e-01 2.22092777e-01 -6.44708633e-01 -1.35021853e+00 -9.11128283e-01 -1.63840234e-01 -2.53294617e-01 6.97292328e-01 7.60453463e-01 1.89554796e-01 2.31820613e-01 8.47191632e-01 -6.40145779e-01 -5.23389459e-01 -6.23068094e-01 -8.86790395e-01 2.14365751e-01 4.25098181e-01 -6.02422953e-01 -2.27557212e-01 9.63588059e-02]
[9.841649055480957, -4.916127681732178]
f29f3e19-99db-445d-aafa-fd7e7b155143
dimensionality-expansion-and-transfer
2204.02802
null
https://arxiv.org/abs/2204.02802v4
https://arxiv.org/pdf/2204.02802v4.pdf
Dimensionality Expansion of Load Monitoring Time Series and Transfer Learning for EMS
Energy management systems (EMS) rely on (non)-intrusive load monitoring (N)ILM to monitor and manage appliances and help residents be more energy efficient and thus more frugal. The robustness as well as the transfer potential of the most promising machine learning solutions for (N)ILM is not yet fully understood as they are trained and evaluated on relatively limited data. In this paper, we propose a new approach for load monitoring in building EMS based on dimensionality expansion of time series and transfer learning. We perform an extensive evaluation on 5 different low-frequency datasets. The proposed feature dimensionality expansion using video-like transformation and resource-aware deep learning architecture achieves an average weighted F1 score of 0.88 across the datasets with 29 appliances and is computationally more efficient compared to the state-of-the-art imaging methods. Investigating the proposed method for cross-dataset intra-domain transfer learning, we find that 1) our method performs with an average weighted F1 score of 0.80 while requiring 3-times fewer epochs for model training compared to the non-transfer approach, 2) can achieve an F1 score of 0.75 with only 230 data samples, and 3) our transfer approach outperforms the state-of-the-art in precision drop by up to 12 percentage points for unseen appliances.
['Carolina Fortuna', 'Jakob Jenko', 'Blaž Bertalanič']
2022-04-06
null
null
null
null
['non-intrusive-load-monitoring', 'non-intrusive-load-monitoring', 'non-intrusive-load-monitoring']
['knowledge-base', 'miscellaneous', 'time-series']
[-4.67378311e-02 -3.37944031e-01 -2.73703665e-01 -1.20997258e-01 -1.04985392e+00 -4.34603304e-01 3.37850392e-01 -9.21026915e-02 -2.63194412e-01 6.49169028e-01 -8.74392986e-02 -2.51402378e-01 -4.95092481e-01 -8.41176391e-01 -7.65500128e-01 -8.42593849e-01 -2.59089619e-01 3.19140315e-01 -1.76378131e-01 -5.99796027e-02 -2.85544485e-01 4.73987460e-01 -1.65418506e+00 2.15495517e-03 8.51443589e-01 1.43328762e+00 1.34775817e-01 5.01085937e-01 3.49859953e-01 8.05585444e-01 -6.53725445e-01 2.26062611e-01 2.29324058e-01 -7.03485757e-02 -9.79514718e-01 -1.18712187e-01 3.56342286e-01 -6.60750210e-01 -3.09910625e-01 7.34566391e-01 7.84411609e-01 2.73461938e-01 7.23979235e-01 -1.35610723e+00 -5.83472073e-01 2.57435471e-01 -4.59347934e-01 3.49800676e-01 2.47573107e-01 2.37278678e-02 5.59767723e-01 -2.50669867e-01 -1.65698186e-01 4.68468040e-01 9.24856305e-01 2.27881670e-01 -1.39532804e+00 -8.64337444e-01 -2.25969344e-01 7.23823130e-01 -1.58237517e+00 -3.78653079e-01 9.70787823e-01 -2.83324450e-01 1.63826787e+00 4.19352114e-01 5.03132284e-01 1.03041482e+00 1.92091975e-04 6.07491195e-01 9.18594360e-01 -5.12451053e-01 5.55451214e-01 2.02683866e-01 -2.44045451e-01 5.65900922e-01 2.66301576e-02 8.15855935e-02 -2.68870831e-01 -1.64254636e-01 6.55566826e-02 -2.26186603e-01 -3.81603390e-02 -1.26568496e-01 -8.42335939e-01 8.61087084e-01 4.29482758e-01 6.51092708e-01 -4.08501685e-01 2.83650577e-01 7.43410170e-01 3.75586182e-01 6.14138842e-01 9.72676203e-02 -8.33478093e-01 -3.41802746e-01 -1.16769254e+00 -2.57122993e-01 6.40275955e-01 7.82444358e-01 6.01195157e-01 3.85766715e-01 1.41274147e-02 8.05984855e-01 -7.86686614e-02 7.62700438e-01 8.68821204e-01 -8.33931863e-01 3.98364007e-01 5.12809515e-01 1.88525468e-01 -3.89893502e-01 -1.01232278e+00 -1.49119675e-01 -1.02321970e+00 2.70001702e-02 -1.14120401e-01 -2.71014571e-01 -4.97580469e-01 1.91227067e+00 2.08890513e-01 1.47691756e-01 -1.55501321e-01 2.80420989e-01 4.22222555e-01 8.32150757e-01 2.25075558e-01 -3.51106018e-01 1.28325188e+00 -7.01814651e-01 -7.38097370e-01 3.83889005e-02 7.53885031e-01 -5.15477180e-01 1.07888651e+00 3.48213017e-01 -7.90968239e-01 -6.09392524e-01 -1.31360018e+00 8.87237489e-02 -7.22816229e-01 2.75779963e-01 3.42031717e-01 9.77519274e-01 -1.13159049e+00 8.19768190e-01 -1.02306235e+00 -7.53037512e-01 5.84418595e-01 7.16606796e-01 -1.54001057e-01 2.82610953e-01 -1.02827322e+00 1.10176289e+00 5.29290736e-01 -3.87862980e-01 -7.55829513e-01 -1.03784156e+00 -7.02368736e-01 1.84974313e-01 -5.44881411e-02 -6.23652697e-01 1.30201852e+00 -4.72614467e-01 -1.68914926e+00 5.01028717e-01 8.23375508e-02 -4.99650747e-01 3.26330841e-01 -4.54323798e-01 -8.61535728e-01 1.98472142e-01 -4.66673309e-03 3.87081057e-01 7.88660705e-01 -8.59339476e-01 -8.68021727e-01 -2.82654226e-01 -1.98390171e-01 -2.86543369e-01 -9.22764122e-01 -2.83202171e-01 3.85421365e-01 -3.99745524e-01 -4.31956261e-01 -1.00497937e+00 4.94683892e-01 -1.82382241e-01 -4.19350564e-02 -3.24434429e-01 1.36982250e+00 -9.49299872e-01 1.37660313e+00 -1.88306069e+00 -2.34291017e-01 8.88513625e-02 -2.95415223e-01 5.18579423e-01 -2.84169316e-02 4.33365285e-01 -5.31476974e-01 -2.32755601e-01 -2.02649802e-01 -2.24767402e-01 3.32869142e-01 1.07446030e-01 2.02906914e-02 7.29299843e-01 8.91599506e-02 9.17629957e-01 -8.46034825e-01 -1.07607603e-01 9.23220634e-01 6.42735600e-01 -1.59578934e-01 8.45116973e-02 1.24256477e-01 3.45430881e-01 -2.63694435e-01 6.14683211e-01 3.94259661e-01 -2.74818569e-01 1.08862564e-01 -7.02244222e-01 -9.07042325e-02 2.20296800e-01 -9.91738498e-01 1.71243799e+00 -1.04944050e+00 7.46242225e-01 -3.69855165e-01 -1.41396892e+00 6.59168184e-01 5.78354299e-01 1.14378560e+00 -1.08813727e+00 2.04831570e-01 2.27608159e-01 -3.98900300e-01 -8.07913959e-01 1.10419400e-01 -1.73934743e-01 -1.90622061e-01 5.43177843e-01 2.15517610e-01 2.09025010e-01 6.00379556e-02 -5.72151661e-01 1.37560856e+00 1.12788910e-02 2.96940625e-01 -5.87941170e-01 5.68764508e-01 -4.57060277e-01 1.10464782e-01 1.47070125e-01 -2.13041395e-01 -1.67291790e-01 -2.38504127e-01 -5.09041905e-01 -1.20836222e+00 -9.42277670e-01 -3.56378287e-01 1.29345226e+00 -4.75237459e-01 -1.34091675e-01 -9.20227647e-01 -4.65402067e-01 4.58824150e-02 1.31801546e+00 -4.83178318e-01 -4.63644952e-01 -5.60688615e-01 -1.00607574e+00 6.53462648e-01 9.04982746e-01 9.12653565e-01 -8.12948644e-01 -1.02434778e+00 3.66513968e-01 -5.97341537e-01 -1.12452614e+00 -2.29699612e-01 6.02593124e-01 -8.39118421e-01 -9.77788925e-01 -4.45103198e-01 -6.19763494e-01 2.87622482e-01 -4.30253102e-03 1.28068495e+00 -3.63525003e-01 -4.34863806e-01 6.54620171e-01 -3.06455493e-01 -4.94250298e-01 -3.81625563e-01 4.70965594e-01 3.65670621e-01 -2.66170055e-01 9.32723403e-01 -1.02682865e+00 -7.98254311e-01 4.16560829e-01 -7.05973625e-01 -5.19265473e-01 4.08074558e-01 6.29050553e-01 3.35926294e-01 7.54845858e-01 9.23331976e-01 -2.04355523e-01 4.00659233e-01 -5.23252785e-01 -7.84045756e-01 1.44080803e-01 -1.10199106e+00 2.80635562e-02 7.93047667e-01 -5.14732480e-01 -9.75314438e-01 6.95896968e-02 -2.69210115e-02 -4.36565816e-01 -2.17187420e-01 -2.79868215e-01 -1.45100728e-01 -6.55126423e-02 6.55697763e-01 1.73639804e-01 -3.83112282e-01 -4.89653587e-01 2.12211996e-01 7.30624497e-01 5.25009513e-01 -6.02598846e-01 9.74882007e-01 5.34707010e-01 2.53189772e-01 -9.65030611e-01 -6.45240426e-01 -5.28156221e-01 -6.67154551e-01 -1.29867777e-01 9.81630683e-01 -1.00340784e+00 -8.57578814e-01 5.13969839e-01 -7.40044832e-01 -6.16952837e-01 -5.78700960e-01 4.53879386e-01 -8.07564437e-01 1.52508721e-01 -4.51422334e-01 -9.92975354e-01 -7.88095713e-01 -7.96243250e-01 1.20927382e+00 -6.22951575e-02 -4.93283004e-01 -1.01061320e+00 1.34732768e-01 3.06660384e-01 7.67798722e-01 5.00796080e-01 1.22119141e+00 -4.83348042e-01 -4.28541303e-02 -3.13927025e-01 -2.39863396e-02 6.54146075e-01 6.42505884e-01 -5.50902486e-01 -1.57016921e+00 -7.08879471e-01 3.26867551e-01 -2.68759042e-01 4.70463097e-01 4.39139962e-01 1.37324977e+00 -2.98986465e-01 -3.82299334e-01 3.45654398e-01 1.55499697e+00 3.44925344e-01 7.82588720e-01 5.18417239e-01 4.11714911e-01 2.70272225e-01 1.75975457e-01 6.33873045e-01 3.45785618e-01 6.37924671e-01 3.68249685e-01 -1.63553864e-01 5.05300052e-02 2.39067003e-02 4.10286129e-01 9.16398466e-01 -8.35161582e-02 -1.54349610e-01 -4.89801884e-01 7.40422070e-01 -1.77507424e+00 -9.95162725e-01 3.15572321e-01 2.15162921e+00 4.63519484e-01 -1.14475684e-02 4.28422153e-01 6.89570546e-01 4.48147714e-01 6.27339631e-02 -9.65662360e-01 -2.35472843e-01 3.19651604e-01 5.47400892e-01 7.35493422e-01 -5.79754896e-02 -1.38657117e+00 9.79415923e-02 6.47419882e+00 7.31301606e-01 -8.28146756e-01 4.67324883e-01 4.28340375e-01 -4.36012655e-01 3.83085728e-01 -7.23112106e-01 -4.51423973e-01 7.81716883e-01 1.58561778e+00 -1.26566082e-01 8.64528179e-01 1.11189651e+00 2.28177488e-01 -9.57664549e-02 -1.46836412e+00 1.26535594e+00 -3.07431724e-02 -8.19257140e-01 -4.85536486e-01 2.67221242e-01 9.00602758e-01 3.25899273e-01 2.23227572e-02 5.48781276e-01 -3.61060575e-02 -8.67301762e-01 4.56567347e-01 4.79009956e-01 8.94946396e-01 -9.68191803e-01 7.71089971e-01 3.92841667e-01 -1.72166264e+00 -6.18020177e-01 -2.82562017e-01 4.84900065e-02 8.48296434e-02 5.00299931e-01 -4.80001986e-01 8.05825591e-01 1.26862073e+00 1.80874497e-01 -3.83595318e-01 5.67937315e-01 3.11426640e-01 7.33444393e-01 -7.82130241e-01 1.83583081e-01 1.73870310e-01 3.81572917e-03 -9.90910307e-02 1.23741961e+00 7.57937431e-01 -7.11288527e-02 -8.63375957e-04 5.78046679e-01 -7.65768215e-02 -1.77039817e-01 -5.09913743e-01 2.41657093e-01 4.19814110e-01 1.30479813e+00 -3.78224552e-01 -2.48405606e-01 -5.82766831e-01 1.11196077e+00 2.21648645e-02 9.57427546e-02 -1.14771342e+00 -2.99342543e-01 6.59315288e-01 1.74839810e-01 6.48334384e-01 1.67734437e-02 1.24541327e-01 -6.08027875e-01 7.62409940e-02 -4.76363689e-01 4.80030686e-01 -5.59316158e-01 -1.55823421e+00 2.78791070e-01 3.18071634e-01 -1.25053024e+00 -4.47472721e-01 -6.35112107e-01 -5.87650061e-01 5.01432061e-01 -1.69609618e+00 -1.12413490e+00 -4.94881362e-01 6.24949217e-01 5.50729752e-01 -1.89522877e-01 1.30500913e+00 9.66345489e-01 -5.07997632e-01 6.01970851e-01 5.11700094e-01 -1.82643548e-01 3.21448594e-01 -1.27716208e+00 3.23077202e-01 3.30911905e-01 -1.37894467e-01 -6.24466725e-02 4.59326625e-01 -1.78175315e-01 -1.19472337e+00 -1.49372697e+00 3.99763823e-01 -6.20433390e-01 5.62743664e-01 -1.73802450e-01 -8.03829253e-01 6.84073269e-01 3.64774436e-01 -2.27334350e-01 1.15307415e+00 -5.52862845e-02 -1.20266177e-01 -4.59564596e-01 -1.67184377e+00 -4.01543081e-02 9.11062002e-01 -7.49870658e-01 -4.79633659e-01 6.62293613e-01 4.43497479e-01 -2.74988897e-02 -1.40472388e+00 4.57265347e-01 5.27422130e-01 -8.14505994e-01 1.06175280e+00 -2.05143228e-01 -2.43402943e-01 -2.91675955e-01 -7.00297654e-01 -1.18080819e+00 -6.62518203e-01 -5.80738306e-01 -7.53170788e-01 1.43195200e+00 7.23422021e-02 -5.51501811e-01 4.64824706e-01 5.85317016e-01 -1.80604551e-02 -4.59653825e-01 -1.26956558e+00 -1.19988823e+00 6.76398650e-02 -6.49455190e-01 1.06451344e+00 8.82500231e-01 -9.18130055e-02 2.47019291e-01 -3.01524550e-01 2.68130869e-01 5.01621962e-01 -2.58046091e-01 3.42967659e-01 -1.31819344e+00 -9.41217765e-02 -2.95991689e-01 -2.93527931e-01 -3.82694304e-01 2.90739655e-01 -8.10550034e-01 -2.93226391e-01 -1.51056004e+00 1.75326094e-01 -9.08175483e-02 -7.68049419e-01 7.03328788e-01 4.55045223e-01 2.23598316e-01 -1.87273279e-01 -9.77606848e-02 -4.43033695e-01 7.58447707e-01 2.24465758e-01 -4.21638817e-01 -1.09902993e-01 -1.31827131e-01 -3.03229421e-01 8.18373322e-01 1.14279640e+00 -2.93103099e-01 -5.88080585e-01 -2.66049922e-01 -1.05245091e-01 -5.54292917e-01 4.33658391e-01 -1.46374810e+00 -1.54322341e-01 3.10558647e-01 8.64237070e-01 -7.57989347e-01 3.06143135e-01 -1.33438051e+00 5.38779736e-01 7.43354499e-01 2.93921024e-01 2.33941242e-01 6.77819431e-01 3.12642127e-01 4.37188804e-01 -1.51802570e-01 7.57846773e-01 4.63600717e-02 -7.02829003e-01 -3.83152738e-02 -3.98183912e-01 -3.71279806e-01 1.15788984e+00 -2.94980425e-02 -3.06071013e-01 -1.78912118e-01 -3.56189996e-01 7.07573816e-02 1.94853306e-01 4.12064970e-01 -1.12637974e-01 -1.75367868e+00 -3.47657174e-01 1.36992261e-01 5.83866611e-02 -3.03827614e-01 4.18152094e-01 6.07643366e-01 -1.52952358e-01 6.78909242e-01 -5.72868586e-02 -8.24384093e-01 -8.75432670e-01 8.16225231e-01 5.44104159e-01 -4.72798318e-01 -5.97362578e-01 2.09387943e-01 -1.69516265e-01 -5.05567491e-01 1.44917786e-01 -5.98616898e-01 1.95168525e-01 2.15633422e-01 5.16730487e-01 1.23863423e+00 6.58371210e-01 -5.81418455e-01 -4.69061553e-01 8.35343897e-01 2.80973017e-01 4.63160813e-01 1.59796202e+00 -3.58616412e-02 1.95507735e-01 5.53308785e-01 1.59374750e+00 -6.63890123e-01 -1.02224171e+00 1.07582016e-02 3.93944196e-02 4.55042087e-02 4.13682491e-01 -8.82176340e-01 -1.11261106e+00 4.94651228e-01 1.64661467e+00 5.44315636e-01 1.78187287e+00 -7.12026376e-03 1.23184478e+00 5.53497136e-01 5.04832566e-01 -1.61394536e+00 1.24453530e-01 -1.26031905e-01 6.53186738e-01 -1.17854178e+00 1.18211277e-01 2.42837653e-01 2.92710047e-02 8.06841671e-01 3.20657521e-01 1.79630920e-01 8.31152260e-01 3.43263477e-01 -3.74129564e-01 -2.08554104e-01 -4.34447616e-01 -5.82522973e-02 1.76359802e-01 9.53197420e-01 1.60871163e-01 1.89184025e-01 3.17171365e-01 3.46568614e-01 -2.44737029e-01 3.74791831e-01 -2.21381679e-01 9.76203680e-01 -2.64911771e-01 -7.30188727e-01 -4.74899501e-01 6.12702906e-01 -3.51478815e-01 3.51647347e-01 4.35716420e-01 8.61860275e-01 1.91187561e-01 1.21168470e+00 1.36355937e-01 -4.42218363e-01 7.09761083e-01 4.08011466e-01 5.72127581e-01 1.00752562e-01 -3.11364263e-01 -1.48222834e-01 -1.00537222e-02 -7.09990919e-01 -7.42413461e-01 -6.44084990e-01 -9.45526838e-01 -5.36259532e-01 -4.40955222e-01 -4.09206212e-01 7.42444873e-01 1.00166070e+00 4.07950163e-01 1.00838959e+00 1.01547372e+00 -1.07715988e+00 -5.93228459e-01 -1.09713948e+00 -7.35464394e-01 4.71940637e-01 2.56185621e-01 -8.14336538e-01 -3.39328527e-01 1.62435770e-01]
[16.05760955810547, 7.575520992279053]
b6d5baa4-fb65-41b4-96cf-b9bdd099577b
real-valued-syntactic-word-vectors-rsv-for
null
null
https://aclanthology.org/W17-0203
https://aclanthology.org/W17-0203.pdf
Real-valued Syntactic Word Vectors (RSV) for Greedy Neural Dependency Parsing
null
['Ali Basirat', 'Joakim Nivre']
2017-05-01
null
null
null
ws-2017-5
['transition-based-dependency-parsing']
['natural-language-processing']
[-8.63703638e-02 1.71006292e-01 -6.22772932e-01 -4.08054382e-01 -8.41685571e-03 -9.08429027e-01 6.55310392e-01 -6.53472245e-01 -2.85945535e-01 1.06888819e+00 -4.63127941e-02 -1.01159286e+00 -3.91567826e-01 -9.63214397e-01 -4.95059669e-01 -6.31337762e-01 -9.79754329e-01 7.25764990e-01 3.30370307e-01 -6.93831444e-01 7.03166842e-01 7.88774848e-01 -1.68942046e+00 7.18545914e-01 7.04417467e-01 8.52217197e-01 2.49141872e-01 1.14950800e+00 -1.95044339e-01 1.55633950e+00 -7.48382092e-01 -5.46825826e-01 3.13719302e-01 -1.23176083e-01 -7.22945035e-01 -1.01074085e-01 9.28529128e-02 -8.59008506e-02 -2.09758401e-01 9.22211111e-01 5.37373662e-01 4.49454933e-02 1.08379531e+00 -1.42548037e+00 -5.91619551e-01 6.10313773e-01 -4.01565880e-02 1.21627934e-01 1.03678203e+00 -5.39447069e-01 1.19919395e+00 -1.13026452e+00 7.20913768e-01 1.26888943e+00 8.66221786e-01 5.44149756e-01 -1.22286928e+00 -1.94712028e-01 -3.26822817e-01 -9.51717794e-02 -1.46558487e+00 -3.25250506e-01 4.25783843e-02 -2.08119690e-01 1.66093647e+00 1.26596653e+00 1.20609856e+00 1.01401424e+00 1.26658809e+00 8.34431887e-01 1.04267764e+00 -5.13792276e-01 3.35295945e-01 3.66983831e-01 1.54683650e-01 6.33519173e-01 8.40953708e-01 5.26628852e-01 -7.06372619e-01 -9.13127720e-01 9.33553874e-01 -2.94925272e-01 1.71355158e-01 -5.05680561e-01 -9.05919552e-01 6.91228509e-01 1.78732842e-01 3.83959889e-01 -1.39880210e-01 9.89067405e-02 1.26390755e-01 5.30987144e-01 -2.58292928e-02 6.47037446e-01 -9.11868811e-01 -1.33165747e-01 -8.71728659e-01 5.10332465e-01 1.25398111e+00 1.52653182e+00 1.24482810e-01 2.94908643e-01 -9.34252143e-02 3.17179203e-01 8.92314315e-01 1.01808000e+00 4.28362608e-01 -1.36146402e+00 -6.87414408e-02 1.72361732e-01 5.01781464e-01 -8.52631688e-01 -6.33224547e-01 -9.64177120e-03 -8.93263519e-01 4.49267089e-01 3.49161088e-01 4.57367361e-01 -8.02827001e-01 5.07305264e-01 4.33481112e-02 -2.34125629e-01 4.53833073e-01 5.55570945e-02 4.99930978e-01 3.76208365e-01 -1.34477139e-01 -5.73289394e-01 1.06082785e+00 -1.36716676e+00 -1.35299087e+00 2.33215362e-01 9.05734658e-01 -1.07320261e+00 4.35900748e-01 5.33875942e-01 -1.55548143e+00 -1.37560293e-01 -1.08699942e+00 1.78573877e-01 -7.27255583e-01 -3.14239264e-01 8.57801437e-01 1.43120694e+00 -1.60129595e+00 9.73287821e-01 -4.91727620e-01 6.59165755e-02 1.20568443e-02 8.24621081e-01 -2.64718989e-03 4.62812334e-01 -1.33193445e+00 1.08501506e+00 2.22979754e-01 -1.21242590e-01 -1.65216476e-01 -2.13068098e-01 -8.23704481e-01 -5.63443303e-01 -4.78693932e-01 -5.29636025e-01 1.44139910e+00 -2.59346128e-01 -1.65295815e+00 9.71794367e-01 -1.42069459e-01 -1.97814897e-01 6.14786744e-01 -1.28011424e-02 -8.31891418e-01 2.42498964e-01 -1.89849049e-01 5.76383233e-01 9.28263724e-01 -1.35132408e+00 -7.59897232e-01 -1.67359829e-01 -1.23336017e-01 2.66287565e-01 -1.25510961e-01 1.89734384e-01 2.11616129e-01 -1.12999000e-01 3.27147305e-01 -7.20919967e-01 -2.53068686e-01 -5.32041907e-01 -1.46512717e-01 -7.10518599e-01 7.70373225e-01 -4.51523662e-01 1.83705616e+00 -1.67618537e+00 -2.06720144e-01 3.98590982e-01 3.57815564e-01 -1.24705513e-03 2.40583986e-01 1.08380008e+00 -2.76906848e-01 7.32199550e-01 4.11965609e-01 -1.10722095e-01 2.24991128e-01 4.85861301e-01 -4.16602850e-01 3.05609167e-01 -1.29282743e-01 1.15307164e+00 -1.16605783e+00 -5.23096442e-01 4.11106765e-01 9.42391157e-02 -4.58732933e-01 4.79237735e-01 2.99364805e-01 1.47170946e-01 -3.56553018e-01 1.39399457e+00 1.15709066e+00 -1.31984919e-01 1.45911396e-01 5.30878425e-01 -3.79135728e-01 3.55090618e-01 -6.96863770e-01 1.05554795e+00 7.08333924e-02 5.00173986e-01 1.02364108e-01 -7.94621468e-01 3.33247900e-01 8.61540735e-01 4.37155962e-01 -9.67555881e-01 -2.26398129e-02 6.92409754e-01 1.12803578e-01 -6.07703328e-01 7.58228302e-01 3.94563079e-02 -3.64872098e-01 6.40070081e-01 -2.37588286e-01 -6.59476995e-01 9.64643434e-02 3.08100313e-01 6.22585893e-01 -5.18246442e-02 5.62923312e-01 -1.05613089e+00 7.86340594e-01 -1.82965681e-01 -1.81299388e-01 1.03415680e+00 -3.09923887e-01 3.19085121e-01 2.99841821e-01 -6.65102363e-01 -6.45341039e-01 -1.12307119e+00 -4.89381433e-01 1.30636716e+00 3.24267983e-01 -4.39044595e-01 -9.54439282e-01 -2.49762803e-01 1.77620783e-01 6.89606130e-01 -5.90509653e-01 3.84124845e-01 -5.03739953e-01 -8.32535863e-01 7.39044368e-01 3.45434904e-01 -5.07752821e-02 -1.33414865e+00 -6.58416986e-01 1.25490099e-01 -2.20292807e-01 -6.63697243e-01 -6.23428151e-02 4.48765576e-01 -1.35989368e+00 -5.18594682e-01 -6.66252747e-02 -8.20914626e-01 5.87345481e-01 2.46782884e-01 1.27047324e+00 5.39230824e-01 -2.31483161e-01 4.26904231e-01 -1.21292919e-01 -4.95818377e-01 -4.59671497e-01 -8.00336525e-02 5.28869390e-01 -5.87835789e-01 5.19427478e-01 -2.50617653e-01 -7.29350567e-01 5.37953973e-01 -6.88540697e-01 1.62748516e-01 1.79803044e-01 1.04410267e+00 1.35816500e-01 -9.34035778e-02 1.22507080e-01 -6.38007045e-01 8.72274399e-01 -1.69219792e-01 -3.78732830e-01 5.77745810e-02 -6.77108407e-01 -3.74140263e-01 3.21430594e-01 -3.25342178e-01 -1.01981449e+00 -4.87835288e-01 -9.82677937e-02 2.45538145e-01 1.11353043e-02 -1.46784872e-01 6.47139177e-02 -5.24923325e-01 8.02199244e-01 9.25758183e-02 1.99174434e-02 -6.80815242e-03 3.01039815e-01 7.09525108e-01 -6.82967342e-03 -6.68678164e-01 8.44880998e-01 4.91470337e-01 7.98524171e-02 -9.57177758e-01 -1.52186140e-01 -2.60129690e-01 -9.51962709e-01 -6.54426932e-01 6.56643391e-01 -6.78531289e-01 -9.10833478e-01 3.91110867e-01 -9.38691139e-01 -3.38627815e-01 -3.91645581e-01 4.25431967e-01 -1.01278400e+00 2.75717527e-02 -3.90154392e-01 -1.27895141e+00 -5.10977268e-01 -1.02017939e+00 9.43384409e-01 5.30070923e-02 -5.10597289e-01 -1.26927447e+00 5.87685481e-02 2.71537274e-01 1.81734428e-01 -1.73075795e-01 6.90226793e-01 -2.38256708e-01 -4.24233019e-01 -1.53791070e-01 2.34436691e-02 -1.39755070e-01 1.70832314e-02 4.95917559e-01 -9.81751978e-01 -5.31145096e-01 6.65065646e-02 -1.92070693e-01 -1.08835101e-01 6.52520418e-01 5.91872573e-01 -2.29931593e-01 -8.56000841e-01 5.40386558e-01 1.38545322e+00 3.85070026e-01 5.32770038e-01 7.28214979e-01 1.41836226e-01 5.53460240e-01 9.17806149e-01 4.63203549e-01 1.30579369e-02 3.28798652e-01 2.40537539e-01 1.49327129e-01 1.11720070e-01 -1.54819340e-01 3.77893507e-01 1.16112018e+00 -8.18235934e-01 -2.69281328e-01 -5.07867396e-01 4.42987174e-01 -1.72482407e+00 -1.40330648e+00 -4.32368398e-01 6.90478683e-01 6.25676990e-01 1.56016424e-01 -1.48347050e-01 3.35214496e-01 4.99015123e-01 -2.03574806e-01 -1.19133167e-01 -1.06291151e+00 -1.43546045e-01 3.15233678e-01 7.37729073e-01 1.00061214e+00 -7.20721722e-01 1.03317809e+00 1.29781246e+01 1.02230716e+00 2.21112028e-01 1.03134915e-01 5.16071796e-01 3.48020852e-01 -4.36954498e-01 -4.56139445e-02 -1.04416132e+00 2.72933897e-02 1.38140702e+00 -4.30666685e-01 6.85999811e-01 5.44219851e-01 3.44648361e-01 -4.23268199e-01 -1.26188684e+00 5.26221812e-01 9.73738134e-02 -1.40886843e+00 -2.83300440e-04 6.85225725e-01 7.73699820e-01 -5.08050561e-01 6.22419357e-01 3.24184299e-01 6.09259963e-01 -1.14389277e+00 8.60300779e-01 2.53660440e-01 1.03040910e+00 -6.05088234e-01 5.67372203e-01 1.68872893e-01 -1.14389896e+00 -2.20873043e-01 -8.77727985e-01 -1.00755692e+00 3.93533185e-02 -1.81779593e-01 -4.29956943e-01 3.48861217e-01 9.58353162e-01 2.99398601e-01 -3.93658698e-01 9.95779395e-01 -4.78694476e-02 1.04875881e-02 -2.71853864e-01 -4.48467314e-01 4.83122796e-01 -3.54241252e-01 4.66730654e-01 1.00164843e+00 2.48499006e-01 3.51035744e-01 -9.84472036e-02 4.01770771e-01 5.45058846e-01 3.29446048e-02 -1.19659424e+00 -1.78908288e-01 2.83276141e-01 9.16795909e-01 -4.83487815e-01 -4.22520459e-01 -2.00212970e-01 8.62069130e-01 -3.55488248e-02 5.01107454e-01 -6.11489356e-01 -4.35615242e-01 9.72222984e-01 -1.27327025e-01 -1.14700586e-01 -3.48497719e-01 -6.23769283e-01 -7.30352640e-01 -5.89872956e-01 -4.54965204e-01 5.93606755e-02 -5.53365827e-01 -1.39813089e+00 5.79277515e-01 -2.27688253e-02 -1.40553558e+00 -6.99901402e-01 -1.27676582e+00 -4.76714373e-01 4.92853165e-01 -1.11898029e+00 -1.10984349e+00 2.50124663e-01 4.52870727e-01 1.64141744e-01 -5.34416080e-01 1.39563632e+00 3.57715860e-02 1.00637585e-01 9.24474537e-01 6.69434488e-01 -7.37814724e-01 5.56605101e-01 -1.27867436e+00 5.68737745e-01 -1.37897313e-01 -4.31265175e-01 9.05828118e-01 6.28349900e-01 -5.39804697e-01 -1.41196322e+00 -3.66917729e-01 1.08350635e+00 -9.83769417e-01 6.55218959e-01 -3.86345625e-01 4.23767231e-02 7.88592756e-01 7.15902448e-01 -6.18741751e-01 8.21781039e-01 -1.83753878e-01 1.80774391e-01 5.75296998e-01 -1.39248300e+00 6.12354755e-01 1.66275799e+00 -4.63594139e-01 -6.25784039e-01 7.60327101e-01 8.13696027e-01 -6.94087505e-01 -1.30082703e+00 3.34633321e-01 8.65424156e-01 -8.75409484e-01 1.61978090e+00 -1.32660246e+00 -4.43697497e-02 2.81152606e-01 -2.61993498e-01 -9.32519078e-01 -5.96193194e-01 -1.23518765e+00 -5.33532679e-01 -5.83747849e-02 5.96577883e-01 -1.13057327e+00 3.42365682e-01 8.74560475e-01 -2.82833427e-01 -6.42737269e-01 -1.06996536e+00 -1.32016802e+00 -3.35779637e-02 -1.45572275e-01 4.90409225e-01 7.63798356e-01 6.83744550e-01 1.09839931e-01 -6.36873543e-02 -1.00294888e-01 5.33176839e-01 9.55312885e-03 4.41501856e-01 -1.34294486e+00 3.86843324e-01 -5.75816095e-01 -3.07655483e-01 -9.45992947e-01 -8.85957032e-02 -8.12076271e-01 -6.53862000e-01 -1.28511906e+00 -8.32044985e-03 -1.91056758e-01 -1.11109078e-01 -1.65725678e-01 3.67937148e-01 2.13746816e-01 1.20859891e-02 1.03788137e-01 -3.71160030e-01 6.18435517e-02 1.29639816e+00 6.91750320e-05 -1.62315920e-01 4.85058486e-01 -4.81304944e-01 7.84440815e-01 8.58408585e-02 -2.99253196e-01 -6.78878546e-01 6.11881316e-02 6.69384480e-01 4.61409837e-02 3.24159935e-02 -7.42885649e-01 5.37211418e-01 -3.75702560e-01 4.78586555e-01 -1.32223868e+00 1.30741090e-01 -9.61415648e-01 6.95283338e-02 9.40189242e-01 2.74610907e-01 1.20424610e-02 8.66204947e-02 5.39500564e-02 -1.44871444e-01 -5.70943117e-01 9.21121240e-01 -4.07591403e-01 -4.92852688e-01 -5.20386267e-03 -1.03226590e+00 8.97834301e-02 9.92593169e-01 -7.84614205e-01 -3.59281451e-01 -4.20183957e-01 -8.29068601e-01 -1.95836127e-02 6.50830388e-01 3.03609259e-02 7.30431557e-01 -1.51530886e+00 -2.30721906e-01 7.18729138e-01 -3.22939813e-01 -3.74400020e-01 -1.70157343e-01 6.58265352e-01 -1.32361674e+00 1.02442718e+00 -5.41665435e-01 -4.55340147e-01 -1.14228773e+00 4.78126436e-01 4.28307921e-01 -2.41845414e-01 -2.15481281e-01 1.11954463e+00 2.71224789e-02 -8.23025763e-01 1.85185194e-01 -9.21545625e-02 -7.54407048e-01 3.55081353e-03 6.88606799e-01 1.05194807e+00 -2.91290224e-01 -6.04341030e-01 -4.56784427e-01 6.65885091e-01 2.32151806e-01 -2.87484169e-01 9.17833567e-01 -2.19243199e-01 -9.89108324e-01 4.28274393e-01 8.38715494e-01 -1.36269778e-01 -3.69319022e-02 4.16855574e-01 1.25943512e-01 -8.02164078e-01 -4.32406247e-01 -3.55811834e-01 -1.85641110e-01 5.54822803e-01 5.34874737e-01 8.99602413e-01 8.57008278e-01 -3.02566767e-01 8.18335712e-01 9.66778398e-01 5.72402716e-01 -1.68019545e+00 -2.13140488e-01 6.89524531e-01 9.12339568e-01 -9.28350806e-01 5.44190466e-01 -7.27165341e-01 -4.14997995e-01 1.32979155e+00 4.68304873e-01 -1.55325383e-01 1.27306652e+00 5.74917436e-01 1.14069022e-02 -3.70670199e-01 -9.44949508e-01 1.12705544e-01 3.75366658e-01 1.11147714e+00 5.06513238e-01 5.07374525e-01 -9.66500878e-01 3.21953118e-01 -7.28706717e-01 -2.34555230e-01 4.90474731e-01 1.41972518e+00 -6.43810987e-01 -1.20391917e+00 -7.23931909e-01 4.61561680e-01 -5.49773455e-01 -1.16372630e-01 -5.28106689e-01 8.46754074e-01 -5.76629937e-02 1.49448860e+00 -1.97535474e-03 -5.03491640e-01 4.11356747e-01 1.41089618e-01 7.21762300e-01 -1.23501487e-01 -9.04846430e-01 4.15413082e-01 3.84890139e-01 -1.23056793e+00 -8.58632207e-01 -1.05834293e+00 -1.40667629e+00 -1.19437599e+00 -5.12782812e-01 1.89310342e-01 3.83317530e-01 3.90289724e-01 -2.06836104e-01 2.85260603e-02 9.80917513e-01 -1.07949340e+00 -3.90341938e-01 -9.39418614e-01 -1.00026262e+00 -7.84516707e-02 2.89751232e-01 -8.00943017e-01 -7.83523321e-01 2.83909619e-01]
[-7.480249881744385, 3.574787139892578]
b1169544-d7c4-49eb-97a8-c83591304670
unsupervised-semantic-scene-labeling-for
null
null
http://openaccess.thecvf.com/content_cvpr_2017/html/Wigness_Unsupervised_Semantic_Scene_CVPR_2017_paper.html
http://openaccess.thecvf.com/content_cvpr_2017/papers/Wigness_Unsupervised_Semantic_Scene_CVPR_2017_paper.pdf
Unsupervised Semantic Scene Labeling for Streaming Data
We introduce an unsupervised semantic scene labeling approach that continuously learns and adapts semantic models discovered within a data stream. While closely related to unsupervised video segmentation, our algorithm is not designed to be an early video processing strategy that produces coherent over-segmentations, but instead, to directly learn higher-level semantic concepts. This is achieved with an ensemble-based approach, where each learner clusters data from a local window in the data stream. Overlapping local windows are processed and encoded in a graph structure to create a label mapping across windows and reconcile the labelings to reduce unsupervised learning noise. Additionally, we iteratively learn a merging threshold criteria from observed data similarities to automatically determine the number of learned labels without human provided parameters. Experiments show that our approach semantically labels video streams with a high degree of accuracy, and achieves a better balance of under and over-segmentation entropy than existing video segmentation algorithms given similar numbers of label outputs.
['Maggie Wigness', 'John G. Rogers III']
2017-07-01
null
null
null
cvpr-2017-7
['scene-labeling']
['computer-vision']
[ 6.11432850e-01 2.06883967e-01 -4.88849461e-01 -7.31386364e-01 -6.34687603e-01 -6.68408990e-01 3.16346854e-01 5.62147081e-01 -3.37389916e-01 5.69707938e-02 1.65189758e-01 2.44897325e-02 -4.22164872e-02 -6.29800975e-01 -6.18852556e-01 -4.40280586e-01 -2.95908093e-01 5.97216010e-01 7.74601161e-01 3.64973634e-01 2.86429375e-01 1.32966474e-01 -1.84092283e+00 5.50132692e-01 8.64247799e-01 9.24996972e-01 2.18396604e-01 8.43798161e-01 -5.52839458e-01 1.18569398e+00 -4.17612165e-01 -3.25062275e-02 3.52544248e-01 -8.47090840e-01 -1.10446465e+00 9.54441726e-01 4.14167225e-01 9.36075766e-03 -3.88054438e-02 1.02861202e+00 -3.32039773e-01 3.15475851e-01 6.32325768e-01 -1.36280394e+00 1.55876964e-01 8.14227581e-01 -3.00834268e-01 2.15834245e-01 4.38429624e-01 -2.50652544e-02 9.08579230e-01 -4.92477864e-01 8.92279744e-01 1.04830515e+00 5.67944169e-01 4.58351374e-01 -1.21065223e+00 -4.72644717e-01 5.17704189e-01 -2.81615369e-02 -1.27944303e+00 -3.54776800e-01 6.73009396e-01 -6.38091207e-01 6.41548395e-01 1.62906051e-01 8.37137401e-01 5.73218405e-01 -2.90133864e-01 8.29186141e-01 6.49020672e-01 -5.45830250e-01 6.68123841e-01 7.79598728e-02 2.84246743e-01 8.41195583e-01 4.51830309e-03 -3.61668885e-01 -6.97183371e-01 7.99398646e-02 7.18298376e-01 4.53270674e-02 1.77315660e-02 -8.07474375e-01 -1.07782030e+00 6.67743206e-01 7.79686049e-02 4.56034690e-01 -3.13188136e-01 1.89428449e-01 4.69690830e-01 2.87625730e-01 5.02992570e-01 4.79677379e-01 -4.97326761e-01 -1.37879819e-01 -1.44281471e+00 1.12391375e-02 9.10519958e-01 1.23592222e+00 1.23249078e+00 3.09506077e-02 6.04446046e-02 5.35083950e-01 3.36038083e-01 5.04001416e-02 3.88334990e-01 -1.58039856e+00 2.41039395e-01 9.84739125e-01 -3.27592969e-01 -8.94762933e-01 -2.21166655e-01 5.48173152e-02 -1.26966223e-01 7.43301436e-02 2.44942620e-01 -5.55933006e-02 -1.27901292e+00 1.56283903e+00 3.32081348e-01 7.13476241e-01 5.03584035e-02 6.83217525e-01 4.91427422e-01 5.80805659e-01 6.54736638e-01 -5.06062150e-01 1.10264838e+00 -1.04003990e+00 -5.61673939e-01 -2.58706778e-01 6.45011723e-01 -4.47464615e-01 7.02083111e-01 3.84830475e-01 -9.76689041e-01 -7.46383846e-01 -9.81854320e-01 2.85826445e-01 -3.53906959e-01 -4.16538000e-01 5.25966465e-01 6.07921720e-01 -1.23843133e+00 6.20148599e-01 -8.97673666e-01 -8.28113616e-01 6.06696010e-01 2.58805811e-01 -1.24403369e-02 1.56688362e-01 -6.45017982e-01 2.05650702e-01 1.03283405e+00 -6.34325206e-01 -1.12030721e+00 -5.95893860e-01 -1.04232275e+00 -1.10901833e-01 7.94157922e-01 -4.88092989e-01 1.29606116e+00 -1.82532704e+00 -1.24540174e+00 9.36809361e-01 -3.36228579e-01 -6.72274292e-01 2.61439949e-01 -9.23303813e-02 -2.48183981e-01 6.68524265e-01 1.09535858e-01 1.16207802e+00 1.12377131e+00 -1.73448145e+00 -1.17484438e+00 -1.61363110e-01 -1.39997572e-01 5.11268139e-01 -2.86902070e-01 1.62497714e-01 -7.88297296e-01 -6.90275788e-01 3.66903484e-01 -6.74340427e-01 -5.78230143e-01 -3.08554858e-01 -4.45105396e-02 -1.87781706e-01 1.08209765e+00 -3.28279197e-01 1.34939754e+00 -2.04517865e+00 1.80411011e-01 5.77760875e-01 3.51693779e-01 -1.23634048e-01 2.10299976e-02 2.72493750e-01 7.99984187e-02 2.48618409e-01 -5.66909492e-01 -3.36119354e-01 -3.80092442e-01 4.62377131e-01 -1.68813244e-01 7.79097676e-02 5.42724840e-02 6.15109205e-01 -1.31317532e+00 -9.39265549e-01 3.26940626e-01 -4.80355732e-02 -6.35629117e-01 4.77737665e-01 -6.53122306e-01 3.54037106e-01 -4.06279176e-01 5.98706841e-01 1.89648107e-01 -3.17946076e-01 5.18998086e-01 5.98043464e-02 1.86711941e-02 -1.22108705e-01 -1.24973643e+00 2.06818175e+00 1.76762909e-01 6.03440106e-01 -2.31832460e-01 -1.40667498e+00 8.29466462e-01 3.07734370e-01 1.16905630e+00 -3.29834253e-01 1.94044605e-01 -1.64984450e-01 -5.48922598e-01 -8.49572241e-01 5.47354221e-01 1.84474513e-01 -2.10629091e-01 5.75225890e-01 5.90649366e-01 -1.44422024e-01 5.43107331e-01 4.97384459e-01 9.66424704e-01 4.02092904e-01 9.96298343e-02 -3.13221097e-01 4.12447274e-01 5.11216223e-01 7.07587719e-01 8.21445227e-01 -1.74971208e-01 7.25808322e-01 3.48924130e-01 -4.07748342e-01 -1.01402211e+00 -1.13231456e+00 2.86044240e-01 1.47003734e+00 6.09281480e-01 -7.63558984e-01 -1.43072975e+00 -9.49051559e-01 -3.76314670e-01 6.09819591e-01 -3.36815715e-01 9.79689695e-03 -4.84568983e-01 -3.08336169e-01 2.84211069e-01 5.46270549e-01 3.63088727e-01 -1.14741874e+00 -1.00145411e+00 2.92495757e-01 -2.70389467e-01 -1.26890218e+00 -4.68712926e-01 3.83778065e-01 -1.10041201e+00 -1.47413385e+00 -3.92476171e-02 -1.18514132e+00 9.28870797e-01 3.72648686e-01 1.23787045e+00 1.60856068e-01 -1.15528077e-01 8.84666920e-01 -7.27507412e-01 -2.71686137e-01 -6.51940286e-01 5.11189029e-02 -1.52640834e-01 2.87792057e-01 8.21858287e-01 -4.03377771e-01 -3.42092901e-01 3.04839909e-01 -1.16891456e+00 1.85503215e-01 7.03970045e-02 1.55746222e-01 7.25695252e-01 3.18955451e-01 3.63440305e-01 -1.03893912e+00 3.23420286e-01 -5.03910601e-01 -4.69693273e-01 3.77578616e-01 -5.88937402e-01 -3.53056230e-02 4.34601873e-01 -4.63387161e-01 -1.07127464e+00 6.19322896e-01 4.57921565e-01 -5.86845040e-01 -7.26529002e-01 2.20320895e-01 -8.63067657e-02 2.68461823e-01 5.41841030e-01 1.60092101e-01 -3.96847650e-02 -2.43245244e-01 6.06665254e-01 4.75522995e-01 6.71415687e-01 -5.31070769e-01 6.18230462e-01 5.98117471e-01 -5.71096718e-01 -7.90302217e-01 -7.15008914e-01 -8.60754013e-01 -1.12934101e+00 -8.26144993e-01 1.28386462e+00 -9.18417871e-01 -1.29569367e-01 3.17676455e-01 -7.06164956e-01 -6.82425022e-01 -6.97649896e-01 3.40328276e-01 -7.76030362e-01 3.49010974e-01 -4.57732737e-01 -7.18400896e-01 3.57517377e-02 -8.98945332e-01 9.03498948e-01 5.07802069e-01 -6.57068253e-01 -1.04403210e+00 -5.26361056e-02 2.52176285e-01 -1.78535298e-01 2.84122884e-01 5.53401113e-01 -9.47961450e-01 -6.91931784e-01 5.75809591e-02 9.09076724e-03 3.14990550e-01 2.60843486e-01 1.95347622e-01 -8.29638124e-01 -1.55758709e-01 -4.86062616e-02 -2.48792812e-01 8.16575587e-01 3.82619023e-01 1.17115629e+00 -2.49359354e-01 -4.79359627e-01 4.44671452e-01 1.42662144e+00 7.24265158e-01 3.79615903e-01 2.34019652e-01 8.49923015e-01 8.33185434e-01 5.59303284e-01 4.35742915e-01 3.83483589e-01 2.33903781e-01 1.47857040e-01 -1.27162144e-01 -9.41201672e-02 -5.51437080e-01 3.06249857e-01 8.16182315e-01 1.53960034e-01 -2.29189992e-01 -9.56711471e-01 6.51365101e-01 -2.12455821e+00 -9.43769932e-01 3.48865502e-02 2.10127616e+00 6.24778569e-01 4.01857376e-01 4.93585080e-01 1.77766755e-01 8.81684124e-01 1.87449828e-01 -4.42306697e-01 -2.60842413e-01 1.15185238e-01 1.71130329e-01 5.83903491e-01 4.38393980e-01 -1.36581492e+00 1.36160350e+00 6.97491789e+00 5.28350472e-01 -6.55707955e-01 1.09751984e-01 7.29735553e-01 1.13521568e-01 -2.68351555e-01 4.42377746e-01 -3.39608103e-01 3.86204958e-01 8.12609613e-01 -2.44744029e-02 4.12771881e-01 7.81430960e-01 1.86833397e-01 -3.98430198e-01 -1.30949438e+00 8.14157724e-01 3.19827646e-01 -1.29111993e+00 2.55622655e-01 -2.38912001e-01 1.04612219e+00 -1.80135727e-01 -4.41267550e-01 -7.86569864e-02 6.21215880e-01 -7.71785319e-01 9.61842716e-01 5.66613138e-01 3.53999466e-01 -6.65357113e-01 2.66601950e-01 3.75726193e-01 -1.51938462e+00 -1.85737997e-01 -2.89921016e-02 4.11760509e-02 2.27233201e-01 2.27756724e-01 -7.74497330e-01 3.55507880e-01 7.69421279e-01 9.70935881e-01 -6.34916484e-01 1.08364582e+00 -1.35800138e-01 7.52138734e-01 -2.53095597e-01 2.23004609e-01 3.55673403e-01 -2.10395187e-01 4.27565455e-01 1.54916608e+00 -9.50011611e-03 1.81402326e-01 8.58098924e-01 4.60798889e-01 -4.05509993e-02 2.07599640e-01 -4.92567241e-01 4.39305268e-02 6.94709718e-01 1.09982228e+00 -1.52606869e+00 -8.91030908e-01 -3.08225721e-01 9.87749338e-01 -5.79898469e-02 6.16871536e-01 -7.48347878e-01 4.71859146e-03 4.33437526e-01 1.46225989e-01 2.84695059e-01 -1.50927827e-01 -2.72050649e-01 -1.10838735e+00 -2.69845843e-01 -7.34193861e-01 8.52827966e-01 -5.80372453e-01 -8.83479178e-01 4.44854826e-01 2.10664168e-01 -1.41132319e+00 -3.35903585e-01 -1.00980997e-01 -5.36514342e-01 -8.27651918e-02 -1.06001496e+00 -7.81329453e-01 -5.45196056e-01 6.59557700e-01 1.11877012e+00 -1.13484174e-01 5.65040112e-01 -8.76623839e-02 -3.35253179e-01 1.68213755e-01 -2.41093591e-01 2.85610676e-01 3.90965223e-01 -1.34153187e+00 1.37569860e-01 1.09733903e+00 5.78463495e-01 1.63564339e-01 6.82098746e-01 -7.34817863e-01 -7.58395135e-01 -1.29625952e+00 6.90474153e-01 -4.24484253e-01 4.11311060e-01 -4.31915075e-01 -1.05812645e+00 7.47231841e-01 3.96116436e-01 -1.85968652e-01 7.92307496e-01 -2.21704155e-01 -3.11279625e-01 -1.57530621e-01 -9.89413619e-01 4.34584826e-01 1.35199726e+00 -4.59403038e-01 -7.95079112e-01 2.04761818e-01 8.20894241e-01 -2.80118853e-01 -7.52886891e-01 4.05631810e-01 1.68451861e-01 -1.00741541e+00 6.94787145e-01 -6.18794560e-01 2.78504491e-01 -4.57147419e-01 -3.22274142e-03 -8.59972596e-01 -5.20388186e-02 -5.65837622e-01 1.54351564e-02 1.29121006e+00 3.29816431e-01 -2.99634449e-02 1.18836904e+00 5.36250830e-01 -7.63594061e-02 -3.43919575e-01 -5.05506337e-01 -6.54800355e-01 -5.55912912e-01 -8.75883758e-01 2.87523329e-01 1.07916951e+00 2.60656059e-01 1.28310353e-01 -1.14308428e-02 1.86476037e-01 9.22547042e-01 4.23134379e-02 7.16376901e-01 -1.31447792e+00 5.08764721e-02 -4.95368540e-01 -5.31411290e-01 -1.09947598e+00 3.11805099e-01 -8.94626796e-01 3.12586814e-01 -1.46389997e+00 1.69501349e-01 -4.21131313e-01 -5.16542137e-01 5.10703087e-01 4.80727069e-02 2.29618534e-01 1.96770132e-01 4.39250857e-01 -1.45161176e+00 1.23443184e-02 5.90616584e-01 -2.93625109e-02 -6.44632757e-01 -2.88155377e-01 -5.01943231e-01 9.91139054e-01 6.67694271e-01 -6.32650614e-01 -8.98055553e-01 -2.90958732e-01 -2.73805171e-01 -3.48798297e-02 1.05984733e-01 -1.38203716e+00 5.30770421e-01 -2.66586810e-01 3.84193659e-01 -4.23922330e-01 -2.87491232e-01 -1.01495135e+00 2.23593384e-01 1.33906320e-01 -6.30459726e-01 -2.72837192e-01 -1.47670984e-01 6.39077306e-01 -2.90013552e-01 -2.27838740e-01 6.90352499e-01 -3.75340581e-01 -1.48478603e+00 2.76938826e-01 -9.21663225e-01 2.15201125e-01 1.43885136e+00 -8.29688370e-01 3.12575042e-01 -2.47622058e-01 -9.05893505e-01 5.11096358e-01 7.55842149e-01 7.63145506e-01 5.94545305e-01 -1.06282794e+00 -2.88875401e-01 3.04562032e-01 2.42122486e-01 3.23905170e-01 4.15014662e-02 1.94762275e-01 -5.88581979e-01 -5.59955128e-02 9.67865437e-02 -1.05740070e+00 -1.18821371e+00 6.07348144e-01 1.33002445e-01 8.22952297e-03 -7.55938649e-01 7.38675475e-01 8.60428140e-02 4.44116369e-02 6.23969018e-01 -2.94002235e-01 -3.82685781e-01 3.69698018e-01 4.37513798e-01 2.13687569e-01 -2.90800959e-01 -7.48402953e-01 -1.75531432e-01 6.96244061e-01 1.21915795e-01 -7.29088783e-02 1.10807335e+00 -4.93507624e-01 5.80074266e-02 6.72500849e-01 1.06991816e+00 -4.31113482e-01 -1.77328575e+00 -3.35197896e-01 6.89291060e-01 -3.73927087e-01 -3.33373159e-01 -3.18231791e-01 -1.08683228e+00 1.64219961e-01 4.27567929e-01 4.52585787e-01 1.43674481e+00 3.84048432e-01 6.70846462e-01 7.71325380e-02 4.01208371e-01 -1.58863401e+00 5.19666016e-01 2.98819691e-01 -3.00950352e-02 -1.09823239e+00 -1.25591189e-01 -5.44169128e-01 -8.16918552e-01 1.06221724e+00 7.27127075e-01 -1.40098304e-01 5.51601231e-01 4.86564070e-01 3.22789818e-01 -2.95059919e-01 -4.97163534e-01 -5.11115134e-01 1.28723621e-01 7.14191675e-01 -2.16387976e-02 -1.18107878e-01 -3.21807563e-02 2.15789542e-01 1.59723029e-01 4.20411751e-02 3.48591417e-01 1.11935127e+00 -1.01388454e+00 -9.68804717e-01 -1.42138466e-01 4.25814241e-01 -2.23840520e-01 2.54422784e-01 -4.17447060e-01 5.00004172e-01 4.64943498e-01 1.05610812e+00 5.18092871e-01 -5.83961010e-01 1.81130797e-01 4.91482109e-01 1.80583000e-01 -8.65071237e-01 -4.41573173e-01 3.75550777e-01 -2.08099201e-01 -7.31095016e-01 -1.15311503e+00 -7.24211872e-01 -1.75232148e+00 2.93813854e-01 -4.38723676e-02 4.56494719e-01 3.36600602e-01 1.12671411e+00 6.46148846e-02 5.06447673e-01 7.43797660e-01 -7.58262396e-01 2.64578402e-01 -4.80237722e-01 -4.08718288e-01 8.75789165e-01 3.95445600e-02 -3.23547214e-01 -1.42574936e-01 1.02814770e+00]
[9.00516128540039, 0.008007034659385681]
1a88641f-658b-49ac-b3ef-f445ad3db2c5
an-evolution-kernel-method-for-graph
2306.14688
null
https://arxiv.org/abs/2306.14688v1
https://arxiv.org/pdf/2306.14688v1.pdf
An Evolution Kernel Method for Graph Classification through Heat Diffusion Dynamics
Autonomous individuals establish a structural complex system through pairwise connections and interactions. Notably, the evolution reflects the dynamic nature of each complex system since it recodes a series of temporal changes from the past, the present into the future. Different systems follow distinct evolutionary trajectories, which can serve as distinguishing traits for system classification. However, modeling a complex system's evolution is challenging for the graph model because the graph is typically a snapshot of the static status of a system, and thereby hard to manifest the long-term evolutionary traits of a system entirely. To address this challenge, we suggest utilizing a heat-driven method to generate temporal graph augmentation. This approach incorporates the physics-based heat kernel and DropNode technique to transform each static graph into a sequence of temporal ones. This approach effectively describes the evolutional behaviours of the system, including the retention or disappearance of elements at each time point based on the distributed heat on each node. Additionally, we propose a dynamic time-wrapping distance GDTW to quantitatively measure the distance between pairwise evolutionary systems through optimal matching. The resulting approach, called the Evolution Kernel method, has been successfully applied to classification problems in real-world structural graph datasets. The results yield significant improvements in supervised classification accuracy over a series of baseline methods.
['Zhiming Zheng', 'Wei Wei', 'Dan Sun', 'Xue Liu']
2023-06-26
null
null
null
null
['graph-classification']
['graphs']
[ 1.64718270e-01 -1.89598814e-01 2.01272771e-01 1.13563143e-01 3.25220436e-01 -7.62374163e-01 9.38741982e-01 5.24027288e-01 2.25675609e-02 6.17619812e-01 -2.31301114e-01 -1.65668547e-01 -4.30984557e-01 -1.13412893e+00 -4.25699323e-01 -1.07768738e+00 -5.08347273e-01 4.92503971e-01 2.82971591e-01 -6.11870527e-01 1.70267656e-01 5.44459641e-01 -1.35057509e+00 -3.42241108e-01 9.25915420e-01 5.83157241e-01 -8.16948488e-02 7.38083959e-01 -2.12998763e-02 3.01901877e-01 -5.10606050e-01 -3.17147344e-01 2.25830227e-01 -7.59006143e-01 -6.10063434e-01 2.57381871e-02 2.49101818e-01 4.32160139e-01 -6.98733509e-01 9.51683640e-01 1.92689657e-01 2.54761517e-01 7.32358634e-01 -1.43199325e+00 -8.53006124e-01 4.38327461e-01 -5.73926508e-01 2.33478665e-01 3.51269484e-01 2.67763883e-01 1.13017058e+00 -3.25171590e-01 7.35504270e-01 1.09016871e+00 8.80004883e-01 2.16151163e-01 -1.61458838e+00 -2.93350101e-01 1.89998955e-01 1.78561077e-01 -1.36639130e+00 -6.17873482e-02 1.12678182e+00 -7.82541335e-01 6.15961373e-01 2.58344024e-01 1.37625635e+00 7.17167675e-01 7.69416511e-01 2.21595094e-01 1.03822148e+00 -1.81868657e-01 3.33767295e-01 -1.88591152e-01 2.89785057e-01 9.61842477e-01 2.01878220e-01 -3.36108692e-02 -4.39342856e-01 -4.17391956e-01 3.53132546e-01 2.13550240e-01 -3.55687022e-01 -4.45057958e-01 -9.19516861e-01 2.86670297e-01 7.31639266e-01 3.81872207e-01 -1.22500435e-01 2.04551339e-01 1.20509930e-01 6.60911202e-01 5.67774773e-01 4.68371361e-01 -2.68224031e-01 -9.72995441e-03 -7.70592928e-01 2.87425458e-01 9.21682179e-01 4.35600877e-01 1.03936255e+00 -1.42394185e-01 8.28275532e-02 4.41731423e-01 2.19179109e-01 3.65705639e-01 3.43524337e-01 -4.30996120e-01 -1.00844212e-01 1.24879324e+00 -3.38747531e-01 -1.37993753e+00 -3.07769537e-01 -6.63867533e-01 -1.08486521e+00 1.24207608e-01 3.57035220e-01 5.85805476e-02 -7.07180083e-01 1.99553204e+00 6.41654730e-01 3.16825956e-01 -2.23236531e-01 4.45342511e-01 3.09480250e-01 7.51513779e-01 -2.29551643e-01 -2.59773195e-01 8.91052365e-01 -6.23665571e-01 -4.34608489e-01 1.66877538e-01 4.16714430e-01 -3.23745191e-01 8.02219272e-01 -9.92998257e-02 -9.61309791e-01 -4.25359786e-01 -1.07790220e+00 5.16458452e-01 -4.75928634e-01 -4.95398700e-01 5.99953830e-01 4.78899151e-01 -1.22291708e+00 1.00497019e+00 -1.05123031e+00 -6.77461445e-01 -2.04352569e-02 1.51410177e-01 -2.24440724e-01 2.83578336e-01 -1.14279735e+00 5.38569808e-01 1.31126314e-01 1.97537884e-01 -7.79030383e-01 -8.51135850e-01 -6.39007449e-01 -9.08297207e-03 3.01654279e-01 -7.01191247e-01 6.22080624e-01 -7.61643767e-01 -1.22952366e+00 5.05841434e-01 -2.94691045e-02 -2.62661581e-03 5.60007036e-01 3.86060625e-01 -4.33927923e-01 -3.00079882e-01 -2.63416171e-01 8.74377191e-02 8.45189989e-01 -1.07437432e+00 -1.33597896e-01 -6.08298242e-01 8.20409879e-03 1.65628642e-01 -5.86708546e-01 -5.04311562e-01 -3.37775350e-01 -5.67386031e-01 2.06026673e-01 -1.41706789e+00 -2.70205796e-01 -1.96782630e-02 -3.65915209e-01 -1.07719161e-01 8.84024739e-01 -3.64180863e-01 1.45002723e+00 -2.04765105e+00 8.36029232e-01 2.92247653e-01 6.58859372e-01 -1.63748115e-02 6.48874789e-02 1.12458920e+00 -7.41042793e-02 2.37186715e-01 -6.83060408e-01 -1.33944377e-01 -2.25646764e-01 1.70916021e-01 -9.51964781e-02 5.84634840e-01 1.19819731e-01 1.05231345e+00 -9.72513735e-01 -3.41304600e-01 -1.16006352e-01 3.92885834e-01 -1.46082163e-01 8.60177577e-02 -1.20902210e-02 4.45120841e-01 -3.81111503e-01 3.32879394e-01 5.27715206e-01 -3.70360643e-01 2.15559214e-01 1.44157588e-01 -1.28884897e-01 -3.11870694e-01 -7.43965626e-01 1.44559753e+00 -1.00450292e-01 6.13177955e-01 -1.92710340e-01 -1.13917565e+00 9.41409826e-01 4.68410552e-02 7.50027359e-01 -4.74014670e-01 1.48462597e-02 -6.41228482e-02 4.35097396e-01 -2.23174319e-01 4.71059829e-01 -2.29735319e-02 -2.03070074e-01 7.35137224e-01 -7.00460747e-02 -1.52290717e-01 2.94800162e-01 4.39170957e-01 1.61023366e+00 -2.07951084e-01 2.14981139e-01 -5.01698077e-01 6.08845651e-01 -4.67802174e-02 5.37658334e-01 5.37412524e-01 -2.36030906e-01 1.81592003e-01 6.23135746e-01 -4.64199334e-01 -1.03682411e+00 -1.16284573e+00 -1.22293942e-01 6.25772119e-01 4.39627975e-01 -3.80692571e-01 -7.74976671e-01 -4.57925022e-01 3.90653521e-01 7.55827799e-02 -1.03175235e+00 -8.89768958e-01 -5.73940694e-01 -9.75194037e-01 2.96946645e-01 3.96535397e-02 4.27770823e-01 -1.02571762e+00 -3.73351455e-01 2.77473927e-01 1.76030084e-01 -5.92576504e-01 -5.75447083e-01 -1.58488050e-01 -8.23119223e-01 -1.15979600e+00 -5.16762137e-01 -5.47184169e-01 7.63808131e-01 2.61805832e-01 8.72259736e-01 5.47894359e-01 -6.91736579e-01 5.48745632e-01 -2.57170051e-01 3.58632118e-01 -6.25244856e-01 1.01026617e-01 3.05260390e-01 3.06515515e-01 -2.84983039e-01 -9.91969824e-01 -7.46415377e-01 3.74890655e-01 -8.35339010e-01 1.57562923e-02 2.86549151e-01 8.66053522e-01 4.92179811e-01 5.80692410e-01 2.50773191e-01 -7.34149694e-01 9.77720082e-01 -6.79412007e-01 -2.69939423e-01 7.23496437e-01 -1.04813278e+00 1.68892518e-01 8.31871986e-01 -6.99637413e-01 -8.44118476e-01 -2.89944232e-01 5.56655049e-01 -7.79967085e-02 2.38981456e-01 7.53162444e-01 -1.40731204e-02 -3.48437160e-01 4.06770498e-01 5.67421496e-01 1.23992309e-01 -2.11351424e-01 4.76445287e-01 2.29595423e-01 4.96402979e-01 -6.82510257e-01 1.24510181e+00 3.87684971e-01 4.25532013e-01 -9.03414726e-01 -1.20130651e-01 -1.44698754e-01 -1.04420960e+00 -6.88788891e-01 3.89860183e-01 -2.22609356e-01 -6.96292698e-01 8.79795730e-01 -7.25096285e-01 -3.77153814e-01 -4.21634734e-01 -1.63166016e-01 -7.73435161e-02 4.82591152e-01 -5.20493686e-01 -7.85529375e-01 -3.13247859e-01 -6.84246719e-01 7.32597828e-01 3.66934508e-01 -1.04000017e-01 -1.37772310e+00 8.46237957e-01 -2.76207089e-01 4.35657144e-01 7.13419437e-01 1.20001733e+00 -1.54087678e-01 -4.93677348e-01 -3.97668749e-01 1.09776862e-01 -1.26292959e-01 4.48980540e-01 6.11669898e-01 -2.58436173e-01 -7.18715370e-01 -1.22561149e-01 2.26065174e-01 6.80740952e-01 6.53662458e-02 7.05523491e-01 -2.99258649e-01 -6.64329231e-01 5.30144691e-01 1.30303276e+00 2.59903818e-01 2.74022073e-01 -1.10274814e-01 7.94094443e-01 7.77687848e-01 1.34989634e-01 3.63528341e-01 3.41942757e-01 7.71300495e-01 1.21664681e-01 1.51693076e-02 -7.95605853e-02 -2.11668625e-01 4.11288857e-01 1.33416724e+00 -1.87841997e-01 -3.29752713e-01 -1.18801951e+00 3.38444918e-01 -1.98028052e+00 -9.25305545e-01 -4.13432807e-01 2.39622664e+00 5.30936718e-01 1.19842082e-01 9.35694501e-02 1.01830615e-02 8.41628790e-01 2.45435745e-01 -1.00402892e+00 -1.57516062e-01 -1.87255010e-01 -2.22698152e-02 8.26471895e-02 3.78962368e-01 -5.46349943e-01 6.47001684e-01 6.48940229e+00 4.61350739e-01 -1.14394581e+00 -2.63361018e-02 4.04324919e-01 2.34274000e-01 -4.40665156e-01 4.33097392e-01 -2.37774163e-01 6.19347572e-01 1.06268322e+00 -9.57963467e-01 6.59152031e-01 3.15348864e-01 5.12609407e-02 -4.72109355e-02 -1.01235163e+00 6.04975522e-01 -1.78291380e-01 -9.83481288e-01 -4.31683362e-02 4.20935124e-01 6.97706342e-01 -1.44552618e-01 2.43349910e-01 6.25111237e-02 3.37217897e-01 -6.18203282e-01 6.51097655e-01 9.14297938e-01 6.65387332e-01 -5.55735409e-01 2.03260481e-01 3.86492461e-01 -1.89822972e+00 2.61374246e-02 -7.28742555e-02 -1.97164163e-01 2.15969548e-01 7.67306864e-01 -6.29865706e-01 7.37446070e-01 4.90636289e-01 1.01189649e+00 -1.08105540e+00 9.10400271e-01 1.46058396e-01 6.26689851e-01 -2.26505950e-01 -1.35754794e-01 -2.07486749e-01 -9.02122676e-01 9.53753829e-01 7.07731426e-01 3.13049734e-01 -1.05994985e-01 5.61642647e-02 1.02496219e+00 1.75638739e-02 -2.24866539e-01 -8.28749835e-01 -4.73994344e-01 3.82660508e-01 1.43093610e+00 -1.14555871e+00 -6.55809045e-02 -6.32249787e-02 1.07372832e+00 5.22610605e-01 3.50713283e-01 -7.82033026e-01 -3.38342518e-01 7.50869155e-01 4.74139750e-02 -2.51279145e-01 -5.45214474e-01 1.85305357e-01 -1.20972383e+00 2.34327912e-01 -4.59805906e-01 1.87138095e-01 -3.38663310e-01 -1.34028363e+00 6.78990424e-01 -9.57972184e-02 -1.09345913e+00 9.99986306e-02 -1.90648451e-01 -9.93359923e-01 5.02887249e-01 -8.97423267e-01 -7.95802355e-01 -7.06439495e-01 4.16658074e-01 6.50796145e-02 -6.47063479e-02 6.57965660e-01 1.14321150e-01 -1.03986192e+00 5.09322286e-01 5.79048574e-01 -2.53868699e-01 3.65124345e-01 -1.52472138e+00 7.97745705e-01 8.10655534e-01 2.68695131e-02 6.99132144e-01 7.55784154e-01 -1.01052034e+00 -1.77087307e+00 -1.01295018e+00 3.87535691e-01 -4.79012877e-01 1.07646072e+00 -7.59036064e-01 -1.23647320e+00 3.39422077e-01 8.49696714e-03 4.10154611e-02 3.68701577e-01 1.36668989e-02 -1.96361288e-01 -2.08808661e-01 -9.48117197e-01 7.17096090e-01 1.46535516e+00 -5.94917953e-01 -1.82385877e-01 1.34886026e-01 7.12365389e-01 2.03608260e-01 -9.88866568e-01 4.68614906e-01 7.96896815e-01 -6.82271779e-01 6.53756499e-01 -6.28318787e-01 2.80652612e-01 -3.99654835e-01 2.92471826e-01 -1.58893573e+00 -5.39069176e-01 -9.12585199e-01 -3.22830796e-01 1.39509082e+00 3.84211212e-01 -9.84448671e-01 6.50423467e-01 5.24591982e-01 1.33377373e-01 -7.88444638e-01 -8.51471901e-01 -8.09060693e-01 8.80286023e-02 2.63336390e-01 6.56061232e-01 1.15052140e+00 5.64731732e-02 7.09915236e-02 -6.58922866e-02 4.64415625e-02 7.49822080e-01 3.05262119e-01 7.85332322e-01 -1.41062570e+00 -3.66284162e-01 -7.47658730e-01 -8.43685448e-01 -4.79942322e-01 1.02475509e-01 -1.12122524e+00 -1.66947559e-01 -1.10700870e+00 4.45251942e-01 -5.47575355e-01 -2.64289021e-01 2.30519031e-03 -3.27900261e-01 -1.47203907e-01 -9.94178951e-02 5.67053854e-01 -2.77185082e-01 9.35748518e-01 1.18959486e+00 -2.79274702e-01 -3.95694852e-01 -1.67806014e-01 -2.76277393e-01 2.50003755e-01 5.98400712e-01 -4.56373930e-01 -5.96547484e-01 9.07297209e-02 3.43551606e-01 7.58021548e-02 2.61036366e-01 -1.17341876e+00 4.51938748e-01 -1.47971109e-01 -3.42255197e-02 -2.03831539e-01 2.06356019e-01 -7.88450718e-01 9.57004607e-01 9.15492237e-01 -1.66191518e-01 4.09064680e-01 -9.05740932e-02 1.11989868e+00 -1.32057518e-01 2.34993994e-01 6.58909380e-01 1.23305112e-01 -3.26508284e-01 6.87585413e-01 -2.74879187e-01 7.51230074e-03 1.31409717e+00 -1.67644665e-01 -4.14138228e-01 -1.93804666e-01 -6.95872962e-01 3.19570869e-01 9.29017663e-01 3.34798843e-01 5.19017994e-01 -1.27989054e+00 -6.04264021e-01 1.71942219e-01 6.13118522e-02 -4.70216721e-01 3.04902256e-01 9.96386111e-01 -4.93750125e-01 -1.44843549e-01 -2.77263165e-01 -6.84411407e-01 -1.24702144e+00 4.32059944e-01 6.37296259e-01 -3.48444223e-01 -8.35200846e-01 6.11203492e-01 2.09895447e-01 -4.55186516e-01 -3.21083069e-01 4.62531298e-02 -1.42226413e-01 1.53410092e-01 3.58237624e-02 4.21878070e-01 -4.50626723e-02 -6.75432324e-01 -2.41060838e-01 8.74759734e-01 -3.50578763e-02 2.05493961e-02 1.31989813e+00 -2.35563010e-01 -5.84290862e-01 9.99874413e-01 1.08136177e+00 -2.02633455e-01 -1.15881300e+00 -3.22549433e-01 3.01788114e-02 -4.53247458e-01 -2.12535158e-01 -3.56043905e-01 -1.16501653e+00 4.88443643e-01 4.72480536e-01 9.11188662e-01 1.04081333e+00 6.47944510e-02 6.77634597e-01 2.95449704e-01 4.50903445e-01 -5.83095133e-01 8.59639421e-02 3.03122014e-01 7.60621727e-01 -7.41069794e-01 1.05406130e-02 -2.79949307e-01 -1.32079214e-01 1.02249002e+00 4.75444198e-01 -1.65783808e-01 1.02740419e+00 -7.34915808e-02 -5.69143891e-01 -4.57302064e-01 -9.09189463e-01 6.40689880e-02 4.05262232e-01 1.91062257e-01 1.36937156e-01 2.30864123e-01 -3.54938805e-01 -7.16484115e-02 -3.45063537e-01 -6.70192540e-01 4.88176018e-01 9.18730378e-01 -2.07616270e-01 -1.22973430e+00 8.87725949e-02 4.07098979e-01 2.62316436e-01 2.66793489e-01 -1.03494525e+00 7.70589888e-01 -1.49083957e-01 6.11252308e-01 7.57232159e-02 -6.99303925e-01 1.80404276e-01 1.26627654e-01 4.85148072e-01 -4.71050262e-01 -4.11233574e-01 -4.37391877e-01 -3.93837005e-01 -3.13514709e-01 -2.38986567e-01 -1.14085650e+00 -1.05723834e+00 -7.55191028e-01 -3.75987351e-01 1.88959435e-01 4.12660092e-01 7.75640011e-01 4.69448745e-01 6.84627414e-01 1.10292220e+00 -6.43212438e-01 -1.02442108e-01 -7.58539677e-01 -7.15701640e-01 7.49336481e-01 3.32712531e-01 -8.90395880e-01 -6.36073232e-01 -1.28161743e-01]
[7.1616010665893555, 5.862022876739502]
ca3066b6-464d-4cc3-b399-95850c50b7dd
deep-learning-based-segmentation-free-license
1912.02441
null
https://arxiv.org/abs/1912.02441v1
https://arxiv.org/pdf/1912.02441v1.pdf
Deep Learning Based Segmentation Free License Plate Recognition Using Roadway Surveillance Camera Images
Smart automated traffic enforcement solutions have been gaining popularity in recent years. These solutions are ubiquitously used for seat-belt violation detection, red-light violation detection and speed violation detection purposes. Highly accurate license plate recognition is an indispensable part of these systems. However, general license plate recognition systems require high resolution images for high performance. In this study, we propose a novel license plate recognition method for general roadway surveillance cameras. Proposed segmentation free license plate recognition algorithm utilizes deep learning based object detection techniques in the character detection and recognition process. Proposed method has been tested on 2000 images captured on a roadway.
['Yusuf Artan', 'Bensu Alkan', 'Alperen Elihos', 'Burak Balci']
2019-12-05
null
null
null
null
['license-plate-recognition']
['computer-vision']
[ 2.41110861e-01 -8.49584818e-01 -4.24327582e-01 -2.30811387e-01 -5.87549150e-01 -6.84507906e-01 4.47561771e-01 -4.54502732e-01 -6.06939733e-01 6.83436632e-01 -4.74770576e-01 -4.27982330e-01 2.18391791e-01 -7.54092872e-01 -3.11100096e-01 -5.36040962e-01 9.46880162e-01 2.45701775e-01 7.36564755e-01 -4.66458611e-02 7.37772942e-01 7.94709027e-01 -1.18996143e+00 1.12433821e-01 6.06512427e-01 8.32641006e-01 -2.44735032e-01 9.01511252e-01 1.22736245e-01 5.34354746e-01 -3.16029221e-01 -5.56652009e-01 5.07046103e-01 1.70312151e-01 1.76513661e-02 5.10870516e-01 8.69765878e-01 -8.44136715e-01 -6.37399852e-01 1.11888874e+00 2.98106343e-01 3.94136995e-01 8.98737788e-01 -1.01762950e+00 -3.32583278e-01 -5.95006764e-01 -8.28373432e-01 7.07502902e-01 6.24600090e-02 3.33295822e-01 4.10679311e-01 -7.48090804e-01 1.55504346e-01 5.28484762e-01 6.78018153e-01 5.90197682e-01 -5.10066748e-01 -9.89624619e-01 -5.54919481e-01 3.62022907e-01 -1.47878432e+00 -8.52455378e-01 6.84933245e-01 -6.01841807e-01 1.11794353e+00 2.24795803e-01 4.40817446e-01 4.76056784e-01 6.09451830e-01 9.13359821e-01 9.91412699e-01 -2.41598442e-01 -3.91262174e-02 4.20856357e-01 5.30069530e-01 9.74047184e-01 7.67013550e-01 -6.76289871e-02 1.82800636e-01 3.96005988e-01 1.12422335e+00 4.91317451e-01 3.32768828e-01 1.06927514e-01 -3.25087607e-01 5.90810299e-01 -2.70121157e-01 2.47310847e-01 -1.71672583e-01 -5.09520099e-02 3.55066836e-01 -2.47826710e-01 6.74231304e-03 -2.00420395e-02 1.83790401e-01 -5.49222410e-01 -1.21778166e+00 -7.55704427e-03 2.43545905e-01 9.08192933e-01 2.82555878e-01 5.03592014e-01 -5.03005832e-02 1.05554080e+00 4.07206714e-01 8.75398099e-01 2.82777011e-01 -5.24525225e-01 6.54718220e-01 4.46527958e-01 3.09323430e-01 -1.18693447e+00 -3.88994932e-01 -9.95207131e-02 -5.89926004e-01 4.86859500e-01 2.87026346e-01 -1.60699248e-01 -1.09500682e+00 3.14559549e-01 -7.71697238e-02 2.17693031e-01 -8.78422856e-02 8.04602563e-01 5.71085751e-01 8.94960940e-01 1.82940438e-02 3.37602973e-01 1.53682423e+00 -9.33822155e-01 -7.13492990e-01 -4.59276110e-01 1.98584963e-02 -9.62131917e-01 5.52467763e-01 6.05601430e-01 -8.11769366e-01 -6.84068203e-01 -1.41842759e+00 1.49831111e-02 -4.37816828e-01 5.62624097e-01 3.85291338e-01 1.45951998e+00 -4.35917675e-01 -2.80859113e-01 -7.19119608e-01 -3.90763789e-01 6.10578477e-01 8.23902905e-01 -3.05840284e-01 -2.14593917e-01 -6.85301602e-01 8.81018579e-01 -7.20565254e-03 2.09203020e-01 -3.55256379e-01 1.63584620e-01 -7.64232814e-01 5.32031730e-02 3.20097685e-01 -2.18472660e-01 1.12573445e+00 -7.58328795e-01 -1.66426563e+00 1.27419686e+00 7.92748258e-02 -7.56681189e-02 4.70821112e-01 -1.38377985e-02 -1.01206112e+00 1.52827485e-03 -3.31945926e-01 1.93823546e-01 1.06551421e+00 -5.73665917e-01 -1.00151050e+00 -5.29569030e-01 -3.69702637e-01 1.46422684e-01 2.22256199e-01 3.23919296e-01 -5.62663615e-01 -1.73644125e-01 -3.66187632e-01 -9.61448193e-01 3.14871728e-01 -4.01283085e-01 -4.84832674e-01 -1.55010089e-01 1.15261173e+00 -7.69555032e-01 7.67512739e-01 -2.16541886e+00 -1.20090842e+00 5.05497873e-01 -5.08790128e-02 1.01281607e+00 3.13928664e-01 -9.30276960e-02 3.78234178e-01 -2.37669095e-01 7.71156698e-03 -2.24944696e-01 7.77404830e-02 -1.00845315e-01 -1.07984878e-01 7.82108188e-01 1.24581121e-01 6.46515906e-01 -8.69345292e-03 -4.97587502e-01 9.32205319e-01 3.81691962e-01 -2.32321754e-01 -2.47611374e-01 5.78816891e-01 -3.47056147e-03 -4.53051776e-01 1.00601351e+00 1.15571749e+00 3.80155712e-01 -3.59659463e-01 1.45550678e-02 -4.46187943e-01 -3.99902880e-01 -1.02773607e+00 6.85738802e-01 -2.41021112e-01 1.41118360e+00 1.67719454e-01 -8.42044771e-01 1.05841613e+00 2.41737932e-01 1.48754656e-01 -8.88698041e-01 4.63361949e-01 3.28301996e-01 -1.47985011e-01 -8.42375278e-01 9.62491512e-01 -4.68062460e-01 -2.86552161e-02 -1.83993652e-01 -4.41233695e-01 -7.44713703e-03 7.46469870e-02 -5.68505406e-01 7.12537050e-01 -2.75883555e-01 2.51545131e-01 2.11444318e-01 9.62479591e-01 2.44786471e-01 5.25871634e-01 6.18294001e-01 -9.80570138e-01 4.07434314e-01 -9.54115540e-02 -6.60730600e-01 -1.31142783e+00 -9.20062184e-01 -1.94508284e-01 6.26568377e-01 2.41955817e-01 6.35641754e-01 -7.40447879e-01 -3.94150347e-01 -1.47020966e-01 5.40138781e-01 -1.57640092e-02 8.05633441e-02 -7.93385923e-01 -3.66751194e-01 9.46163774e-01 6.02346480e-01 1.00950491e+00 -9.66240764e-01 -5.61858296e-01 9.18940976e-02 4.06807959e-01 -1.40289879e+00 -7.06441760e-01 -4.18936759e-01 -5.22550523e-01 -1.08993542e+00 -6.82339251e-01 -1.20068610e+00 5.42848766e-01 6.13787711e-01 3.24112535e-01 -1.77377053e-02 -5.18545806e-01 3.70729417e-01 3.57482910e-01 -3.16685438e-01 -2.15102375e-01 -9.41582695e-02 1.43982753e-01 3.46701860e-01 8.62428486e-01 8.09591357e-03 -6.66482031e-01 3.64372462e-01 -8.15003872e-01 -2.26751283e-01 7.93345451e-01 2.13568777e-01 2.38432631e-01 2.56991833e-01 2.97605336e-01 -7.85952270e-01 7.89795220e-01 1.09745385e-02 -1.35135984e+00 1.82527989e-01 -2.41247252e-01 -5.47445834e-01 6.21728599e-01 2.46289700e-01 -1.12417972e+00 1.26658723e-01 -4.78010803e-01 -2.64623612e-01 -7.78445125e-01 -1.25994787e-01 -1.06442884e-01 -2.63292402e-01 9.05761197e-02 5.98509133e-01 -1.48212209e-01 -2.07551613e-01 -4.72034723e-01 1.19761455e+00 9.02829111e-01 6.73386082e-02 6.80180073e-01 4.80688542e-01 -9.32496712e-02 -1.65833199e+00 -1.32572474e-02 -1.15260136e+00 -5.47650278e-01 -5.26268184e-01 1.23536098e+00 -5.49772859e-01 -1.18590391e+00 9.26914155e-01 -9.67094302e-01 3.43314499e-01 5.09575427e-01 5.84584355e-01 -1.01727031e-01 3.78419966e-01 -3.67389828e-01 -1.24595118e+00 -2.65273988e-01 -1.11475682e+00 8.97920430e-01 7.31253684e-01 2.21123651e-01 -9.47574556e-01 3.66251200e-01 1.17590737e+00 4.83139098e-01 -1.99556425e-02 3.56138706e-01 -6.84406340e-01 -7.94745982e-01 -1.07538605e+00 -4.82430518e-01 5.37252784e-01 2.47224241e-01 3.47449690e-01 -9.52510893e-01 1.33810386e-01 -1.63464099e-01 2.13734329e-01 8.45698595e-01 7.55674005e-01 7.16061056e-01 -2.67343998e-01 -3.48694503e-01 4.03536379e-01 1.64612663e+00 8.24793100e-01 1.01516700e+00 4.97624516e-01 7.03627467e-01 1.67421997e-01 3.14212799e-01 2.87098199e-01 4.48010638e-02 5.43380439e-01 -1.30730435e-01 -1.30369946e-01 1.40678763e-01 7.19953999e-02 5.44736922e-01 6.26226604e-01 -9.28750411e-02 -3.21809173e-01 -1.19658720e+00 1.65759370e-01 -1.57567263e+00 -1.49964416e+00 -3.01794529e-01 2.10701799e+00 -6.83906451e-02 3.99159133e-01 2.17306033e-01 2.19772831e-01 1.09114635e+00 -1.01265065e-01 -4.18056607e-01 -8.50276947e-01 1.82302862e-01 8.70168656e-02 9.61731195e-01 6.67011201e-01 -1.47520125e+00 1.04427171e+00 5.88737631e+00 8.76498997e-01 -1.39665401e+00 1.00099109e-02 5.41492999e-01 1.23490199e-01 3.74164701e-01 -7.01029658e-01 -1.19942141e+00 3.77944738e-01 6.34432733e-01 3.88754070e-01 -5.23929531e-03 7.82094061e-01 4.84530479e-01 -4.35366809e-01 -5.18501341e-01 1.56484187e+00 2.60566026e-01 -1.57626641e+00 -2.06939295e-01 1.11927249e-01 5.32108307e-01 -9.95458066e-02 5.24079442e-01 3.90638769e-01 -3.42380375e-01 -8.36429834e-01 1.89605385e-01 6.52257085e-01 8.28605771e-01 -1.00869977e+00 8.79852533e-01 3.02269012e-01 -8.23645294e-01 7.45469481e-02 -3.21314216e-01 1.21195175e-01 -1.21628940e-01 -2.15096980e-01 -1.20563996e+00 -2.87282556e-01 -4.89436612e-02 5.45697868e-01 -2.24763870e-01 1.41760647e+00 3.37274909e-01 6.87733531e-01 -2.57026196e-01 -2.17988402e-01 6.69207811e-01 -5.59883237e-01 4.63257283e-01 1.61918247e+00 1.25104025e-01 2.28789911e-01 -4.21407558e-02 6.11302376e-01 -8.38988125e-02 -5.02240993e-02 -9.08499479e-01 -2.51501780e-02 -5.03320359e-02 1.22547579e+00 -1.01260149e+00 -2.78708160e-01 -7.67610133e-01 8.89049590e-01 -6.67035580e-01 1.24346793e-01 -1.20746291e+00 -7.64882386e-01 6.37860477e-01 6.15157187e-01 4.46798921e-01 -5.74110925e-01 -2.67632395e-01 -1.01804233e+00 -2.65555799e-01 -4.33079541e-01 2.25179285e-01 -5.93365490e-01 -8.07097137e-01 -2.29353234e-02 -3.16725254e-01 -1.41431963e+00 1.72946930e-01 -1.09890819e+00 -8.23923945e-01 5.19138455e-01 -1.55520749e+00 -1.05978119e+00 -3.07906568e-01 8.09313715e-01 1.15018201e+00 -7.21957207e-01 1.77499577e-01 6.85326397e-01 -1.20769215e+00 7.22293615e-01 6.65907502e-01 6.46087170e-01 4.37861532e-01 -6.28652275e-01 1.95487514e-01 1.15419245e+00 -1.97190002e-01 5.12706816e-01 2.78048366e-01 -6.16491616e-01 -1.35474408e+00 -9.21863139e-01 2.57735431e-01 -1.81831867e-01 1.45397246e-01 -1.65383548e-01 -4.64456856e-01 5.68805397e-01 2.37854674e-01 -1.87611226e-02 8.94959867e-01 -6.12116396e-01 1.67867571e-01 -5.20533562e-01 -1.45228159e+00 3.03199589e-01 4.00368590e-03 -3.73861372e-01 -4.42612678e-01 8.02768320e-02 -5.22622168e-01 -1.35175213e-01 -6.98874742e-02 -2.51393884e-01 9.91467595e-01 -7.17150807e-01 6.98032677e-01 -2.89783090e-01 -8.88674110e-02 -4.48097736e-01 3.50058004e-02 -3.60614657e-01 -3.40342343e-01 -3.82126093e-01 4.90895480e-01 9.49169517e-01 2.92678773e-01 -6.24825537e-01 1.34655929e+00 1.19639754e+00 -3.09431553e-01 -3.41006331e-02 -1.23075962e+00 -7.35923290e-01 -2.64678657e-01 -4.47594434e-01 -2.27825399e-02 7.01908052e-01 -1.15732148e-01 1.45794377e-01 -5.21394253e-01 1.97888359e-01 6.49674356e-01 -3.09282690e-01 6.86405122e-01 -8.89856398e-01 1.08326644e-01 -7.04715669e-01 -1.05242503e+00 -9.20605183e-01 -7.85478801e-02 -3.43380094e-01 1.24936633e-01 -1.36102867e+00 4.18827116e-01 -7.63383135e-02 -3.57472450e-01 8.76151919e-02 1.92053035e-01 7.27629244e-01 5.96552268e-02 2.79689550e-01 -8.05245697e-01 -1.75534543e-02 8.53955269e-01 -6.24921739e-01 -4.99556847e-02 5.04796267e-01 3.87461632e-02 8.49366486e-01 1.13193703e+00 -2.16839701e-01 7.91669451e-03 -2.60814577e-01 2.66610533e-02 3.21338959e-02 1.57428950e-01 -1.29477727e+00 6.03951395e-01 -1.98675558e-01 6.79086149e-01 -6.72825098e-01 6.02221251e-01 -1.02833462e+00 -2.84148425e-01 4.41358924e-01 2.39689559e-01 -1.12244919e-01 7.15284646e-01 4.42980468e-01 -1.79789662e-01 -3.64765972e-01 1.06356192e+00 5.23250997e-02 -1.27296889e+00 7.32247010e-02 -1.29935169e+00 -3.84772539e-01 1.29414093e+00 -1.36516774e+00 -3.12941074e-01 -1.55524328e-01 -3.14025253e-01 -2.43999362e-01 3.38128060e-01 5.99526048e-01 1.09538352e+00 -1.09031332e+00 -5.56736529e-01 4.42456275e-01 8.96728188e-02 -6.20264769e-01 2.60698497e-01 7.08676219e-01 -1.28680539e+00 1.16245413e+00 -7.80332744e-01 -3.54408056e-01 -1.61901379e+00 1.85594678e-01 4.79679316e-01 1.44851282e-01 -4.11390513e-01 4.78795826e-01 -2.56902993e-01 1.32924300e-02 -5.87692969e-02 -4.11507398e-01 -4.29054826e-01 -4.58200485e-01 4.52835381e-01 8.58368397e-01 1.03168376e-01 -8.85499179e-01 -6.24566078e-01 1.10294557e+00 -1.71542600e-01 -1.71656311e-01 9.38152015e-01 1.38579858e-02 4.23934937e-01 1.17615171e-01 1.01999962e+00 3.22090268e-01 -1.08571589e+00 3.31868291e-01 -1.07304305e-02 -7.31098354e-01 3.79994929e-01 -5.39921343e-01 -1.05017817e+00 9.68971968e-01 8.39839756e-01 -9.68219191e-02 6.26193702e-01 -7.68888414e-01 1.17331803e+00 7.21875727e-01 1.63915411e-01 -1.54872060e+00 -3.22240680e-01 3.19736034e-01 2.23941341e-01 -1.66679382e+00 -5.71804261e-03 3.62803675e-02 -6.54048741e-01 1.54570711e+00 6.94592118e-01 -3.49386364e-01 4.10589427e-01 3.50343287e-01 -1.32684648e-01 -1.68237805e-01 1.70210155e-03 1.07135870e-01 4.77906942e-01 5.46942294e-01 2.77998984e-01 1.41111583e-01 1.54750824e-01 3.29500914e-01 4.67833996e-01 4.13944349e-02 6.62627637e-01 1.01380396e+00 -9.66528118e-01 -6.40925646e-01 -8.49121690e-01 5.29008329e-01 -6.45664036e-01 1.57404289e-01 -3.94635499e-01 8.88804138e-01 8.27171728e-02 1.05521429e+00 3.48590612e-01 -2.49009192e-01 3.96350414e-01 1.43663079e-01 4.34233546e-01 -2.99893498e-01 -3.86974961e-01 -1.25716567e-01 1.56743512e-01 5.44756576e-02 6.67828098e-02 -6.97018981e-01 -1.10726738e+00 -7.37492025e-01 -2.50730664e-01 -2.14681715e-01 9.34891522e-01 9.38847184e-01 -1.06139384e-01 1.37065485e-01 4.07115281e-01 -3.78609180e-01 -4.85689677e-02 -5.23802698e-01 -8.19689631e-01 2.09121034e-03 5.52598834e-01 -4.59271520e-01 1.87669605e-01 7.83126056e-02]
[9.825849533081055, -4.964443206787109]
d1feafef-067b-47bf-91b6-2c184c020a17
electricity-price-forecasting-the-dawn-of
2204.00883
null
https://arxiv.org/abs/2204.00883v1
https://arxiv.org/pdf/2204.00883v1.pdf
Electricity Price Forecasting: The Dawn of Machine Learning
Electricity price forecasting (EPF) is a branch of forecasting on the interface of electrical engineering, statistics, computer science, and finance, which focuses on predicting prices in wholesale electricity markets for a whole spectrum of horizons. These range from a few minutes (real-time/intraday auctions and continuous trading), through days (day-ahead auctions), to weeks, months or even years (exchange and over-the-counter traded futures and forward contracts). Over the last 25 years, various methods and computational tools have been applied to intraday and day-ahead EPF. Until the early 2010s, the field was dominated by relatively small linear regression models and (artificial) neural networks, typically with no more than two dozen inputs. As time passed, more data and more computational power became available. The models grew larger to the extent where expert knowledge was no longer enough to manage the complex structures. This, in turn, led to the introduction of machine learning (ML) techniques in this rapidly developing and fascinating area. Here, we provide an overview of the main trends and EPF models as of 2022.
['Rafał Weron', 'Grzegorz Marcjasz', 'Jesus Lago', 'Arkadiusz Jędrzejewski']
2022-04-02
null
null
null
null
['electrical-engineering']
['miscellaneous']
[-4.81550872e-01 -1.79759726e-01 -1.29187673e-01 -2.88681030e-01 -3.69481474e-01 -9.26175058e-01 8.64190936e-01 3.22750479e-01 -1.01708323e-01 1.11413991e+00 -6.33199960e-02 -7.30220914e-01 -3.85513246e-01 -1.04830897e+00 -1.13068469e-01 -5.48255086e-01 -6.65768206e-01 5.01713514e-01 -1.97525054e-01 -5.02454996e-01 3.22593600e-01 5.22668958e-01 -1.29994893e+00 8.08595419e-02 5.71992755e-01 1.60658967e+00 3.60747650e-02 3.19346905e-01 -3.33592296e-01 5.71835339e-01 -3.89144689e-01 -5.35322785e-01 5.24223983e-01 -8.45218375e-02 -4.08975959e-01 -4.11081493e-01 -7.42228508e-01 -4.42300290e-01 -1.68763295e-01 6.07106805e-01 1.75177708e-01 -4.68798011e-04 3.09733868e-01 -1.48466945e+00 -5.28589547e-01 9.23737705e-01 -6.43056631e-01 3.26993942e-01 4.48300168e-02 -1.08604945e-01 1.21704137e+00 -8.40921223e-01 -3.93472239e-02 6.07927740e-01 5.93874097e-01 -2.58915305e-01 -1.13472378e+00 -4.53296810e-01 8.08605179e-02 3.71885955e-01 -1.09588897e+00 1.24427170e-01 8.39646101e-01 -5.45229673e-01 1.33620811e+00 4.59171981e-01 9.25251901e-01 3.64384204e-01 4.22358215e-01 4.50882614e-01 1.29910278e+00 -4.38155413e-01 4.56183046e-01 3.55168134e-01 -2.20280200e-01 -3.68793517e-01 1.19191952e-01 4.43767011e-01 -7.50779957e-02 -4.95313048e-01 5.27285993e-01 2.32149780e-01 -1.88080549e-01 -4.49504443e-02 -1.00977695e+00 1.18420875e+00 5.74341081e-02 5.66872716e-01 -7.57530332e-01 -3.78034532e-01 4.88087833e-01 7.24477112e-01 5.28886139e-01 4.47962046e-01 -1.16143417e+00 -4.74863827e-01 -1.38054705e+00 4.71110523e-01 1.18934453e+00 5.05355895e-01 4.83463913e-01 3.21916431e-01 1.97729006e-01 5.62965810e-01 2.25762296e-02 3.90937299e-01 4.97581184e-01 -5.60419679e-01 4.02716368e-01 4.55432981e-02 5.49019575e-01 -6.83281362e-01 -6.75078154e-01 -6.07876182e-01 -1.09919226e+00 2.37550050e-01 5.30473769e-01 -2.93735385e-01 -2.32069299e-01 1.12599254e+00 -6.08833209e-02 -1.21618189e-01 -2.38132164e-01 5.28806865e-01 -9.77962017e-02 9.65910912e-01 -1.35374695e-01 -6.78160965e-01 1.19853294e+00 -6.17341042e-01 -4.46989983e-01 1.66070580e-01 3.85000795e-01 -9.08730924e-01 3.99232656e-01 6.45832181e-01 -1.22736526e+00 -2.83834934e-01 -7.61674464e-01 5.21531284e-01 -7.75502801e-01 -5.28407097e-01 5.52995980e-01 4.50916052e-01 -8.79261494e-01 1.01122749e+00 -5.20763755e-01 2.67633468e-01 9.30098295e-02 2.42214501e-01 2.15700924e-01 4.90279436e-01 -1.66199446e+00 1.23417163e+00 3.23597342e-01 4.22376573e-01 -1.33446194e-02 -8.46266031e-01 -5.07748842e-01 3.53002310e-01 1.62510768e-01 -2.52969056e-01 1.27932608e+00 -5.99693775e-01 -1.33881593e+00 1.61064103e-01 1.50604710e-01 -9.08690155e-01 6.53987944e-01 2.83544868e-01 -7.97556758e-01 -6.12426162e-01 -2.56542891e-01 5.92002571e-02 4.00047809e-01 -5.44622004e-01 -9.17897463e-01 -2.98160940e-01 -2.28513870e-02 -1.55262634e-01 2.14952126e-01 1.37290388e-01 6.36601448e-01 -9.04882729e-01 -1.97537988e-03 -5.35037458e-01 -2.99098015e-01 -5.31730711e-01 6.00369535e-02 -5.07537961e-01 5.58052599e-01 -1.04041886e+00 1.20508814e+00 -1.69834423e+00 -2.50063270e-01 6.70950353e-01 -2.12082043e-01 -2.09100842e-01 3.39019746e-01 9.48406518e-01 -5.39861202e-01 -2.03119703e-02 -1.32646844e-01 2.89597154e-01 4.25258964e-01 1.65266588e-01 -6.55863464e-01 2.38793731e-01 1.12031847e-01 1.11615467e+00 -6.61497653e-01 4.91085470e-01 3.55267107e-01 7.84477144e-02 5.77048771e-02 -2.61577338e-01 -3.30106407e-01 3.35705459e-01 -3.17060351e-01 5.65770924e-01 7.12300777e-01 -3.29435468e-01 1.47661224e-01 2.01283246e-01 -6.49678946e-01 2.54877001e-01 -1.13432336e+00 9.31676686e-01 -4.99884546e-01 4.81698900e-01 -1.64844796e-01 -1.38489854e+00 9.12694871e-01 4.46873575e-01 9.93929803e-01 -1.19599950e+00 4.63475361e-02 6.74465299e-01 -3.40477303e-02 -2.22780630e-01 4.93230283e-01 -2.19918087e-01 -1.26663849e-01 6.58493698e-01 -6.71973884e-01 -8.83098170e-02 4.75825906e-01 -4.53681588e-01 6.73757911e-01 -9.78823528e-02 4.98130172e-01 -3.10376853e-01 4.95362222e-01 -7.51041062e-03 5.64335525e-01 2.01301888e-01 1.21909440e-01 5.18847965e-02 4.41684216e-01 -1.14405668e+00 -1.18700671e+00 -8.44968438e-01 -6.10904694e-01 7.72224009e-01 -3.90597463e-01 3.72798629e-02 -2.10330024e-01 -5.60092553e-02 5.50767064e-01 9.72055554e-01 -2.60213494e-01 4.88010496e-01 -4.16744083e-01 -9.92212057e-01 -2.06462353e-01 5.61159492e-01 3.81466985e-01 -1.26145554e+00 -5.83804429e-01 8.12437117e-01 2.58384854e-01 -9.80390847e-01 -1.36009157e-01 4.28943425e-01 -8.15076411e-01 -7.63458073e-01 -1.06479645e+00 -3.29688549e-01 1.50294900e-01 -3.72785419e-01 1.33903337e+00 -2.89506912e-01 -2.01708108e-01 -1.03498966e-01 -1.75030529e-01 -7.58361340e-01 -9.26015154e-02 -1.94184706e-01 1.13669477e-01 -9.51846763e-02 3.75463247e-01 -8.61088097e-01 -5.69244504e-01 1.12521365e-01 -5.63892901e-01 -2.45273411e-01 5.02521753e-01 7.75324285e-01 4.15820837e-01 4.61332828e-01 1.25028110e+00 -5.20751417e-01 7.50583470e-01 -6.51215911e-01 -1.44801784e+00 3.00917268e-01 -1.28158939e+00 -2.00476274e-01 7.65869081e-01 -1.00907870e-01 -5.04738331e-01 -4.22380000e-01 -1.40076745e-02 2.80197449e-02 1.47270933e-01 9.72114086e-01 3.42210054e-01 1.49247348e-01 2.32859980e-02 5.03232241e-01 -2.46689126e-01 -6.24715686e-01 1.85440838e-01 8.08729827e-01 3.96086514e-01 -1.90275505e-01 1.01794529e+00 -2.88403872e-02 1.63117439e-01 -5.65778852e-01 -3.07123035e-01 2.10560355e-02 -6.49571657e-01 -7.55248964e-02 4.05980825e-01 -7.88406909e-01 -7.69512951e-01 6.02848291e-01 -8.20509732e-01 -3.31753850e-01 -4.87412721e-01 5.12050271e-01 -3.16870391e-01 1.07613236e-01 -7.15758085e-01 -1.23381138e+00 -3.60842079e-01 -7.07092941e-01 2.60407925e-01 2.92169988e-01 -3.33936274e-01 -1.20802164e+00 -3.35439458e-04 1.04751721e-01 9.16759372e-01 5.39645493e-01 1.22905219e+00 -7.36161053e-01 -4.44826692e-01 -4.10946548e-01 -1.78368598e-01 4.79756057e-01 2.58127630e-01 -2.50382602e-01 -5.82415342e-01 -3.53362769e-01 1.35505155e-01 9.45907682e-02 2.15147570e-01 4.38684821e-01 1.00300014e+00 -5.51380634e-01 -4.19744253e-02 1.05375841e-01 1.33367646e+00 7.56554782e-01 5.86155772e-01 6.57304704e-01 -1.62184119e-01 5.52476406e-01 4.11431015e-01 7.18346179e-01 7.44274914e-01 4.06362802e-01 2.03017578e-01 -3.51056531e-02 1.02570152e+00 -1.08437426e-02 -1.95784587e-03 8.12043548e-01 -3.51849496e-01 1.85313776e-01 -8.82354081e-01 5.19689083e-01 -1.55691934e+00 -1.02893329e+00 1.23707496e-01 2.35267043e+00 6.92921698e-01 5.14566779e-01 3.01524252e-01 4.72621828e-01 3.72817427e-01 -8.90747309e-02 -7.36302733e-01 -8.77802730e-01 -1.49595037e-01 2.60581642e-01 6.93021059e-01 2.36772448e-01 -9.36389506e-01 -4.76345420e-02 6.29119539e+00 5.54367244e-01 -1.12572563e+00 -3.57891947e-01 1.01914656e+00 2.29174957e-01 -5.32860994e-01 3.54557410e-02 -4.49457914e-01 8.73609543e-01 1.22001064e+00 -7.55823195e-01 1.01168311e+00 7.81239629e-01 5.73988855e-01 -2.00769007e-01 -9.55862403e-01 9.57282007e-01 -5.93854189e-01 -1.38426614e+00 -5.77134967e-01 4.65777844e-01 8.80398571e-01 3.84392500e-01 -1.84818581e-01 4.02556062e-01 6.92573786e-02 -1.00970364e+00 6.30046129e-01 7.43445456e-01 3.32802325e-01 -1.01706445e+00 1.13654935e+00 7.71670341e-01 -1.22565627e+00 -5.26635528e-01 -2.41112664e-01 -5.07053494e-01 4.02484655e-01 9.65843499e-01 -2.66093820e-01 7.84444571e-01 8.73400331e-01 4.46971118e-01 2.27174573e-02 9.09592271e-01 2.25959048e-01 2.54416883e-01 -5.96754074e-01 -1.15004353e-01 3.86629164e-01 -5.28253436e-01 -8.99246782e-02 5.94795167e-01 4.78407562e-01 3.07169586e-01 5.68487085e-02 6.87036157e-01 8.27846751e-02 -1.07984908e-01 -2.85974950e-01 -8.32317248e-02 5.23196578e-01 1.27836561e+00 -5.81912756e-01 -1.70218304e-01 -9.08353508e-01 3.14544737e-01 -2.42845669e-01 4.48895305e-01 -7.58472025e-01 -4.56925899e-01 4.60589200e-01 4.68963295e-01 4.70592380e-01 -2.79504120e-01 -4.13845390e-01 -9.43695724e-01 1.90913543e-01 -6.58592761e-01 3.23123217e-01 -4.97639179e-01 -1.71533704e+00 3.99586916e-01 1.28120974e-01 -1.08472991e+00 -8.75857115e-01 -5.37165642e-01 -8.36220503e-01 1.24441648e+00 -2.06866670e+00 -3.18836421e-01 4.00233537e-01 2.51903504e-01 3.06038082e-01 -4.35185492e-01 9.72450554e-01 3.32551301e-01 -1.15939617e-01 1.68252662e-01 8.66678715e-01 2.13330016e-01 -1.26537785e-01 -1.20962775e+00 7.23526061e-01 1.09683163e-01 6.09000772e-02 9.43652764e-02 6.12334132e-01 -2.80298233e-01 -1.12875521e+00 -5.32947302e-01 1.32579875e+00 1.20324418e-01 1.07747996e+00 -2.62130201e-01 -8.20580959e-01 4.77711439e-01 4.21443224e-01 -1.12934694e-01 4.45520967e-01 -4.49233167e-02 2.39573810e-02 -4.66392934e-01 -1.26268435e+00 4.25529405e-02 -7.84024503e-03 -3.68067503e-01 -5.51038146e-01 1.27432197e-01 6.60415515e-02 -1.07721463e-01 -1.32163405e+00 4.48192120e-01 7.38135934e-01 -8.17870855e-01 6.91532016e-01 -1.93802446e-01 -1.22943416e-01 -5.18116280e-02 -8.56570303e-02 -1.27616704e+00 -3.21134686e-01 -9.85927105e-01 -2.08878160e-01 9.18562174e-01 7.84515262e-01 -1.30219269e+00 6.03216946e-01 1.00721753e+00 2.25744009e-01 -1.21109664e+00 -1.20322025e+00 -7.49157667e-01 3.06021392e-01 -2.95701861e-01 1.28745461e+00 1.01217568e+00 3.05249304e-01 -1.70606643e-01 -1.91108465e-01 -2.45201409e-01 6.60490930e-01 8.09147894e-01 9.36649963e-02 -1.55022979e+00 -3.60551983e-01 -6.70578778e-01 -3.02199334e-01 -7.47805178e-01 -2.85401702e-01 -6.50840521e-01 -5.52566886e-01 -1.35479319e+00 -2.45719030e-01 -3.00822079e-01 -6.52534425e-01 4.17276204e-01 4.83830482e-01 -6.53564173e-04 1.75812826e-01 2.23158136e-01 4.24676001e-01 2.73139119e-01 8.30976784e-01 -1.53221607e-01 -5.63851073e-02 5.73409855e-01 -3.73669326e-01 5.78908026e-01 1.05999827e+00 1.19506633e-02 -2.16337949e-01 -7.49472752e-02 6.50651515e-01 3.58821124e-01 1.36943221e-01 -4.46753770e-01 1.65737495e-01 -3.20584059e-01 7.55040944e-01 -1.02527905e+00 -3.18048932e-02 -9.32423115e-01 6.18578255e-01 4.87638533e-01 1.31444618e-01 6.98219061e-01 1.64457127e-01 2.37629414e-01 -3.58562410e-01 7.10475817e-02 4.58751589e-01 -1.89539492e-01 -5.08732677e-01 2.56425858e-01 -2.47192621e-01 -2.27713376e-01 1.21176481e+00 -1.79407090e-01 3.31321894e-03 -5.59369087e-01 -1.06035292e+00 6.89456165e-01 -3.57623659e-02 5.42717040e-01 1.58349007e-01 -1.13356888e+00 -7.97658563e-01 3.65861803e-01 -4.99583751e-01 -1.99046195e-01 6.42848238e-02 8.67402971e-01 -3.86359960e-01 8.81252825e-01 -5.30742249e-03 -1.50940299e-01 -5.47676921e-01 3.92421752e-01 3.31530780e-01 -5.90105832e-01 -6.85432732e-01 2.49623477e-01 -4.80904043e-01 -1.48711398e-01 7.87775964e-02 -5.24128735e-01 -1.76074207e-01 4.94917780e-01 5.02714038e-01 5.71967781e-01 1.68318674e-01 -3.12029958e-01 -1.82437465e-01 4.57655936e-01 2.10451201e-01 1.04081653e-01 1.89839947e+00 -9.58052948e-02 -3.09521079e-01 6.68330073e-01 9.34986711e-01 -4.54907000e-01 -1.16085327e+00 3.88390273e-02 5.80908120e-01 -3.75753343e-02 1.10182250e-02 -1.02051926e+00 -1.05268764e+00 5.40403843e-01 2.86431640e-01 1.20128977e+00 1.17952740e+00 -2.81059682e-01 1.09309042e+00 2.06565052e-01 8.09447825e-01 -1.50524187e+00 -1.00789201e+00 3.75797480e-01 8.71657729e-01 -9.08087134e-01 7.78257996e-02 2.40362853e-01 -3.05415422e-01 1.40168667e+00 -5.22097230e-01 1.26497790e-01 1.34104836e+00 5.48115432e-01 -1.86601475e-01 2.29713038e-01 -8.62630606e-01 3.00492257e-01 1.98972207e-02 9.61243361e-02 3.53550702e-01 2.75625288e-01 -2.93851972e-01 9.62582231e-01 -6.73948765e-01 3.11278701e-01 4.75923538e-01 8.72674167e-01 -2.86771357e-01 -1.23241603e+00 -2.01145425e-01 8.61730933e-01 -5.89774370e-01 -2.26612285e-01 2.43124336e-01 8.84838700e-01 -1.04401559e-01 9.11835492e-01 3.90973002e-01 1.53727874e-01 4.65785235e-01 3.20947468e-01 1.12024015e-02 6.47224337e-02 -5.19824386e-01 2.36026019e-01 -1.34165600e-01 -2.89978594e-01 -5.44753559e-02 -9.65722501e-01 -1.13334513e+00 -8.00199926e-01 -4.51141357e-01 6.05170548e-01 8.18686962e-01 9.52588677e-01 -4.95384634e-02 1.51951447e-01 1.39249551e+00 -9.08625603e-01 -1.06821907e+00 -9.28296268e-01 -1.28978419e+00 -3.64344835e-01 2.75580168e-01 -2.09277511e-01 -6.83423698e-01 -2.98670530e-01]
[4.682650089263916, 4.0253472328186035]
24320d10-97d3-4c2c-8dfb-cae8ac12eb34
iterative-soft-shrinkage-learning-for
2303.09650
null
https://arxiv.org/abs/2303.09650v1
https://arxiv.org/pdf/2303.09650v1.pdf
Iterative Soft Shrinkage Learning for Efficient Image Super-Resolution
The field of image super-resolution (SR) has witnessed extensive neural network designs from CNN to transformer architectures. However, prevailing SR models suffer from prohibitive memory footprint and intensive computations, which limits further deployment on computational-constrained platforms. In this work, we investigate the potential of network pruning for super-resolution to take advantage of off-the-shelf network designs and reduce the underlying computational overhead. Two main challenges remain in applying pruning methods for SR. First, the widely-used filter pruning technique reflects limited granularity and restricted adaptability to diverse network structures. Second, existing pruning methods generally operate upon a pre-trained network for the sparse structure determination, failing to get rid of dense model training in the traditional SR paradigm. To address these challenges, we adopt unstructured pruning with sparse models directly trained from scratch. Specifically, we propose a novel Iterative Soft Shrinkage-Percentage (ISS-P) method by optimizing the sparse structure of a randomly initialized network at each iteration and tweaking unimportant weights with a small amount proportional to the magnitude scale on-the-fly. We observe that the proposed ISS-P could dynamically learn sparse structures adapting to the optimization process and preserve the sparse model's trainability by yielding a more regularized gradient throughput. Experiments on benchmark datasets demonstrate the effectiveness of the proposed ISS-P compared with state-of-the-art methods over diverse network architectures.
['Zhiqiang Tao', 'Yun Fu', 'Yulun Zhang', 'Huan Wang', 'Jiamian Wang']
2023-03-16
null
null
null
null
['image-super-resolution']
['computer-vision']
[ 5.97642958e-01 3.89237218e-02 -1.98519915e-01 -2.96010911e-01 -4.48714346e-01 -5.51479794e-02 1.94286957e-01 -2.18178630e-01 -3.48531514e-01 7.12901294e-01 9.37523618e-02 -1.43379241e-01 -2.62835801e-01 -9.07593906e-01 -7.33590424e-01 -5.23408771e-01 -2.51089931e-02 4.90786172e-02 4.85934138e-01 -2.54038572e-01 1.47949412e-01 6.21484339e-01 -1.68175638e+00 5.41653872e-01 9.87042785e-01 1.24217844e+00 5.29481053e-01 2.63303518e-01 -1.81724817e-01 9.36014175e-01 -3.66757631e-01 -2.39936829e-01 6.00429595e-01 -2.51357049e-01 -4.24097359e-01 -1.26460418e-01 7.97441423e-01 -3.04794580e-01 -4.70817059e-01 1.17443430e+00 4.65803564e-01 7.97205344e-02 6.77945977e-03 -4.87725019e-01 -5.52469075e-01 9.36358750e-01 -7.58610129e-01 6.31250739e-01 -1.24852583e-01 -1.21876488e-04 8.24735463e-01 -1.13910854e+00 6.50749803e-01 1.14980042e+00 9.44317698e-01 5.29880464e-01 -1.43230355e+00 -8.77658546e-01 4.35473800e-01 1.18848592e-01 -1.64358497e+00 -7.49562681e-01 8.22774172e-01 -5.04285693e-02 9.62923646e-01 1.83859482e-01 6.59787595e-01 7.91081250e-01 -1.98446974e-01 4.06399488e-01 1.16800344e+00 -2.43537247e-01 2.87519604e-01 2.12082155e-02 3.98505758e-03 8.84341240e-01 4.49250281e-01 -2.88997963e-02 -9.81357276e-01 5.48357284e-03 1.10027027e+00 -1.23496637e-01 -3.72167677e-01 -2.04202414e-01 -7.94452727e-01 6.48435175e-01 5.35411358e-01 2.93935061e-01 -3.23226810e-01 -1.06054060e-02 3.14597696e-01 2.93357044e-01 5.33179879e-01 2.65465707e-01 -5.57234943e-01 7.02623352e-02 -1.50374794e+00 1.66924521e-01 6.99165523e-01 8.09009790e-01 7.32764900e-01 4.52144414e-01 -5.32910377e-02 9.53764141e-01 -2.66053796e-01 6.53037727e-02 4.76365447e-01 -9.40794706e-01 6.08119607e-01 6.37042046e-01 -3.76069963e-01 -1.25586748e+00 -2.19068557e-01 -9.77571309e-01 -1.27662337e+00 1.56962276e-01 1.52851269e-01 7.99795911e-02 -8.91321421e-01 1.64866889e+00 4.50655669e-01 4.93035942e-01 -8.77596065e-02 8.70931506e-01 8.40806067e-01 4.74508166e-01 -1.34005919e-01 -2.70406276e-01 1.07994938e+00 -1.02466524e+00 -2.12697551e-01 -4.30138081e-01 2.47480348e-01 -5.87678671e-01 1.03295624e+00 6.18401051e-01 -1.31675386e+00 -5.99288702e-01 -1.16483617e+00 -1.42264426e-01 1.13711327e-01 4.04108405e-01 5.76631725e-01 4.52861816e-01 -1.07785463e+00 1.01916301e+00 -1.01658106e+00 -1.47912443e-01 9.06511188e-01 6.13822639e-01 -1.74782738e-01 -7.99531341e-02 -8.34057570e-01 5.50935507e-01 4.77373958e-01 2.25621119e-01 -7.85154819e-01 -1.08789122e+00 -5.80552161e-01 5.02350092e-01 6.01065993e-01 -7.79995799e-01 1.03382075e+00 -1.05235922e+00 -1.78442037e+00 4.91900116e-01 -2.55415291e-01 -8.27225149e-01 3.13339740e-01 -3.87906700e-01 -2.77073085e-01 3.55252713e-01 -1.76633343e-01 4.54198599e-01 1.24442708e+00 -1.09499812e+00 -7.10104764e-01 -2.89617360e-01 1.64251611e-01 1.44192532e-01 -8.22522819e-01 2.27718763e-02 -3.45895767e-01 -7.96286762e-01 4.14774776e-01 -6.40812516e-01 -4.59282726e-01 1.31642893e-01 -2.18950644e-01 1.75480321e-01 8.83060098e-01 -4.92552340e-01 1.64913809e+00 -2.07913065e+00 1.69316515e-01 1.50151089e-01 5.95533431e-01 5.43864965e-01 -1.90218940e-01 -6.76625445e-02 1.67843103e-02 -1.59507822e-02 -1.47542074e-01 -4.83042777e-01 -5.53677440e-01 1.02555312e-01 -2.79528171e-01 3.13496858e-01 3.25973302e-01 4.26782578e-01 -6.44551158e-01 -6.86836243e-01 -2.27438752e-02 7.68078864e-01 -8.67573559e-01 1.29435780e-02 -8.81848559e-02 2.43258089e-01 -4.93005753e-01 7.34167993e-01 6.91835344e-01 -6.41081989e-01 3.52270365e-01 -5.66541135e-01 -2.83879995e-01 4.64148104e-01 -1.38878083e+00 1.64715934e+00 -5.23818612e-01 2.92220563e-01 3.85244370e-01 -1.22806180e+00 1.04430592e+00 -1.60748663e-03 3.03362399e-01 -7.08481729e-01 -1.78303551e-02 3.76503319e-01 -3.76881324e-02 -3.49745117e-02 5.30949771e-01 2.30567027e-02 5.06007612e-01 4.74763550e-02 2.67861150e-02 4.06251878e-01 3.17832291e-01 1.64294705e-01 1.25993359e+00 6.12341315e-02 3.30948532e-01 -4.09113884e-01 5.84639966e-01 -1.31452799e-01 9.17094052e-01 7.61872530e-01 1.47048533e-01 5.47910392e-01 2.36100584e-01 -6.72491491e-01 -1.03914702e+00 -8.07153940e-01 -2.06810579e-01 1.09769499e+00 -1.85668301e-02 -5.46336710e-01 -7.16397345e-01 -4.40362990e-01 -2.35820249e-01 1.33138433e-01 -4.60656315e-01 1.59022529e-02 -1.11245799e+00 -6.99423969e-01 3.35525393e-01 5.37495136e-01 8.07889402e-01 -8.31866503e-01 -8.56194913e-01 3.18244696e-01 2.21110225e-01 -1.25074029e+00 -2.23695502e-01 1.88236848e-01 -1.32272470e+00 -6.61863565e-01 -4.09185946e-01 -8.12593400e-01 7.88539946e-01 4.50519204e-01 1.10586417e+00 2.95776546e-01 -2.27384612e-01 -3.23997468e-01 -1.17924310e-01 6.95599243e-02 -3.69548835e-02 5.16231120e-01 8.20033625e-03 -1.48705440e-03 1.03528768e-01 -1.22519362e+00 -9.80804861e-01 8.82244632e-02 -7.34866381e-01 3.57703567e-01 9.06356215e-01 8.49612474e-01 1.01769733e+00 2.90252745e-01 5.87276757e-01 -1.16371822e+00 4.31472659e-01 -3.80147964e-01 -6.68772638e-01 6.14920408e-02 -7.68313766e-01 4.90645245e-02 1.12056518e+00 -6.32010996e-01 -1.11080754e+00 1.90188006e-01 1.06883787e-01 -7.13413537e-01 1.67037487e-01 7.03394949e-01 6.35684282e-02 -4.99253154e-01 7.46352255e-01 4.43007648e-01 -9.61347520e-02 -6.86542571e-01 2.67443597e-01 1.78561777e-01 7.22838402e-01 -6.47467375e-01 9.85414922e-01 6.68737531e-01 1.88272715e-01 -7.74278820e-01 -1.17803013e+00 -2.26137280e-01 -4.81749505e-01 1.31941155e-01 3.16287369e-01 -1.17067778e+00 -2.46776611e-01 1.09464921e-01 -8.73184443e-01 -2.40978211e-01 -3.77275825e-01 1.71954423e-01 -1.29320234e-01 3.40659499e-01 -7.29841828e-01 -4.94364798e-01 -8.07720602e-01 -9.66315210e-01 7.73740411e-01 3.27331543e-01 -3.98777463e-02 -4.84791368e-01 -1.88216016e-01 3.09337258e-01 9.10936117e-01 9.31999981e-02 7.85698771e-01 -3.22895736e-01 -8.85836840e-01 4.24189009e-02 -5.80497622e-01 3.63329321e-01 -7.54879862e-02 -2.06092179e-01 -7.18705416e-01 -4.57726747e-01 1.88315928e-01 -2.88803369e-01 9.82976794e-01 3.66794050e-01 1.56488967e+00 -5.38741767e-01 -1.22263685e-01 1.12676299e+00 1.75794733e+00 -2.75179356e-01 4.34474319e-01 5.11733651e-01 7.71275818e-01 1.97524130e-01 2.70826668e-01 6.14555180e-01 1.11720569e-01 5.03972232e-01 4.23179179e-01 6.20239452e-02 -3.79398555e-01 -2.51026064e-01 1.52039230e-01 1.07066071e+00 -3.06106389e-01 2.62042463e-01 -6.07978880e-01 4.66022402e-01 -1.63614786e+00 -8.94845247e-01 2.82906502e-01 2.11539292e+00 1.12428486e+00 4.38971221e-01 -1.66800499e-01 5.53355925e-02 5.08065104e-01 4.35492337e-01 -5.96591830e-01 -1.35665879e-01 -1.62196621e-01 7.38826752e-01 5.27373075e-01 1.55963078e-01 -8.44098091e-01 9.23906922e-01 5.34010696e+00 1.09208846e+00 -1.38790166e+00 2.72240520e-01 6.34815037e-01 -6.15668774e-01 -1.15872703e-01 8.25348869e-02 -1.25652599e+00 2.59450644e-01 8.71341050e-01 -1.21033758e-01 7.62739897e-01 1.08357465e+00 2.80553192e-01 1.42298609e-01 -8.92376781e-01 9.85322893e-01 -3.52341682e-02 -1.77582836e+00 2.50709802e-01 -1.81017458e-01 9.44106817e-01 1.67377725e-01 1.38960302e-01 2.40614012e-01 -1.73426103e-02 -1.00836313e+00 8.15730751e-01 2.97697663e-01 9.44736421e-01 -7.34315991e-01 4.09842342e-01 2.68430948e-01 -1.56575203e+00 -3.24187309e-01 -6.88743234e-01 -9.00459960e-02 4.27889377e-02 9.56467688e-01 -3.38979810e-01 3.19579452e-01 9.92724061e-01 6.51632488e-01 -5.24701297e-01 8.43142807e-01 7.80601799e-02 7.15922892e-01 -6.08075798e-01 3.05051178e-01 2.09679857e-01 -1.89155996e-01 6.20595694e-01 1.18772066e+00 4.32879835e-01 9.60266590e-02 1.96261585e-01 9.14854944e-01 -4.03245211e-01 6.98702559e-02 -2.75571138e-01 7.13120550e-02 8.33762646e-01 1.33602929e+00 -8.14323425e-01 -2.64825821e-01 -5.35337150e-01 5.94817460e-01 7.70791531e-01 2.01362252e-01 -5.61682820e-01 -1.23648390e-01 4.89253581e-01 5.40889561e-01 7.73474336e-01 -1.53959647e-01 -7.17687070e-01 -1.12016487e+00 2.70116359e-01 -1.16717660e+00 2.29232162e-01 -2.16073036e-01 -1.04885983e+00 9.86150384e-01 -2.08060369e-01 -1.09965837e+00 2.50373423e-01 -2.05304578e-01 -5.23262441e-01 7.07674503e-01 -1.85599875e+00 -1.09149218e+00 -5.12702465e-01 6.08457744e-01 7.63211668e-01 -3.20792496e-01 5.19479811e-01 4.61760491e-01 -7.93565691e-01 7.22784519e-01 -5.10911681e-02 -2.03806907e-01 2.78682798e-01 -8.06112707e-01 2.23552808e-01 1.13157225e+00 3.40550952e-02 8.60210538e-01 4.79954004e-01 -5.76842844e-01 -1.34822702e+00 -1.21947801e+00 5.58435917e-01 2.69015342e-01 7.76789308e-01 -2.95364201e-01 -1.15390849e+00 3.77301574e-01 -1.68764725e-01 4.76066530e-01 3.02305907e-01 1.74293190e-01 -4.92732853e-01 -5.44883549e-01 -1.09529459e+00 6.25310481e-01 1.41717911e+00 -3.95565748e-01 -2.81290889e-01 1.18667476e-01 7.91953146e-01 -5.64772546e-01 -9.02765572e-01 6.03849471e-01 5.72838187e-01 -1.11592352e+00 1.12636447e+00 -3.84315163e-01 8.31068873e-01 -2.79002190e-01 -1.31839916e-01 -7.67104268e-01 -5.49229801e-01 -7.32225895e-01 -6.86407268e-01 1.00982881e+00 3.69723529e-01 -4.12298232e-01 1.17630768e+00 4.28468049e-01 -1.53379932e-01 -1.29284489e+00 -9.88217354e-01 -6.59790397e-01 -2.78272986e-01 -1.18216261e-01 6.40222788e-01 8.29568863e-01 -4.31861997e-01 3.01273763e-01 -4.11292911e-01 4.12590683e-01 8.09281468e-01 1.77131519e-01 6.25938475e-01 -1.18739772e+00 -6.54036820e-01 -4.51847255e-01 -1.84885100e-01 -1.17167723e+00 -7.14358613e-02 -7.30644524e-01 -2.30988845e-01 -1.16520524e+00 2.51900941e-01 -6.84639275e-01 -3.57353389e-01 4.40080702e-01 -1.86777458e-01 3.95676315e-01 2.22598314e-01 4.53688473e-01 -6.10960662e-01 4.72992599e-01 1.13109756e+00 8.23233873e-02 -3.80202085e-01 -2.13540435e-01 -8.49939466e-01 9.54002082e-01 9.08722281e-01 -6.57666504e-01 -7.23736584e-01 -7.06430316e-01 4.02514130e-01 -1.33257687e-01 3.43783259e-01 -1.27884126e+00 3.88231665e-01 2.48578489e-02 2.10241958e-01 -4.17279929e-01 1.79834113e-01 -7.47501850e-01 1.17040321e-01 2.69433707e-01 -2.48956233e-01 -4.50001396e-02 2.50950694e-01 6.36555672e-01 -2.39938855e-01 -2.24913642e-01 1.07741046e+00 -2.74787307e-01 -5.85085630e-01 4.68978703e-01 1.52431101e-01 1.09991387e-01 5.15532017e-01 -5.00732780e-01 -3.18951160e-01 1.59801811e-01 -5.53676844e-01 -6.91176727e-02 3.29254270e-01 -5.06316200e-02 7.70525396e-01 -1.02954650e+00 -6.39413714e-01 2.42844775e-01 -4.73902106e-01 4.87144858e-01 3.85425538e-01 8.10114145e-01 -5.74312031e-01 1.76325083e-01 -1.80297777e-01 -4.61989194e-01 -1.26211107e+00 4.09628451e-01 2.06717670e-01 -7.10762382e-01 -1.10036838e+00 1.08944237e+00 1.09978326e-01 6.64551556e-02 2.76361376e-01 -2.35568017e-01 -1.46817133e-01 -3.03502530e-01 6.33266628e-01 4.16054398e-01 2.64853805e-01 -3.39584738e-01 -8.21666270e-02 4.96234477e-01 -4.96521533e-01 4.50106382e-01 1.75101256e+00 -4.90571000e-02 -2.33087227e-01 3.71122360e-03 8.87053847e-01 -1.51983306e-01 -1.43563735e+00 -6.69076085e-01 -1.06578015e-01 -6.43788695e-01 4.85190570e-01 -4.31275189e-01 -1.67729056e+00 4.77674425e-01 5.31390011e-01 -5.46563640e-02 1.49849296e+00 -3.82831544e-01 1.08218408e+00 4.46969777e-01 5.20965219e-01 -1.15227413e+00 -8.38619992e-02 3.53568405e-01 6.90147460e-01 -9.21171248e-01 4.67751145e-01 -7.73911953e-01 -1.79050997e-01 1.05031598e+00 8.00605476e-01 -3.63116711e-01 6.92886531e-01 6.54163063e-01 -3.89741540e-01 -1.94533676e-01 -8.73383760e-01 1.04142316e-01 1.38304532e-01 3.77074689e-01 2.59777367e-01 -2.69726098e-01 -2.65441030e-01 6.01955593e-01 -3.01954746e-01 4.43102896e-01 2.21839443e-01 9.46463823e-01 -4.80336189e-01 -8.98688316e-01 -3.73183824e-02 8.67716730e-01 -6.76718652e-01 -4.78117794e-01 1.04183212e-01 4.91341799e-01 1.83006123e-01 4.17945921e-01 -1.25036076e-01 -2.55630672e-01 3.62590641e-01 -3.57616335e-01 4.49293882e-01 -7.27676511e-01 -7.27631390e-01 -3.19839120e-02 8.70647728e-02 -8.12661529e-01 -5.50281405e-01 -4.81366605e-01 -9.34974372e-01 -4.04670328e-01 -3.15601110e-01 -2.06291690e-01 3.25367540e-01 7.86295712e-01 6.63845122e-01 5.54057717e-01 3.92404467e-01 -1.00279331e+00 -8.79395306e-01 -7.61517763e-01 -4.01763767e-01 2.02189572e-02 1.73373431e-01 -5.50890088e-01 -3.77786577e-01 4.07478623e-02]
[10.898613929748535, -1.6566725969314575]
643782cc-a28b-4725-a633-02facd88e25d
polarimetric-pose-prediction
2112.03810
null
https://arxiv.org/abs/2112.03810v2
https://arxiv.org/pdf/2112.03810v2.pdf
Polarimetric Pose Prediction
Light has many properties that vision sensors can passively measure. Colour-band separated wavelength and intensity are arguably the most commonly used for monocular 6D object pose estimation. This paper explores how complementary polarisation information, i.e. the orientation of light wave oscillations, influences the accuracy of pose predictions. A hybrid model that leverages physical priors jointly with a data-driven learning strategy is designed and carefully tested on objects with different levels of photometric complexity. Our design significantly improves the pose accuracy compared to state-of-the-art photometric approaches and enables object pose estimation for highly reflective and transparent objects. A new multi-modal instance-level 6D object pose dataset with highly accurate pose annotations for multiple objects with varying photometric complexity is introduced as a benchmark.
['Benjamin Busam', 'Arturo Guridi', 'Pengyuan Wang', 'HyunJun Jung', 'Magdalena Wysock', 'Iuliia Skobleva', 'Patrick Ruhkamp', 'Yitong Li', 'Daoyi Gao']
2021-12-07
null
null
null
null
['transparent-objects', '6d-pose-estimation']
['computer-vision', 'computer-vision']
[ 2.97060996e-01 -1.31661654e-01 1.06635556e-01 -4.10599172e-01 -7.48864114e-01 -7.33688354e-01 8.01430166e-01 -4.15115327e-01 -3.53136271e-01 2.91520417e-01 -9.95424092e-02 3.07438105e-01 -2.78637171e-01 -4.02083546e-01 -7.75684476e-01 -9.58882630e-01 1.88833043e-01 8.87157679e-01 5.26855469e-01 1.81571737e-01 4.86310601e-01 8.89478683e-01 -1.74314606e+00 -4.82365228e-02 2.50878334e-01 1.28338706e+00 1.75112396e-01 7.07865417e-01 1.70855343e-01 2.60677278e-01 -5.50183356e-01 -4.18308794e-01 7.26612210e-01 3.63421381e-01 -4.57592160e-01 3.04077387e-01 1.19534039e+00 -3.01518083e-01 -3.98682654e-02 6.82149172e-01 6.82809651e-01 -2.13309452e-01 8.52079511e-01 -8.86831284e-01 -3.88728589e-01 -2.42026493e-01 -6.03147030e-01 -1.66558653e-01 8.07766199e-01 5.31007767e-01 7.84996510e-01 -7.89725423e-01 5.36987185e-01 1.14149761e+00 7.94537246e-01 3.95433933e-01 -1.37776339e+00 -2.94388056e-01 -7.20889494e-02 2.08929554e-01 -1.04693460e+00 -2.88260430e-01 1.08886099e+00 -5.77551425e-01 9.80789781e-01 3.42809618e-01 6.95538223e-01 1.11595893e+00 2.28434950e-01 2.18184873e-01 1.70227563e+00 -3.84690702e-01 -1.05877481e-01 3.87355722e-02 -2.64469028e-01 4.44844186e-01 3.89102489e-01 3.68475258e-01 -9.16510761e-01 -1.96479157e-01 5.03871799e-01 -3.47957104e-01 -3.31983685e-01 -8.86063635e-01 -1.34144175e+00 9.75472108e-02 3.56398553e-01 -5.65668583e-01 -1.20097220e-01 3.50064248e-01 -1.76692262e-01 -1.33777112e-01 5.04925907e-01 7.08605468e-01 -6.70624256e-01 -1.54972985e-01 -3.28388363e-01 2.52907842e-01 6.87286258e-01 8.75299156e-01 8.83100331e-01 -2.87079006e-01 -1.43321693e-01 7.67491519e-01 9.29228365e-01 1.04608321e+00 -2.49164879e-01 -1.17850578e+00 1.85319379e-01 5.96804082e-01 5.18944085e-01 -4.74366814e-01 -5.61334550e-01 -3.25633585e-01 1.60532117e-01 5.60710609e-01 6.99217260e-01 2.53383547e-01 -9.79264319e-01 1.39653730e+00 7.99829543e-01 5.45331836e-02 -1.76531956e-01 1.32581508e+00 1.14899969e+00 1.60398811e-01 -4.46770847e-01 -1.17773764e-01 1.50037754e+00 -4.05629247e-01 -1.05707906e-02 -4.11282778e-01 -4.00979295e-02 -1.33669329e+00 9.02569234e-01 6.70021176e-01 -1.14267421e+00 -1.81099519e-01 -8.57740521e-01 -1.48513764e-01 -6.23586960e-02 1.02927506e-01 7.90653646e-01 9.37158585e-01 -5.43653488e-01 2.25765333e-01 -5.53524613e-01 -6.36607260e-02 4.17064816e-01 4.27991539e-01 -2.08276153e-01 5.14456481e-02 -6.13160491e-01 1.17865777e+00 1.95329189e-01 -7.79765397e-02 -6.75966024e-01 -1.17668021e+00 -3.95735145e-01 -7.03603148e-01 5.20662785e-01 -9.15500760e-01 1.20131934e+00 -3.05131048e-01 -2.06319857e+00 1.57022774e+00 2.39188895e-01 -1.60428807e-01 7.32040644e-01 -4.72771376e-01 -1.20594939e-02 4.21163738e-01 -3.60307425e-01 5.47088444e-01 1.27541935e+00 -1.60449028e+00 -2.02863678e-01 -7.31514156e-01 2.77682394e-01 4.91804540e-01 5.15899584e-02 9.19247791e-02 -5.81585348e-01 1.07339785e-01 5.04309237e-01 -1.28807855e+00 3.02564412e-01 3.19581717e-01 -3.34135920e-01 -2.17363626e-01 7.66179442e-01 -6.01836257e-02 2.18500957e-01 -1.80376697e+00 1.55601636e-01 7.09988549e-02 9.30102915e-02 6.55717030e-02 7.71040469e-02 1.46474794e-01 1.79687336e-01 -6.71498299e-01 1.02247000e-01 -4.03254598e-01 1.49500817e-01 1.41366068e-02 4.19059247e-02 9.87371981e-01 4.90889288e-02 8.04535568e-01 -5.83896160e-01 -2.47863322e-01 4.43911672e-01 6.95945680e-01 -4.09663945e-01 2.61235237e-01 -7.41073906e-01 6.92117393e-01 -6.53050914e-02 8.85795176e-01 1.08956242e+00 -7.99257830e-02 -2.14437261e-01 -8.01377058e-01 -2.06754267e-01 4.71835554e-01 -1.23632157e+00 1.71973407e+00 -3.54426682e-01 4.83683109e-01 4.04841565e-02 -1.67625085e-01 1.03499460e+00 -6.96484074e-02 7.10521162e-01 -5.72829187e-01 1.14672266e-01 1.30564317e-01 -2.50006855e-01 -5.75978041e-01 3.19615394e-01 -2.37659752e-01 1.41559392e-01 1.33186489e-01 -8.19073468e-02 -9.68222678e-01 -3.15312445e-01 -3.96235317e-01 7.74793923e-01 8.34601581e-01 -4.52148020e-02 -2.47966945e-01 3.99256557e-01 -1.27756715e-01 3.45645159e-01 4.87717718e-01 -3.60874496e-02 1.01361382e+00 1.26140118e-01 -4.29119051e-01 -8.48456144e-01 -1.43775535e+00 -6.59254789e-01 7.76675642e-01 5.06182790e-01 4.87031676e-02 -3.50239605e-01 -1.53407365e-01 6.34951115e-01 2.92932153e-01 -4.02236134e-01 -8.42113569e-02 -3.72407854e-01 -9.60482895e-01 1.17769592e-01 1.45524755e-01 2.56315172e-01 -6.74426615e-01 -1.12599683e+00 -1.64544463e-01 8.81526768e-02 -1.43909264e+00 1.94157884e-01 8.68999884e-02 -6.71658456e-01 -1.24077582e+00 -4.65797722e-01 -1.15267150e-01 2.10869431e-01 4.89341825e-01 1.41126418e+00 -5.69548845e-01 -6.00524306e-01 1.08835649e+00 -1.83476955e-01 -8.52781594e-01 -7.50128645e-03 -1.99435025e-01 2.61919975e-01 2.23895133e-01 3.81147027e-01 -4.55815613e-01 -9.26465094e-01 4.57291394e-01 -4.10593480e-01 -4.43648137e-02 6.15960717e-01 2.95739800e-01 6.61028266e-01 -5.02202034e-01 -3.54709983e-01 -5.73052764e-01 -1.05571650e-01 1.56254977e-01 -1.09814870e+00 7.04419613e-02 -4.46552038e-01 6.24650507e-04 -1.78392142e-01 -5.37917733e-01 -1.38536692e+00 2.70496011e-01 3.46028358e-01 -4.67077911e-01 -4.63362962e-01 -4.95103784e-02 -1.86990276e-01 -7.40971148e-01 9.43900943e-01 -1.60348594e-01 -1.85581967e-02 -6.03036761e-01 1.75776243e-01 4.69634503e-01 5.39210856e-01 -6.94614351e-01 1.13021278e+00 8.79926383e-01 7.34329402e-01 -1.01613557e+00 -1.04944980e+00 -6.48637831e-01 -7.80125499e-01 -6.27746999e-01 8.01159620e-01 -1.00170159e+00 -1.43609107e+00 8.00325453e-01 -1.14865851e+00 -2.08693117e-01 -4.75508906e-02 8.42732131e-01 -6.77442968e-01 2.71488547e-01 -2.26100877e-01 -9.27932143e-01 7.06889629e-02 -9.95809793e-01 1.88544846e+00 1.12412319e-01 1.29801527e-01 -6.32760167e-01 1.34748816e-01 1.00245941e+00 -2.62314305e-02 4.73736048e-01 2.70629942e-01 3.46374989e-01 -1.31787670e+00 -3.42494473e-02 -2.91347533e-01 -5.80924079e-02 -1.44284353e-01 1.69415116e-01 -1.60287917e+00 -1.28885269e-01 4.58500050e-02 -4.94808286e-01 7.65152574e-01 5.27865052e-01 9.04600859e-01 1.25877053e-01 -1.40006423e-01 8.28209341e-01 1.56649959e+00 -3.31438273e-01 6.55183196e-01 5.29974699e-01 1.02850902e+00 8.43291938e-01 7.86297202e-01 6.05811954e-01 2.54115522e-01 1.13537157e+00 1.07020557e+00 4.22033817e-01 -1.96830228e-01 4.46500838e-01 2.79118031e-01 2.39636526e-01 -4.98775572e-01 1.84219390e-01 -1.08833694e+00 -1.02077395e-01 -1.28665066e+00 -6.44710958e-01 -5.48717916e-01 2.30695415e+00 8.57014656e-01 1.92298323e-01 4.59413417e-02 3.22815627e-02 1.76490754e-01 1.64458677e-01 -5.15088141e-01 2.27298979e-02 -1.53791174e-01 1.31308317e-01 9.39411283e-01 4.56848532e-01 -1.10176587e+00 5.89272916e-01 6.27634001e+00 3.95300478e-01 -1.01724267e+00 -5.38332425e-02 -5.79896979e-02 -5.03460348e-01 -2.29964390e-01 -1.09331869e-02 -1.03240240e+00 2.08679691e-01 5.87496102e-01 6.79680228e-01 3.05878222e-01 6.14849508e-01 -1.30813373e-02 -5.67832232e-01 -1.18829429e+00 1.36437726e+00 3.10663044e-01 -9.52149808e-01 -3.98440063e-01 1.86866730e-01 7.04074383e-01 3.72630745e-01 4.19376612e-01 -6.22517526e-01 -1.74368724e-01 -7.05261052e-01 1.07353055e+00 9.71696734e-01 5.79945922e-01 -4.56155896e-01 2.72663653e-01 1.31282419e-01 -1.00749898e+00 2.06858926e-02 -3.96469772e-01 -3.71458419e-02 4.93888035e-02 7.35497594e-01 -8.65248024e-01 5.34459889e-01 1.09559357e+00 6.17447436e-01 -8.37706625e-01 1.33743143e+00 -3.23298275e-01 2.63402581e-01 -7.23730385e-01 -3.35237756e-02 -1.48782909e-01 -3.05664867e-01 1.04479229e+00 9.06492472e-01 -1.41098082e-01 -1.06987566e-01 -3.57523151e-02 9.03467774e-01 2.27350414e-01 -3.75235885e-01 -3.03169906e-01 4.66611803e-01 2.94617891e-01 1.41799748e+00 -4.19091523e-01 4.38962609e-01 -5.14364660e-01 5.35383463e-01 4.41691987e-02 2.57498950e-01 -6.92831457e-01 3.66378784e-01 8.62518907e-01 3.24001372e-01 2.79318810e-01 -5.10305822e-01 -4.29811895e-01 -8.99226725e-01 3.04083049e-01 -2.99185008e-01 -1.19619392e-01 -1.36319804e+00 -1.35337055e+00 -1.11108951e-01 2.99279153e-01 -1.45087421e+00 2.84005255e-01 -1.33287787e+00 -5.24626896e-02 8.71503174e-01 -1.80867338e+00 -1.36369932e+00 -6.35808587e-01 3.09774727e-01 -8.15659463e-02 7.27717951e-02 6.18869126e-01 -1.75120860e-01 -7.34021440e-02 1.70245349e-01 -6.73808679e-02 -3.41703296e-01 9.20454443e-01 -1.37658310e+00 -7.47484714e-02 3.94135326e-01 2.64930725e-01 4.16968018e-01 1.00972986e+00 -2.67754227e-01 -2.21135879e+00 -7.09614038e-01 2.29497507e-01 -1.24999070e+00 3.92737508e-01 -4.95960474e-01 -5.23773015e-01 2.88267255e-01 -1.87624320e-01 1.66908413e-01 3.72107714e-01 1.47667825e-02 -6.20104313e-01 -4.57735717e-01 -1.08357096e+00 4.16427404e-01 1.26227081e+00 -7.94595897e-01 -7.29525506e-01 6.40262425e-01 3.61420125e-01 -1.00152409e+00 -1.05676651e+00 7.54168510e-01 9.24750984e-01 -1.34095216e+00 1.67915201e+00 4.10667658e-02 -8.63453150e-02 -4.10271823e-01 -1.93982825e-01 -8.21329772e-01 -2.35214576e-01 -7.10998178e-01 -2.92403519e-01 9.38769042e-01 1.27486717e-02 -7.78839946e-01 9.23195481e-01 5.06701589e-01 -5.59125803e-02 -4.78748530e-01 -1.05994844e+00 -9.19919193e-01 -3.52703214e-01 -5.87900460e-01 4.44206834e-01 3.79766345e-01 -8.08706105e-01 2.50709325e-01 -1.70091376e-01 6.27415717e-01 1.18425071e+00 6.55605674e-01 1.09916747e+00 -1.90719211e+00 -3.41812432e-01 -3.12392235e-01 -7.37857401e-01 -9.83820140e-01 1.14310995e-01 -5.72394907e-01 3.45057957e-02 -1.05956626e+00 2.10143894e-01 -2.46753514e-01 1.55674160e-01 3.27961892e-02 6.72984272e-02 8.13243747e-01 -2.57948749e-02 3.72222036e-01 -5.75239897e-01 3.43401045e-01 1.34578657e+00 1.23003080e-01 -5.15188165e-02 1.00731380e-01 -1.63606793e-01 9.06115651e-01 5.27300298e-01 -3.20363075e-01 -1.10424943e-01 -4.09245670e-01 8.52939367e-01 -3.86642307e-01 8.29246521e-01 -1.16404045e+00 -2.99321637e-02 -3.00891280e-01 4.63962704e-01 -8.29728067e-01 1.00688207e+00 -1.18472087e+00 3.05642664e-01 2.63917595e-01 -7.29085803e-02 -6.67862773e-01 3.30296755e-02 7.17767537e-01 4.69331950e-01 -1.17007315e-01 1.09772599e+00 -3.00169349e-01 -6.10728264e-01 4.28654552e-01 2.05364406e-01 1.09906584e-01 9.11750376e-01 -6.94091737e-01 -4.65292484e-01 8.80104955e-03 -4.03564572e-01 -2.55916238e-01 8.02507699e-01 3.39251906e-01 4.13076133e-01 -1.20115006e+00 -5.87574363e-01 -6.45799786e-02 7.51424193e-01 8.90970230e-02 -2.42395654e-01 1.00700998e+00 -7.03452587e-01 4.14145768e-01 2.02755239e-02 -1.38299954e+00 -1.60954022e+00 1.64753512e-01 6.07300222e-01 4.84273195e-01 -3.79624248e-01 1.06643176e+00 -9.52151790e-02 -5.90328574e-01 1.52943596e-01 -3.03358972e-01 1.36009410e-01 7.19738826e-02 4.32137260e-03 6.27100825e-01 3.61064166e-01 -9.08093631e-01 -4.95955914e-01 1.25644767e+00 3.84179920e-01 -7.24989772e-02 1.38891566e+00 -3.84329677e-01 -1.13010019e-01 6.11436605e-01 9.27524924e-01 3.30723636e-02 -1.97821200e+00 -3.12667102e-01 -2.95511633e-01 -9.68439400e-01 1.22870803e-01 -9.52366292e-01 -7.00239837e-01 6.91532314e-01 8.01826179e-01 -5.34028150e-02 8.23666930e-01 4.10076946e-01 2.17460111e-01 5.02554536e-01 7.35540271e-01 -1.28962207e+00 4.63203043e-01 5.16264498e-01 8.59740615e-01 -1.52089047e+00 4.42114085e-01 -7.84748316e-01 -2.73274511e-01 1.04625297e+00 6.37059987e-01 -7.68195167e-02 5.69229960e-01 2.65062720e-01 1.23953685e-01 -3.93921882e-01 -2.63072580e-01 -3.22022378e-01 9.59774852e-01 9.73728240e-01 7.44766146e-02 -8.16383362e-02 1.35898173e-01 -1.23426497e-01 -3.09677660e-01 -6.04828119e-01 6.16988875e-02 7.82310665e-01 -4.19760972e-01 -9.10368681e-01 -9.57933426e-01 2.46735409e-01 -3.49217564e-01 2.01304123e-01 -6.88339233e-01 4.21062976e-01 3.79497439e-01 6.97128892e-01 -1.17057888e-02 -3.12255081e-02 3.67805213e-01 8.47955048e-02 1.53084421e+00 -7.23712087e-01 -3.79478991e-01 1.44934177e-01 2.09081829e-01 -8.88624191e-01 -1.08402407e+00 -7.77018905e-01 -8.12479019e-01 8.69649798e-02 -5.43624282e-01 -6.94917619e-01 8.71327102e-01 8.17419350e-01 9.38121527e-02 4.76179039e-03 5.13812482e-01 -1.51050031e+00 -4.87751156e-01 -7.80939877e-01 -4.26424652e-01 5.91467500e-01 3.51242542e-01 -1.19251478e+00 -5.54885864e-01 -5.61108142e-02]
[7.205202579498291, -2.365288019180298]
35269bb0-75a4-44a6-bc73-e20bc05ccdd2
rethinking-of-pedestrian-attribute-1
2107.03576
null
https://arxiv.org/abs/2107.03576v2
https://arxiv.org/pdf/2107.03576v2.pdf
Rethinking of Pedestrian Attribute Recognition: A Reliable Evaluation under Zero-Shot Pedestrian Identity Setting
Pedestrian attribute recognition aims to assign multiple attributes to one pedestrian image captured by a video surveillance camera. Although numerous methods are proposed and make tremendous progress, we argue that it is time to step back and analyze the status quo of the area. We review and rethink the recent progress from three perspectives. First, given that there is no explicit and complete definition of pedestrian attribute recognition, we formally define and distinguish pedestrian attribute recognition from other similar tasks. Second, based on the proposed definition, we expose the limitations of the existing datasets, which violate the academic norm and are inconsistent with the essential requirement of practical industry application. Thus, we propose two datasets, PETA\textsubscript{$ZS$} and RAP\textsubscript{$ZS$}, constructed following the zero-shot settings on pedestrian identity. In addition, we also introduce several realistic criteria for future pedestrian attribute dataset construction. Finally, we reimplement existing state-of-the-art methods and introduce a strong baseline method to give reliable evaluations and fair comparisons. Experiments are conducted on four existing datasets and two proposed datasets to measure progress on pedestrian attribute recognition.
['Kaiqi Huang', 'Xiaotang Chen', 'Houjing Huang', 'Jian Jia']
2021-07-08
null
null
null
null
['pedestrian-attribute-recognition']
['computer-vision']
[ 4.16944355e-01 -3.74768406e-01 -1.75402626e-01 -9.62962925e-01 -4.95757520e-01 -4.61850196e-01 8.05593431e-01 1.17153101e-01 -4.31559563e-01 9.31528270e-01 1.44108906e-01 -1.05070725e-01 7.74616152e-02 -8.18438530e-01 -4.24997061e-01 -7.24822998e-01 2.48399805e-02 3.55042964e-01 7.40024596e-02 -8.26362222e-02 2.16035500e-01 2.85355747e-01 -2.08657932e+00 3.34119767e-01 5.23594916e-01 1.16946757e+00 -2.70564139e-01 5.84171176e-01 1.44834235e-01 7.46127665e-01 -6.00639701e-01 -1.19060588e+00 3.43500227e-01 -2.17722699e-01 -9.48159397e-01 4.05776769e-01 1.01958144e+00 -3.94766033e-01 -3.89755040e-01 1.01442611e+00 5.75289786e-01 1.21538557e-01 9.18662190e-01 -1.85959148e+00 -9.32400525e-01 1.10917771e-02 -2.73759753e-01 4.99595292e-02 7.16128409e-01 3.18972409e-01 9.44260418e-01 -8.30972433e-01 4.09364969e-01 1.10361266e+00 8.98742318e-01 6.87871695e-01 -9.51587200e-01 -6.06384099e-01 5.24408758e-01 6.84828162e-01 -1.62640941e+00 -6.20939851e-01 5.25653362e-01 -5.33766031e-01 7.15250313e-01 6.61711454e-01 8.39949787e-01 1.66134083e+00 -3.69097978e-01 9.83503819e-01 1.15938473e+00 -3.42301697e-01 8.30467138e-03 1.76997080e-01 4.56641942e-01 4.35515702e-01 4.60692972e-01 2.18863636e-02 -2.98512638e-01 -1.91441223e-01 3.21529746e-01 -2.45154247e-01 -1.70816341e-03 -6.13570631e-01 -1.29805589e+00 5.89164793e-01 -1.43612340e-01 -3.08577329e-01 -2.32848495e-01 -2.46504948e-01 7.43323445e-01 1.27415553e-01 9.82798710e-02 -2.04557348e-02 -7.39161149e-02 -2.30056614e-01 -5.44660032e-01 2.89473772e-01 7.18766868e-01 1.57176721e+00 7.52968192e-01 5.24635464e-02 -4.37636413e-02 8.07269871e-01 1.78379416e-01 6.13285005e-01 -4.56004925e-02 -1.14652097e+00 2.02615753e-01 4.94126171e-01 2.40873516e-01 -8.99655044e-01 5.17122187e-02 2.09368691e-01 -8.04884672e-01 2.59377778e-01 5.48344970e-01 1.04154609e-02 -6.40410066e-01 1.39024389e+00 2.13889033e-01 1.85218051e-01 3.87252010e-02 8.59139919e-01 9.86674845e-01 3.06461304e-01 7.42140532e-01 1.10738136e-01 1.73706555e+00 -7.28671849e-01 -5.06433904e-01 -2.08893582e-01 2.19281226e-01 -7.80129194e-01 1.06965220e+00 1.48455888e-01 -6.35705709e-01 -7.67438829e-01 -1.30700493e+00 2.14657590e-01 -7.98990548e-01 2.69066364e-01 8.72073770e-01 1.35076988e+00 -7.49447763e-01 6.08508326e-02 -3.16035956e-01 -9.06074047e-01 5.63712895e-01 4.04634088e-01 -6.73389733e-01 6.75938949e-02 -1.22918010e+00 9.74451125e-01 1.94854707e-01 -9.60653052e-02 -8.84545028e-01 -3.77652854e-01 -8.13677192e-01 -4.01093096e-01 1.94274843e-01 -1.05430651e+00 1.01933265e+00 -9.04826701e-01 -1.20436561e+00 1.54933739e+00 -3.38340133e-01 -4.09761578e-01 5.55106282e-01 -1.93991601e-01 -8.18960130e-01 3.74004021e-02 3.75445247e-01 7.51164317e-01 4.93790716e-01 -1.44629705e+00 -1.23739314e+00 -4.00033593e-01 4.03676778e-01 1.28356442e-01 -2.70843148e-01 4.01962876e-01 -1.86295077e-01 -6.13720953e-01 -3.00105542e-01 -9.23520982e-01 -1.16511963e-01 2.82581430e-02 -5.17863333e-01 -1.79980785e-01 8.16246510e-01 -2.51176864e-01 1.09620810e+00 -2.08197308e+00 -4.52350348e-01 -3.68853584e-02 1.28706872e-01 3.74407858e-01 1.64014831e-01 2.34120250e-01 -2.05218196e-01 6.21918403e-02 -3.25192332e-01 -3.84392709e-01 2.46854872e-01 4.16943014e-01 -1.30845770e-01 6.82303369e-01 7.64239803e-02 6.09722197e-01 -8.60165775e-01 -7.84537613e-01 4.07770544e-01 4.20221418e-01 -2.89814435e-02 1.14116848e-01 3.58510911e-01 2.56296396e-01 -6.03533447e-01 1.25491059e+00 7.69145727e-01 3.41033548e-01 -1.32947892e-01 -4.27568078e-01 -2.84852594e-01 -1.13949604e-01 -1.41011405e+00 1.22504914e+00 2.16482490e-01 7.08518863e-01 -1.59712464e-01 -1.21276283e+00 1.07016015e+00 3.37578118e-01 6.71852589e-01 -3.83082867e-01 -6.88856235e-03 1.03269227e-01 -3.78180653e-01 -5.70105851e-01 7.44040310e-01 1.08823933e-01 -3.99735391e-01 1.01398237e-01 -2.36435458e-02 4.18493509e-01 3.19606811e-01 -1.29467733e-02 5.87836683e-01 2.70570397e-01 6.72751367e-01 -5.07243931e-01 1.10941148e+00 5.55468276e-02 5.52446365e-01 9.50958371e-01 -1.13482463e+00 7.39027202e-01 2.81928629e-01 -1.11270285e+00 -1.41706264e+00 -1.33481610e+00 -2.36491591e-01 1.40567148e+00 2.61999190e-01 -4.08800960e-01 -1.01733947e+00 -8.92825425e-01 -1.55671731e-01 5.94124317e-01 -7.91663408e-01 7.90187195e-02 -5.89298129e-01 -1.11969721e+00 9.28480089e-01 7.94441879e-01 9.57017779e-01 -8.33374977e-01 -6.55187070e-01 -1.06126241e-01 -6.39492929e-01 -1.42105794e+00 -4.12095606e-01 -4.59805787e-01 -9.40611809e-02 -1.26679254e+00 -5.55222571e-01 -7.58532345e-01 5.37215769e-01 4.99557853e-01 1.45937026e+00 1.33634672e-01 -1.55128241e-01 8.98793638e-01 -4.53018427e-01 -6.51987493e-01 -1.60294548e-01 -2.23307707e-03 4.74066705e-01 3.75857800e-01 1.18782175e+00 -3.36675972e-01 -6.97612345e-01 7.96329916e-01 -2.59046465e-01 -4.29441221e-02 5.22215486e-01 5.74102581e-01 5.12245357e-01 -4.28330928e-01 6.00619793e-01 -7.44381964e-01 3.07466477e-01 -2.93257266e-01 -2.10994720e-01 3.38237852e-01 -4.75785971e-01 -3.68382126e-01 3.30219746e-01 -2.76959032e-01 -1.09856486e+00 3.96755785e-01 -2.73887694e-01 1.41041055e-01 -9.97893870e-01 -3.48456293e-01 -8.08732688e-01 -6.79986402e-02 6.14349782e-01 3.90864223e-01 -9.83052105e-02 -1.28792644e-01 3.11417490e-01 8.83517265e-01 9.75877941e-01 -9.83282864e-01 5.05807579e-01 7.90081680e-01 8.23140983e-03 -9.59008932e-01 -7.63279378e-01 -6.02129936e-01 -1.19494069e+00 -5.73791087e-01 1.02766585e+00 -8.82288575e-01 -1.14625156e+00 6.96682811e-01 -9.50167239e-01 1.79362029e-01 -2.33891323e-01 3.40387613e-01 -1.03821254e+00 6.39204323e-01 -2.43636161e-01 -8.07266235e-01 -2.09356368e-01 -1.20452952e+00 8.43390048e-01 1.03190012e-01 -3.74299824e-01 -6.43897772e-01 -4.04207975e-01 4.94769752e-01 3.26054454e-01 5.44381738e-01 5.88289201e-01 -7.30088830e-01 -3.97857934e-01 -3.21869671e-01 -4.01648134e-01 2.70749629e-01 1.26356333e-01 2.91910261e-01 -1.21671546e+00 -1.28195405e-01 -3.75476867e-01 -1.34492114e-01 5.81804752e-01 3.20859045e-01 9.54627335e-01 -2.35780135e-01 -5.16467690e-01 5.24372458e-01 9.45129573e-01 2.69900054e-01 9.31147277e-01 9.36145365e-01 5.94075322e-01 8.20207894e-01 7.50686049e-01 5.35281479e-01 9.36783612e-01 8.39847326e-01 2.94192761e-01 7.18183741e-02 -2.73360640e-01 -5.24706692e-02 2.41647199e-01 2.87048399e-01 -7.76838601e-01 -4.60537858e-02 -8.39388132e-01 4.17453945e-01 -1.79618263e+00 -1.52621889e+00 -4.72087651e-01 2.58304691e+00 3.67539108e-01 -2.60978825e-02 7.21807599e-01 2.20799506e-01 1.03548658e+00 1.60013754e-02 -1.25959724e-01 -2.26191550e-01 -5.26617765e-01 -3.50178331e-01 3.73867333e-01 2.73946524e-01 -1.85986316e+00 7.43864775e-01 7.10052395e+00 5.34610331e-01 -4.91281182e-01 -2.28554219e-01 9.35363233e-01 3.42500031e-01 4.07331049e-01 -1.13573231e-01 -1.04739344e+00 6.23661220e-01 8.06921363e-01 -2.99373209e-01 -1.25246391e-01 1.02082753e+00 -3.23230237e-01 5.53727672e-02 -1.43060517e+00 1.15158951e+00 1.76895469e-01 -9.59898651e-01 3.72027069e-01 -3.47415917e-02 1.66419134e-01 -6.00434005e-01 2.13049620e-01 4.04021263e-01 2.89315134e-01 -9.84637439e-01 8.14933717e-01 4.58877742e-01 8.36905301e-01 -8.12086344e-01 8.89002979e-01 -1.30491778e-01 -1.72001100e+00 6.16322570e-02 -5.60824990e-01 -1.24077044e-01 1.23377822e-01 -1.59916788e-01 -4.25414681e-01 7.31337428e-01 9.77846384e-01 6.63295329e-01 -8.83869052e-01 1.08455682e+00 5.25728390e-02 3.29313099e-01 3.29116434e-02 3.27288173e-02 2.10393578e-01 -3.25589865e-01 5.68448544e-01 1.71633828e+00 3.45648736e-01 4.33001429e-01 2.48591676e-01 3.86399835e-01 3.23694915e-01 -3.85850407e-02 -7.66718090e-01 5.08916318e-01 7.25949466e-01 1.21941447e+00 -5.14601111e-01 -5.62101424e-01 -9.28307116e-01 9.51047838e-01 -9.69154164e-02 3.18067282e-01 -1.01425147e+00 -1.26714855e-01 1.35475600e+00 1.19742289e-01 9.55581143e-02 -1.75320074e-01 -4.32833940e-01 -9.64525163e-01 -3.10858376e-02 -9.43827868e-01 8.15073311e-01 -4.29007828e-01 -1.60702181e+00 5.59207976e-01 3.59667838e-01 -1.62487924e+00 -2.19451562e-01 -7.44869351e-01 -3.70497316e-01 7.19800532e-01 -1.19078052e+00 -1.72426701e+00 -7.25865841e-01 7.43187726e-01 6.42643034e-01 -5.76159120e-01 1.14713395e+00 6.26760840e-01 -6.00109458e-01 9.11277711e-01 -3.45873982e-01 5.50343573e-01 8.50500166e-01 -1.14684558e+00 8.57668161e-01 7.04682529e-01 -2.82392889e-01 5.37551463e-01 9.61421728e-01 -3.69681180e-01 -1.17380905e+00 -1.06948411e+00 7.86733329e-01 -1.12557554e+00 4.11097258e-01 -3.56195033e-01 -7.31292725e-01 8.50075901e-01 4.22056586e-01 1.57172188e-01 1.13069117e+00 4.46411921e-03 -4.05208588e-01 -1.45810440e-01 -1.26372623e+00 5.51618636e-01 1.32619286e+00 -2.20908031e-01 -5.71179748e-01 -1.73621420e-02 8.35084766e-02 2.18533259e-02 -9.05664384e-01 4.88353133e-01 9.08701539e-01 -1.08063519e+00 1.71225190e+00 -8.43729854e-01 2.72774082e-02 -3.93181890e-01 -7.54948080e-01 -9.46701586e-01 -5.65415621e-01 -3.48768651e-01 1.29453242e-01 1.64142931e+00 1.81540236e-01 -5.54622531e-01 9.84154224e-01 9.81852651e-01 -2.06368566e-01 -4.95678127e-01 -9.75076735e-01 -1.00509262e+00 8.13050494e-02 -2.70384759e-01 9.08569157e-01 1.05740547e+00 -2.32556865e-01 2.69770294e-01 -6.52563572e-01 1.26646161e-01 1.29661465e+00 -2.16098964e-01 1.11041069e+00 -1.39016521e+00 3.99986982e-01 -5.54653466e-01 -1.07769489e+00 -8.93854558e-01 1.81463420e-01 -2.62300551e-01 -2.36092016e-01 -1.32968330e+00 5.73395252e-01 -4.91722077e-01 -2.45972693e-01 2.12048247e-01 -3.61515105e-01 4.87573594e-01 1.22071944e-01 1.96387470e-01 -8.10319006e-01 4.06530231e-01 7.07097054e-01 -3.72803003e-01 5.96881688e-01 1.70484811e-01 -7.74714470e-01 9.01590347e-01 8.18108976e-01 -7.05829561e-02 -2.30300084e-01 -1.06600203e-01 -3.78330320e-01 -5.81974387e-01 6.28283441e-01 -1.33248985e+00 2.60813326e-01 -3.65874261e-01 6.84949279e-01 -5.21880388e-01 5.89404821e-01 -8.23505878e-01 9.36766118e-02 5.04976474e-02 -4.97496277e-02 4.68466729e-01 -7.39177689e-02 4.50487942e-01 -1.26835153e-01 -1.65063336e-01 9.43416238e-01 -2.28440002e-01 -1.46531320e+00 5.06566048e-01 -6.08305633e-01 3.84698994e-03 1.40270782e+00 -9.40989673e-01 -4.53707486e-01 -3.98646206e-01 -5.92237413e-01 4.06827591e-02 6.77843690e-01 4.77759689e-01 7.09003866e-01 -1.68466640e+00 -9.89392102e-01 1.72724113e-01 5.68287790e-01 -8.42006445e-01 7.36664161e-02 5.12442708e-01 -3.25447023e-01 4.45001423e-01 -8.76251757e-01 -5.70184886e-01 -1.56934369e+00 7.98437893e-01 1.26352295e-01 4.71108675e-01 -5.75561523e-01 5.48896015e-01 2.78927535e-01 -3.25935960e-01 1.86821476e-01 4.46372658e-01 -5.13490081e-01 7.75676966e-02 6.35056734e-01 7.18010545e-01 -1.72313631e-01 -1.35741746e+00 -5.93592286e-01 4.13746387e-01 -7.34725734e-03 1.33042812e-01 8.24406505e-01 -4.95368153e-01 1.35838181e-01 1.92977875e-01 9.25101757e-01 -2.53068566e-01 -1.30664444e+00 1.99503265e-03 1.87088326e-01 -7.50592589e-01 -8.00819337e-01 -4.37564760e-01 -7.26617157e-01 6.68803692e-01 8.67437065e-01 2.30455384e-01 9.99183714e-01 -1.73129067e-01 5.93527675e-01 4.60222602e-01 5.35328925e-01 -1.28135788e+00 -2.25388348e-01 4.08486664e-01 4.68875289e-01 -1.51858604e+00 2.24570259e-01 -8.02105606e-01 -9.58991349e-01 9.39561367e-01 9.35121357e-01 2.61128664e-01 2.13017553e-01 2.13627309e-01 1.34710550e-01 1.30454049e-01 -2.09212527e-01 -3.18807840e-01 8.93591568e-02 1.33970630e+00 5.81753135e-01 3.25184226e-01 -1.24049097e-01 6.86990261e-01 -2.35664964e-01 -2.89201051e-01 5.88187516e-01 8.43851268e-01 -4.78070378e-01 -1.39539635e+00 -5.53354502e-01 3.17023039e-01 -3.25772017e-01 1.69250816e-01 -5.45984507e-01 1.01540339e+00 2.76022226e-01 1.07967865e+00 -4.81273495e-02 -5.25967360e-01 6.59452915e-01 1.56381130e-01 4.54339504e-01 3.70500758e-02 -2.14726672e-01 -7.71067500e-01 6.07992291e-01 -2.45573387e-01 -7.83026934e-01 -1.02742612e+00 -5.36138773e-01 -8.20081472e-01 2.79665411e-01 -7.27905035e-02 1.55132994e-01 6.44737363e-01 1.74642950e-01 -3.17696109e-02 2.82423228e-01 -6.75882518e-01 -2.45863318e-01 -6.87959850e-01 -2.08174989e-01 9.08105135e-01 -4.05013599e-02 -1.01177597e+00 7.88161755e-02 7.22770691e-01]
[14.519291877746582, 0.9557169079780579]
74cc0d7c-8644-4f37-9386-a30d6b4f0873
sparse-coding-for-alpha-matting
1604.02898
null
http://arxiv.org/abs/1604.02898v1
http://arxiv.org/pdf/1604.02898v1.pdf
Sparse Coding for Alpha Matting
Existing color sampling based alpha matting methods use the compositing equation to estimate alpha at a pixel from pairs of foreground (F) and background (B) samples. The quality of the matte depends on the selected (F,B) pairs. In this paper, the matting problem is reinterpreted as a sparse coding of pixel features, wherein the sum of the codes gives the estimate of the alpha matte from a set of unpaired F and B samples. A non-parametric probabilistic segmentation provides a certainty measure on the pixel belonging to foreground or background, based on which a dictionary is formed for use in sparse coding. By removing the restriction to conform to (F,B) pairs, this method allows for better alpha estimation from multiple F and B samples. The same framework is extended to videos, where the requirement of temporal coherence is handled effectively. Here, the dictionary is formed by samples from multiple frames. A multi-frame graph model, as opposed to a single image as for image matting, is proposed that can be solved efficiently in closed form. Quantitative and qualitative evaluations on a benchmark dataset are provided to show that the proposed method outperforms current state-of-the-art in image and video matting.
['Hisham Cholakkal', 'Ehsan Shahrian Varnousfaderani', 'Jubin Johnson', 'Deepu Rajan']
2016-04-11
null
null
null
null
['video-matting']
['computer-vision']
[ 5.25907815e-01 -1.93746641e-01 -1.90244377e-01 -2.29847655e-01 -6.58155560e-01 -2.37342626e-01 4.28372800e-01 4.95559245e-04 -1.88221321e-01 7.03677356e-01 -6.91910684e-02 2.56461442e-01 1.54927105e-01 -8.61056268e-01 -7.95773625e-01 -1.15356445e+00 1.94647759e-01 3.14886868e-01 2.20758885e-01 2.47002721e-01 8.80505443e-02 2.99893230e-01 -1.50080085e+00 5.70750654e-01 8.27978194e-01 1.12356997e+00 2.17173010e-01 5.42449594e-01 -3.07595193e-01 7.23829150e-01 -5.10653198e-01 -4.33396548e-01 4.79071498e-01 -8.67558122e-01 -4.70308632e-01 6.48948848e-01 7.94134200e-01 -3.38136047e-01 -2.52840787e-01 1.30571353e+00 -9.10758525e-02 5.43398559e-02 6.74932659e-01 -1.19694841e+00 -2.15981796e-01 3.58559787e-01 -1.03918481e+00 2.36894693e-02 4.14674699e-01 -8.87828171e-02 7.36257672e-01 -8.89924824e-01 5.53159952e-01 1.09473538e+00 3.74324739e-01 -9.28446371e-03 -1.41189241e+00 -3.62818152e-01 1.26149431e-01 2.21813798e-01 -1.48331749e+00 -2.93293983e-01 8.66061151e-01 -6.15748405e-01 3.38237464e-01 4.00227487e-01 1.14535654e+00 5.93694627e-01 3.51689100e-01 8.50176573e-01 1.37805915e+00 -6.19124174e-01 4.31606203e-01 -3.29742730e-02 8.64025205e-02 9.39330399e-01 3.49068999e-01 -9.78802592e-02 -8.17474902e-01 -2.74277419e-01 9.01471138e-01 -3.85849439e-02 -4.25734371e-01 -5.73403895e-01 -1.25990772e+00 7.86921263e-01 3.65040839e-01 2.45520204e-01 -5.53613842e-01 1.22814201e-01 1.20263085e-01 -3.63415480e-02 4.42007065e-01 -3.88893448e-02 1.36935592e-01 6.85600713e-02 -1.58080685e+00 7.55450502e-03 7.86543608e-01 8.85075152e-01 1.22604394e+00 4.06065494e-01 -1.43947124e-01 6.41147256e-01 3.33847284e-01 7.72722483e-01 6.46927133e-02 -1.24037778e+00 1.09773844e-01 4.19386804e-01 1.45981044e-01 -1.30024934e+00 2.33404756e-01 -3.97176258e-02 -9.85014439e-01 3.81349653e-01 6.42873228e-01 1.80602014e-01 -1.29214323e+00 1.48685741e+00 4.64625776e-01 6.00181282e-01 -1.69620171e-01 1.01318729e+00 4.76968050e-01 9.32154894e-01 -2.97629863e-01 -6.81793809e-01 1.28037393e+00 -8.15814376e-01 -9.88063693e-01 -1.42155319e-01 -1.85437128e-01 -9.47222590e-01 5.90247095e-01 7.80942798e-01 -1.22409320e+00 -5.64935088e-01 -1.07603562e+00 4.45954502e-02 3.71067300e-02 1.20363757e-01 5.04628241e-01 7.29966998e-01 -8.84449303e-01 3.46493393e-01 -8.88590932e-01 -1.13793977e-01 1.91265985e-01 3.11818331e-01 -2.23412797e-01 -3.22651505e-01 -7.38264799e-01 8.03494930e-01 2.19405025e-01 2.87686199e-01 -1.06637537e+00 -3.71644169e-01 -8.71278822e-01 -2.09166080e-01 1.38365284e-01 -8.65635514e-01 5.07974446e-01 -1.60987937e+00 -1.32285833e+00 7.96155512e-01 -4.84120071e-01 -4.68999505e-01 5.22944570e-01 -1.69463322e-01 -1.99163049e-01 7.21647561e-01 2.14659199e-01 7.32088327e-01 1.48734403e+00 -1.72907603e+00 -6.28699064e-01 -1.93895549e-01 -4.02241983e-02 1.85189858e-01 5.99237718e-03 -1.59526780e-01 -7.24612296e-01 -7.64194250e-01 3.71479630e-01 -8.98890913e-01 -1.35075465e-01 2.76582778e-01 -2.90196300e-01 5.16882837e-01 9.05734241e-01 -1.02478385e+00 1.11427581e+00 -2.26360369e+00 4.91304517e-01 6.54317737e-01 3.64983797e-01 -9.64164585e-02 9.78015959e-02 4.65716273e-02 3.66003215e-02 -5.34871578e-01 -7.40594685e-01 -2.29684964e-01 -5.62612772e-01 4.13741708e-01 -5.11175916e-02 8.90382826e-01 3.06137875e-02 3.49212170e-01 -9.74331796e-01 -8.02332401e-01 4.77288276e-01 5.97040772e-01 -3.36774409e-01 2.82285839e-01 -2.11516768e-01 4.34993207e-01 -2.44765371e-01 6.26423836e-01 8.90395761e-01 -5.27201667e-02 1.63519800e-01 -5.66714048e-01 -1.05402604e-01 -3.28397155e-01 -1.37470412e+00 1.74281180e+00 -2.22726002e-01 5.68950236e-01 4.06847388e-01 -8.13896596e-01 9.21823561e-01 2.04688847e-01 8.94100666e-01 -2.58246124e-01 1.47149071e-01 1.74632579e-01 -1.16920017e-01 -1.83900312e-01 3.31893295e-01 -5.21357879e-02 2.93377221e-01 2.96288729e-01 6.35874495e-02 -2.94251531e-01 5.59347928e-01 4.17318881e-01 9.76831794e-01 1.48800969e-01 3.21653366e-01 -3.66892695e-01 5.66451252e-01 -2.50503817e-03 7.09358156e-01 5.69440246e-01 -9.93741825e-02 8.45043600e-01 3.03176761e-01 -4.12956446e-01 -1.10857224e+00 -1.10225463e+00 -8.00765604e-02 4.00903106e-01 5.13902068e-01 -3.71846318e-01 -1.09763908e+00 -4.64424461e-01 -7.05003291e-02 4.79660362e-01 -8.52352321e-01 9.73929316e-02 -6.12368405e-01 -5.83798468e-01 3.44253145e-02 1.65406227e-01 6.32429659e-01 -6.11760378e-01 -6.26590371e-01 1.08413428e-01 -5.67244828e-01 -1.01087165e+00 -6.18470967e-01 1.30432189e-01 -9.24538732e-01 -1.01464438e+00 -7.83475339e-01 -6.69343531e-01 9.66730058e-01 5.93426526e-01 1.01857984e+00 1.13503270e-01 -1.02077253e-01 5.30550659e-01 -4.04858410e-01 2.01382965e-01 -3.76542866e-01 -5.92907250e-01 -1.90056369e-01 7.62265325e-01 9.48707946e-03 -4.31305557e-01 -5.08725464e-01 3.21639597e-01 -1.01860905e+00 4.46305126e-01 6.30257607e-01 9.50192153e-01 1.01886213e+00 -8.95683244e-02 9.73339286e-03 -8.29155624e-01 -1.42694458e-01 -3.36927295e-01 -5.56452811e-01 3.11007082e-01 -2.65640736e-01 -2.58076517e-03 3.62194091e-01 -4.37456787e-01 -1.18206692e+00 4.28869486e-01 3.86167228e-01 -8.26798081e-01 5.59135899e-03 4.00155604e-01 -2.30057314e-01 -2.86866248e-01 3.41544002e-01 2.36985788e-01 2.13916257e-01 -1.84929684e-01 5.31951845e-01 1.97547033e-01 7.20399678e-01 -6.23254955e-01 7.96079755e-01 7.80719459e-01 1.35223344e-01 -9.53519642e-01 -7.40437686e-01 -5.42575359e-01 -6.75149858e-01 -6.50266647e-01 1.00768137e+00 -9.54045177e-01 -1.20291375e-01 4.09793764e-01 -1.05090404e+00 -1.35032713e-01 -3.58722836e-01 5.86327076e-01 -7.23032475e-01 7.22347379e-01 -6.56374454e-01 -7.85085320e-01 -7.01614395e-02 -1.29214156e+00 1.00846434e+00 7.52988607e-02 -1.61644191e-01 -7.49834657e-01 -1.83469635e-02 5.92312574e-01 -6.04359619e-03 3.94471884e-01 6.83997989e-01 1.64523363e-01 -8.21467578e-01 -1.19706370e-01 -1.74456581e-01 5.25305808e-01 2.67841697e-01 3.30078363e-01 -8.26526880e-01 -3.18124861e-01 3.35163951e-01 6.62409961e-02 1.12461174e+00 7.96401799e-01 8.39019001e-01 -4.65091020e-01 -2.37747654e-01 6.82692707e-01 1.48275185e+00 1.82374671e-01 8.72764230e-01 1.73825920e-01 8.79188001e-01 3.56459469e-01 9.15032923e-01 5.88152945e-01 -3.39832366e-03 6.50003433e-01 3.72018337e-01 -1.95557013e-01 -3.64392579e-01 1.18984915e-01 5.96732557e-01 8.86300564e-01 -1.31618872e-01 -1.01808093e-01 -5.10857761e-01 4.03645515e-01 -1.85144901e+00 -9.00869012e-01 -3.90112579e-01 2.33438635e+00 9.63556647e-01 -9.02416483e-02 2.04578474e-01 2.41695598e-01 1.02574420e+00 2.21445784e-01 -2.39052340e-01 -1.76116213e-01 -3.07462722e-01 2.16915786e-01 3.75627339e-01 7.12657392e-01 -1.02351296e+00 5.43960154e-01 6.10364199e+00 9.92540061e-01 -1.03776169e+00 8.26578587e-02 8.79204750e-01 7.85005987e-02 -4.28352147e-01 2.85569370e-01 -4.99833882e-01 6.91638529e-01 5.79719126e-01 3.61723155e-02 5.38033485e-01 5.24719059e-01 1.94084853e-01 -7.41629004e-01 -9.32130098e-01 1.02136827e+00 4.37174261e-01 -1.18553507e+00 1.18483163e-01 -9.88909788e-03 9.91082728e-01 -4.86892223e-01 -8.27415287e-02 -4.45188075e-01 3.11048944e-02 -7.15372503e-01 1.11037457e+00 6.56760097e-01 7.08569586e-01 -6.98888063e-01 6.17126286e-01 -5.49781360e-02 -1.27849793e+00 1.45997852e-01 -2.12019607e-01 1.96208730e-01 3.53323460e-01 1.22443974e+00 -5.53641140e-01 6.33135676e-01 6.16315961e-01 7.49488056e-01 -4.31792647e-01 1.01781571e+00 -1.20875262e-01 5.94030678e-01 -3.92163128e-01 3.92901599e-01 -3.63824656e-03 -8.35737467e-01 5.97838521e-01 1.25081229e+00 3.59506756e-01 7.22945482e-02 4.29694682e-01 8.02231669e-01 2.84805208e-01 8.09365958e-02 -3.15112263e-01 7.78663233e-02 2.46686578e-01 1.43317389e+00 -1.16598248e+00 -6.12071812e-01 -4.85917181e-01 1.12882698e+00 -1.95115283e-02 4.67510730e-01 -7.86979437e-01 1.78933322e-01 3.87197793e-01 2.97192127e-01 3.14128935e-01 -3.25654000e-01 -5.45568347e-01 -1.23334777e+00 -1.78504974e-01 -1.08754516e+00 3.08919698e-01 -7.56358802e-01 -1.09568131e+00 4.36267644e-01 1.67694524e-01 -1.40228176e+00 -5.46989515e-02 -2.86329031e-01 -3.70219737e-01 6.99005961e-01 -1.04328346e+00 -1.21108365e+00 -5.32932043e-01 7.15589881e-01 4.37960446e-01 -3.07805091e-02 5.51614702e-01 2.60900795e-01 -3.62676710e-01 1.53838441e-01 7.98358321e-02 -3.33701931e-02 5.18641651e-01 -1.22661865e+00 -2.52748907e-01 1.31685734e+00 4.21238542e-01 4.01628852e-01 8.42394829e-01 -9.17552531e-01 -1.45436406e+00 -9.78193283e-01 4.10936713e-01 -5.83186606e-03 3.23918849e-01 -1.29033208e-01 -8.60187888e-01 4.93847579e-01 4.20479000e-01 1.33113384e-01 5.37066102e-01 -3.04364055e-01 -2.13200226e-01 -3.18260223e-01 -1.21725512e+00 3.62865984e-01 4.93477792e-01 -2.44616494e-01 -1.04196064e-01 1.13848530e-01 3.03643644e-01 -3.07818919e-01 -8.42430294e-01 2.38272995e-01 5.31043589e-01 -1.08641112e+00 8.57682526e-01 7.69603848e-02 3.25872302e-01 -8.86466265e-01 -3.16564023e-01 -1.10077870e+00 -3.11585963e-01 -5.04008830e-01 -4.30750340e-01 1.08331621e+00 -2.05320716e-02 -6.20841719e-02 7.29501188e-01 2.54154593e-01 -1.06918365e-01 -5.02285421e-01 -1.04468381e+00 -4.71067369e-01 -5.13686478e-01 -7.54125491e-02 1.43481353e-02 8.94750714e-01 -3.60983789e-01 3.90506983e-02 -6.38696373e-01 1.83642432e-01 1.01554799e+00 4.34336841e-01 6.76573992e-01 -8.77870262e-01 -4.85020369e-01 -9.60292071e-02 -5.30593991e-01 -8.65740955e-01 8.06330815e-02 -5.36731005e-01 2.71730244e-01 -1.32970631e+00 4.76192415e-01 -3.73918205e-01 -2.85637304e-02 2.15392023e-01 -2.36686572e-01 5.81999421e-01 2.94911087e-01 3.00580740e-01 -3.51073861e-01 5.30578256e-01 1.23242772e+00 -4.14340019e-01 2.95496732e-02 -1.96557477e-01 -1.72378510e-01 6.47512734e-01 4.84573841e-01 -4.31151748e-01 -2.47528717e-01 -2.28148028e-01 -3.68816592e-02 3.61197770e-01 1.86349094e-01 -1.03596592e+00 1.04008690e-01 -3.23278099e-01 6.76357687e-01 -4.17689443e-01 7.08064556e-01 -7.84782171e-01 5.83632886e-01 3.37562650e-01 -5.74474148e-02 -1.98466182e-02 -9.45942700e-02 7.41822004e-01 -3.47329944e-01 -5.21732748e-01 1.11273730e+00 -1.80283785e-01 -4.68176603e-01 2.60792136e-01 -3.91721785e-01 -3.63339514e-01 1.10830319e+00 -5.20846426e-01 8.85945633e-02 -5.68400323e-01 -7.60906637e-01 -3.19771260e-01 7.10450768e-01 -1.87396199e-01 8.99433196e-01 -1.36537802e+00 -7.04798520e-01 3.71327341e-01 -2.26230979e-01 4.65469584e-02 2.09255710e-01 1.06326485e+00 -7.38809824e-01 -2.18178362e-01 -4.25712675e-01 -8.75069022e-01 -1.30148840e+00 4.27904785e-01 2.77424995e-02 -1.04259692e-01 -5.47593832e-01 7.03970909e-01 4.32943642e-01 5.09361506e-01 7.74522349e-02 -2.67366171e-01 1.27420291e-01 1.99156836e-01 4.36731666e-01 4.12853509e-01 -4.74202037e-02 -1.01638222e+00 -1.61375791e-01 4.79161710e-01 1.96014512e-02 -4.72467571e-01 1.07906866e+00 -2.64271259e-01 -4.89062756e-01 6.49544120e-01 8.97064924e-01 1.71047762e-01 -1.44151783e+00 -7.59591535e-02 -3.92944813e-01 -1.00926983e+00 2.62929797e-01 -4.59385067e-01 -1.46124434e+00 6.89366460e-01 5.89605212e-01 5.10000624e-02 1.32935333e+00 -2.99223930e-01 6.42340720e-01 -1.52014434e-01 4.54103619e-01 -9.50060666e-01 2.81393558e-01 8.55476484e-02 7.43014216e-01 -9.46392417e-01 2.75162786e-01 -6.83236182e-01 -6.68251216e-01 1.16342843e+00 3.55640948e-01 -4.00010318e-01 5.00831962e-01 4.66972500e-01 -2.76684724e-02 -3.09489220e-02 -3.15764040e-01 -1.03667840e-01 3.11433047e-01 5.17628729e-01 3.13164264e-01 4.05291580e-02 -2.40658104e-01 -2.97991652e-02 1.67514980e-01 -1.86478063e-01 6.63828015e-01 9.11870360e-01 -5.81333280e-01 -1.00392735e+00 -7.96865940e-01 5.74296713e-01 -2.32428923e-01 -4.19760048e-02 -1.86518326e-01 3.90246630e-01 2.49273524e-01 8.63724947e-01 1.42218202e-01 -1.77905738e-01 -2.73470283e-01 -1.81228980e-01 9.17910755e-01 -5.16871452e-01 -3.95170391e-01 6.21701181e-01 -1.92535426e-02 -5.11348069e-01 -9.26874220e-01 -8.55925977e-01 -1.06036818e+00 -2.89495111e-01 -5.21579266e-01 1.79308787e-01 5.77111304e-01 8.44741940e-01 -1.18329950e-01 1.75846666e-01 6.99692488e-01 -1.04843497e+00 8.96603167e-02 -6.15186572e-01 -9.94419813e-01 4.37048942e-01 2.60447383e-01 -8.56636703e-01 -3.04889798e-01 6.36295855e-01]
[10.791590690612793, -1.6867763996124268]
5d09561c-af38-4f58-932a-45258b17d61c
spatio-temporal-prediction-in-video-coding-by-2
2207.09727
null
https://arxiv.org/abs/2207.09727v1
https://arxiv.org/pdf/2207.09727v1.pdf
Spatio-temporal prediction in video coding by best approximation
Within the scope of this contribution we propose a novel efficient spatio-temporal prediction algorithm for video coding. The algorithm operates in two stages. First, motion compensation is performed on the block to be predicted in order to exploit temporal correlations. Afterwards, in order to exploit spatial correlations, this preliminary estimate is spatially refined by forming a joint model of the motion compensated block and spatially adjacent already decoded blocks. Compared to an earlier refinement algorithm, the novel one only needs very little iteration, leading to a speedup of factor 17. The implementation of this new algorithm into the H.264/AVC leads to a maximum reduction in data rate of up to nearly 13% for the considered sequences.
['André Kaup', 'Haricharan Lakshman', 'Jürgen Seiler']
2022-07-20
null
null
null
null
['motion-compensation']
['computer-vision']
[ 5.61364710e-01 1.86739698e-01 3.03705446e-02 -1.18437164e-01 -2.48359576e-01 -4.85606417e-02 4.24024254e-01 4.92151201e-01 -6.08487666e-01 7.73840249e-01 1.63142085e-01 -1.90787047e-01 1.51987355e-02 -5.31643271e-01 -3.13846439e-01 -8.15866351e-01 -3.09135258e-01 7.32690736e-04 9.74240065e-01 1.98626310e-01 4.88213032e-01 7.44741559e-01 -1.51904356e+00 4.84577805e-01 5.40674806e-01 1.11613297e+00 6.12534583e-01 1.03341246e+00 2.41010129e-01 1.14019012e+00 -1.34374574e-01 7.57823214e-02 1.13365561e-01 -7.17182994e-01 -9.23657835e-01 4.30584818e-01 9.36342627e-02 -4.46394026e-01 -2.13498250e-01 8.30897808e-01 1.07523367e-01 2.09588766e-01 4.90918398e-01 -4.20856357e-01 4.80881304e-01 4.40223128e-01 -7.36740172e-01 4.37431723e-01 3.44036847e-01 -1.48509338e-01 7.30075419e-01 -9.59596872e-01 7.01696336e-01 7.32132733e-01 3.96290809e-01 2.67779976e-01 -1.29245222e+00 -2.30853111e-01 -2.39402220e-01 5.08308828e-01 -1.49309766e+00 -7.25956798e-01 6.46635711e-01 -4.02765572e-01 1.03166032e+00 1.77677736e-01 1.04219472e+00 1.82973146e-01 3.43163937e-01 5.49240649e-01 8.44461322e-01 -6.59592390e-01 4.00398374e-01 -2.89601535e-01 -1.20019205e-01 4.23483521e-01 -1.69941649e-01 1.78517982e-01 -3.47314447e-01 6.73159510e-02 9.81634200e-01 -2.23031923e-01 -3.91429961e-01 -4.68282104e-01 -1.28738701e+00 5.27081072e-01 3.78750354e-01 8.20192814e-01 -7.36873269e-01 2.89471805e-01 4.30884421e-01 8.51028562e-02 4.47698206e-01 -8.68000016e-02 -1.03346772e-01 -2.39059612e-01 -1.73828459e+00 1.11971065e-01 5.76739967e-01 6.87638819e-01 8.27442706e-01 8.44980329e-02 9.95753855e-02 6.07178509e-01 5.61330616e-01 6.46236762e-02 2.04260945e-01 -1.09838581e+00 4.46170598e-01 2.75507450e-01 2.21859396e-01 -1.01496720e+00 -3.53547215e-01 -2.31447265e-01 -1.12891877e+00 5.19992888e-01 4.01035845e-01 1.88185163e-02 -6.57503247e-01 1.05511224e+00 3.38336319e-01 4.85131025e-01 -2.93058623e-02 7.55235493e-01 1.47927314e-01 9.84857500e-01 3.15441824e-02 -7.45491564e-01 1.08905447e+00 -9.29485679e-01 -7.35880613e-01 8.38532671e-02 7.30538368e-01 -1.19037986e+00 -7.34504536e-02 5.75383127e-01 -1.46737063e+00 -7.61339486e-01 -1.15692008e+00 6.84004873e-02 3.22981775e-01 1.29511997e-01 -1.08096212e-01 4.20656413e-01 -1.40061164e+00 7.25485027e-01 -9.40120935e-01 -1.20337076e-01 8.90444666e-02 5.83064198e-01 -2.83161134e-01 7.16267303e-02 -8.01901937e-01 6.77834928e-01 9.08616245e-01 1.16080038e-01 -5.03239512e-01 -5.84755361e-01 -7.63535500e-01 2.56298870e-01 1.82351351e-01 -4.64570761e-01 9.56140876e-01 -1.27299011e+00 -1.48910534e+00 5.55817246e-01 -8.58265877e-01 -7.97723889e-01 6.23442471e-01 -2.38683280e-02 -2.13984832e-01 5.23326755e-01 -2.66134501e-01 6.95673287e-01 1.06801689e+00 -1.08716846e+00 -1.05031502e+00 1.00135960e-01 -2.53700405e-01 9.57614630e-02 3.94409709e-02 1.38218477e-01 -7.35666513e-01 -8.98122430e-01 1.98743001e-01 -7.42444515e-01 -4.78832096e-01 -2.14665607e-02 2.61832952e-01 3.11145127e-01 8.12687755e-01 -8.87890697e-01 1.49294960e+00 -2.23990560e+00 5.25484741e-01 3.22427303e-01 2.36856252e-01 4.26611215e-01 2.02979729e-01 3.04050297e-01 -1.52005076e-01 -3.98310632e-01 -3.99835080e-01 -3.03959817e-01 -7.13359773e-01 1.07019857e-01 -9.57540274e-02 5.12249887e-01 -4.33058999e-02 3.51181537e-01 -8.23114634e-01 -7.01347530e-01 7.06859767e-01 6.93427742e-01 -6.87178791e-01 6.31952137e-02 3.20124701e-02 6.50193512e-01 -3.03045541e-01 1.46487862e-01 9.25534666e-01 -5.71743101e-02 3.57526243e-01 -1.05408452e-01 -7.08819270e-01 8.05020407e-02 -1.27747989e+00 1.46001804e+00 -4.58982319e-01 7.87244081e-01 3.49907987e-02 -9.33303237e-01 9.37300086e-01 7.26797819e-01 9.30808842e-01 -4.58900571e-01 8.84508807e-03 4.37073320e-01 -1.18080556e-01 -7.87458420e-02 7.88264811e-01 1.08385850e-02 6.02866411e-01 4.69892956e-02 -2.04804465e-01 1.95973054e-01 4.58060563e-01 -1.39773712e-01 1.09720528e+00 1.85948342e-01 9.26028371e-01 -3.19116801e-01 1.29030836e+00 -7.43171126e-02 4.69805151e-01 1.38726011e-01 -1.30882561e-01 5.00251293e-01 3.14272046e-01 -6.32884145e-01 -1.42842031e+00 -6.45570993e-01 -3.26743647e-02 4.13473547e-01 1.22211151e-01 -5.33263505e-01 -6.87447846e-01 -3.39967608e-01 -5.19925296e-01 4.59351689e-01 -4.92387980e-01 3.17624003e-01 -1.01924658e+00 -3.56019020e-01 -4.73464429e-02 4.31507885e-01 7.29033113e-01 -9.70206916e-01 -1.06990921e+00 6.17909133e-01 -4.73055318e-02 -9.72004116e-01 -2.38295615e-01 8.67957473e-02 -1.42508829e+00 -8.23718190e-01 -1.16631329e+00 -8.38189781e-01 4.07779336e-01 3.02881151e-01 9.72410560e-01 3.82092595e-01 6.63578287e-02 -1.48117021e-01 -5.74746370e-01 5.13240218e-01 -7.99642980e-01 -2.35566780e-01 -5.65228641e-01 1.48345783e-01 8.75831097e-02 -5.17681479e-01 -9.23117220e-01 4.46659267e-01 -9.27367210e-01 3.25730592e-01 6.15552545e-01 6.99495912e-01 6.05242193e-01 2.57931560e-01 -2.08921254e-01 -5.35490036e-01 -2.65871763e-01 -3.68771851e-01 -9.72977042e-01 -4.12832759e-02 -3.30813140e-01 1.02334134e-02 6.66760147e-01 -5.13294451e-02 -1.30003440e+00 6.32346332e-01 -5.53626716e-01 -2.52614439e-01 -1.63554311e-01 3.49801004e-01 1.55582815e-01 -2.54772812e-01 2.58960485e-01 2.49530599e-01 -3.28049719e-01 -4.96296555e-01 5.52501306e-02 2.87523448e-01 4.79278296e-01 1.63407907e-01 6.65328443e-01 4.07753378e-01 4.41923499e-01 -1.16670239e+00 4.82429303e-02 -8.60174239e-01 -1.09103251e+00 -4.21500683e-01 1.08808565e+00 -8.18818569e-01 -3.55273098e-01 3.23226213e-01 -1.54126716e+00 -4.16406333e-01 -2.86633909e-01 8.09361577e-01 -7.93180227e-01 7.51987696e-01 -6.31349862e-01 -7.31682897e-01 7.65107796e-02 -1.22215748e+00 7.35045135e-01 -2.12957069e-01 -3.77062440e-01 -1.12427998e+00 1.21620663e-01 1.45954937e-01 1.78988874e-01 -1.08893350e-01 6.14994943e-01 -3.09105520e-03 -9.39035118e-01 5.87208197e-02 -3.63896489e-01 2.48718813e-01 8.46872255e-02 -8.12550634e-03 -5.56192875e-01 -2.36313954e-01 2.63947874e-01 3.74023557e-01 9.73224282e-01 5.05269527e-01 7.25771189e-01 1.05520040e-02 -3.50264281e-01 5.39265215e-01 1.77205944e+00 6.15503371e-01 1.07242990e+00 2.87293613e-01 4.69338626e-01 4.34685290e-01 1.03188682e+00 7.89434433e-01 -1.92130193e-01 1.07241476e+00 4.50265914e-01 -1.77663087e-03 -3.53921413e-01 2.68873781e-01 1.70399293e-01 8.81643593e-01 -6.77026570e-01 -2.55670577e-01 -8.73729169e-01 7.01301634e-01 -1.89820647e+00 -1.13010335e+00 -6.00116491e-01 2.31440639e+00 3.77966076e-01 5.18209636e-02 -3.84581573e-02 6.94196463e-01 6.15245759e-01 5.03573775e-01 2.49927267e-01 -6.39526248e-01 1.98582605e-01 3.48177016e-01 8.14415395e-01 1.07151854e+00 -9.86249924e-01 6.54204786e-01 6.18775558e+00 9.49689984e-01 -1.00391126e+00 2.21324041e-02 4.59989041e-01 1.72744438e-01 6.67162463e-02 3.37762713e-01 -6.71071172e-01 3.90553862e-01 1.05935419e+00 4.01035026e-02 2.00635895e-01 5.95885873e-01 6.51222587e-01 -7.81371951e-01 -8.24561954e-01 9.42457438e-01 -3.76095027e-02 -1.26650131e+00 -5.37747443e-02 2.48367667e-01 8.15712154e-01 -3.69905561e-01 -2.67292261e-01 -5.23154020e-01 -3.07313472e-01 -5.74445426e-01 7.75903523e-01 4.10438657e-01 4.21334982e-01 -9.51671302e-01 7.99479127e-01 2.88943410e-01 -1.70920098e+00 2.93569285e-02 -3.10018539e-01 -1.57996655e-01 6.48938239e-01 6.57765865e-01 -6.22127235e-01 5.67652643e-01 5.41769564e-01 8.90322804e-01 -1.58470958e-01 1.34676528e+00 -4.15174253e-02 5.58922887e-01 -9.48290452e-02 3.62368464e-01 4.08724725e-01 -1.20569274e-01 7.10703731e-01 1.43153071e+00 6.95528030e-01 6.70033768e-02 -3.17243546e-01 2.66315818e-01 3.70513707e-01 4.06021059e-01 -2.34137371e-01 6.37636602e-01 -1.71405777e-01 8.13943744e-01 -9.61969972e-01 -6.65798783e-01 -5.08753300e-01 1.47016144e+00 7.26813003e-02 2.67593175e-01 -7.21258283e-01 8.50216392e-03 2.11884230e-01 3.81620735e-01 9.25389647e-01 -5.36966383e-01 -1.96066365e-01 -8.58780980e-01 -8.43121707e-02 -4.89194870e-01 5.10322526e-02 -4.06958371e-01 -1.99089035e-01 6.30176723e-01 5.54690510e-02 -1.51905501e+00 -7.41402149e-01 -3.41334015e-01 -2.20618039e-01 1.19921362e+00 -1.46768761e+00 -5.62765956e-01 -9.63907763e-02 5.10197937e-01 8.95868599e-01 -1.36662228e-02 6.00444317e-01 4.24607337e-01 -1.88779749e-03 -8.43721926e-02 3.62612486e-01 -2.40785077e-01 3.58284563e-01 -9.28910077e-01 2.32982695e-01 1.20655835e+00 -6.36110082e-02 1.04025461e-01 1.00514722e+00 -6.40579164e-01 -7.62649655e-01 -8.51364911e-01 1.55048919e+00 4.81705934e-01 5.99867105e-01 1.91336021e-01 -1.12112248e+00 3.93151432e-01 4.60091412e-01 4.42306995e-02 3.16998661e-01 -6.07551575e-01 2.01584160e-01 -1.15297176e-01 -8.85582507e-01 4.53339100e-01 6.35317564e-01 -2.46600211e-01 -3.14064294e-01 -7.54050389e-02 2.07236648e-01 -3.79292279e-01 -1.11040461e+00 3.50975871e-01 4.98387963e-01 -1.51230001e+00 9.36564982e-01 4.18471843e-01 5.54990530e-01 -5.72547317e-01 -5.13484478e-02 -8.80540073e-01 -6.36665523e-01 -7.59447813e-01 -2.19812542e-01 8.80252600e-01 2.17778906e-01 -2.21738353e-01 9.14513409e-01 2.58978426e-01 -1.56518426e-02 -5.09005010e-01 -1.17059577e+00 -4.27668154e-01 -4.53058660e-01 -5.46730518e-01 1.15632340e-01 4.46736842e-01 -1.40522137e-01 -1.05584621e-01 -6.77761793e-01 1.17514551e-01 5.62590659e-01 -3.93665619e-02 5.03782749e-01 -7.95116186e-01 -6.00280344e-01 -4.01721865e-01 -8.37828457e-01 -1.59513152e+00 -2.42167652e-01 -4.02440757e-01 7.14858398e-02 -1.25280881e+00 4.95061018e-02 -2.90403783e-01 -2.96925962e-01 -1.45074502e-01 7.15387836e-02 6.36227489e-01 2.99578100e-01 2.05986306e-01 -3.36642057e-01 2.04044968e-01 1.01777637e+00 8.49244297e-02 -2.78869838e-01 2.56867945e-01 3.12707275e-01 8.31479788e-01 6.40216529e-01 -3.07954311e-01 -3.19137186e-01 -2.54850119e-01 3.08060609e-02 6.18072391e-01 3.61888021e-01 -1.28050804e+00 2.07041547e-01 1.43906489e-01 5.40466964e-01 -8.43379498e-01 5.14501810e-01 -1.23839068e+00 4.93389666e-01 8.40184927e-01 -1.05500564e-01 1.01904767e-02 1.98818251e-01 4.93961483e-01 -6.48545086e-01 -7.06057012e-01 1.10873461e+00 7.16153979e-02 -1.11749661e+00 -8.29478875e-02 -1.04623497e+00 -6.82337880e-01 1.27415073e+00 -6.13694906e-01 6.62640631e-01 -3.38832915e-01 -1.02119851e+00 -3.30616087e-01 6.55191183e-01 -2.26617128e-01 6.74902260e-01 -1.08112907e+00 -6.14191353e-01 8.88437219e-03 -3.39260817e-01 -3.06167781e-01 6.39924824e-01 1.12248266e+00 -1.19909441e+00 7.25059927e-01 -2.87477762e-01 -8.03149700e-01 -1.92960644e+00 9.74123240e-01 2.56185587e-02 -5.55737257e-01 -7.56393015e-01 3.49298030e-01 1.33001611e-01 9.10254002e-01 -1.05835520e-01 -3.46153647e-01 -5.54198861e-01 -9.50810164e-02 6.62639618e-01 7.05358267e-01 -2.07521785e-02 -1.27046955e+00 -2.62856841e-01 8.85294795e-01 2.20194340e-01 -3.90871048e-01 1.12505388e+00 -5.30361831e-01 -1.86045662e-01 2.47486204e-01 1.24622214e+00 2.96195000e-01 -1.47232676e+00 6.43854141e-02 2.50435889e-01 -8.10490489e-01 1.98793605e-01 -9.86216366e-02 -1.07332301e+00 7.91595697e-01 5.46756148e-01 1.48646161e-01 1.64312637e+00 -3.47463906e-01 6.24685943e-01 -2.09204569e-01 4.00044680e-01 -7.42764413e-01 -3.02961200e-01 5.07593751e-01 7.27724016e-01 -8.84269536e-01 2.96072572e-01 -8.83950114e-01 -2.57182062e-01 1.56127703e+00 -9.63755101e-02 -1.43590406e-01 7.64273345e-01 3.03144068e-01 -3.57604355e-01 3.35729420e-01 -7.58639932e-01 -2.56890178e-01 4.30780500e-01 3.32002193e-01 8.54578495e-01 -3.50247741e-01 -8.86271477e-01 -6.84355438e-01 3.03128988e-01 3.16391051e-01 3.10255110e-01 8.30696940e-01 -6.09241545e-01 -1.40268719e+00 -5.85914671e-01 -4.95807268e-02 -5.19228041e-01 -1.61622182e-01 1.85161918e-01 7.32645988e-01 2.25388020e-01 1.05931330e+00 2.13566780e-01 -1.28346562e-01 1.71592366e-02 -1.86533943e-01 4.93704677e-01 -3.28916162e-01 -5.11723638e-01 6.98706329e-01 5.54989800e-02 -8.77270937e-01 -9.44985151e-01 -8.40827227e-01 -1.26612353e+00 -1.71135843e-01 4.18834686e-02 1.44052550e-01 4.67573285e-01 8.45375776e-01 -1.42061651e-01 5.22960484e-01 6.77229583e-01 -1.24425054e+00 3.88441116e-01 -6.26787961e-01 -5.32307327e-01 8.44175890e-02 3.90576601e-01 -3.72510910e-01 -1.19717397e-01 4.29512441e-01]
[11.324528694152832, -2.0726277828216553]
9f657809-b346-4186-90b8-bc566f7e2947
a-unifying-partially-interpretable-framework
2208.07581
null
https://arxiv.org/abs/2208.07581v3
https://arxiv.org/pdf/2208.07581v3.pdf
Regression modelling of spatiotemporal extreme U.S. wildfires via partially-interpretable neural networks
Risk management in many environmental settings requires an understanding of the mechanisms that drive extreme events. Useful metrics for quantifying such risk are extreme quantiles of response variables conditioned on predictor variables that describe, e.g., climate, biosphere and environmental states. Typically these quantiles lie outside the range of observable data and so, for estimation, require specification of parametric extreme value models within a regression framework. Classical approaches in this context utilise linear or additive relationships between predictor and response variables and suffer in either their predictive capabilities or computational efficiency; moreover, their simplicity is unlikely to capture the truly complex structures that lead to the creation of extreme wildfires. In this paper, we propose a new methodological framework for performing extreme quantile regression using artificial neutral networks, which are able to capture complex non-linear relationships and scale well to high-dimensional data. The ``black box" nature of neural networks means that they lack the desirable trait of interpretability often favoured by practitioners; thus, we unify linear, and additive, regression methodology with deep learning to create partially-interpretable neural networks that can be used for statistical inference but retain high prediction accuracy. To complement this methodology, we further propose a novel point process model for extreme values which overcomes the finite lower-endpoint problem associated with the generalised extreme value class of distributions. Efficacy of our unified framework is illustrated on U.S. wildfire data with a high-dimensional predictor set and we illustrate vast improvements in predictive performance over linear and spline-based regression techniques.
['Raphaël Huser', 'Jordan Richards']
2022-08-16
null
null
null
null
['additive-models']
['methodology']
[ 2.87563503e-01 -1.29256487e-01 -1.70064869e-03 -5.10800481e-01 -3.69974554e-01 -5.88311136e-01 6.92089140e-01 4.10630405e-01 -4.46547508e-01 1.01310849e+00 3.01233590e-01 -6.61658645e-01 -6.47319853e-01 -1.09345424e+00 -6.87748194e-01 -7.71352232e-01 -4.65065569e-01 3.35428387e-01 -2.91119456e-01 -3.93100560e-01 8.74620155e-02 7.02547252e-01 -1.62656116e+00 2.66422462e-02 9.51823056e-01 9.18639541e-01 -1.82878643e-01 4.95114893e-01 -2.54363030e-01 2.55441487e-01 -4.26192790e-01 -2.51822352e-01 2.21940160e-01 -1.00781702e-01 2.34004203e-02 -7.21679032e-01 2.30589524e-01 -2.19373181e-01 2.18561083e-01 8.09027672e-01 5.52695096e-01 5.40122613e-02 1.28809309e+00 -1.08841670e+00 -7.81152487e-01 5.44276357e-01 -4.45571154e-01 -8.39390606e-02 -2.33579408e-02 2.80004650e-01 1.02774870e+00 -6.28874481e-01 1.33525580e-01 1.24605477e+00 1.22102380e+00 1.32055447e-01 -1.83255351e+00 -4.56161737e-01 1.01354644e-02 -2.44960055e-01 -1.25195515e+00 -1.31940007e-01 5.37288010e-01 -8.08481693e-01 7.48692930e-01 3.60996097e-01 6.67061985e-01 1.20170844e+00 6.17169559e-01 1.84569672e-01 1.12078023e+00 -4.21406850e-02 5.76364994e-01 -2.51447618e-01 -9.28395689e-02 -4.27987091e-02 3.92395496e-01 7.60295808e-01 -1.86385050e-01 -3.60670060e-01 6.15648031e-01 4.45818782e-01 -1.70837030e-01 -1.13843940e-01 -7.74082959e-01 9.74663258e-01 5.91668546e-01 -5.77814318e-02 -6.41681314e-01 4.22103226e-01 4.40341502e-01 2.82198429e-01 7.93942750e-01 4.38209355e-01 -5.65099776e-01 8.75517726e-02 -1.09931636e+00 3.61319751e-01 8.42153668e-01 3.80898952e-01 5.97305536e-01 1.80208936e-01 -2.27649420e-01 7.41606653e-01 3.80431920e-01 6.43542945e-01 5.19265346e-02 -6.97713673e-01 5.98420426e-02 5.07459283e-01 3.42192709e-01 -9.25455987e-01 -5.99821568e-01 -5.93363404e-01 -1.26624858e+00 6.90133333e-01 6.66860580e-01 -2.91188985e-01 -9.56982136e-01 1.89000905e+00 2.34196514e-01 -6.25424311e-02 9.43076834e-02 7.31454551e-01 2.78845906e-01 6.80921555e-01 6.62279665e-01 -1.84232697e-01 1.16602671e+00 1.69330984e-01 -4.83520508e-01 -2.58547097e-01 1.43156901e-01 -1.36803225e-01 1.26550078e+00 2.45181844e-01 -7.42760837e-01 -2.86863923e-01 -8.03968072e-01 2.98927218e-01 -7.20179677e-01 -3.88616800e-01 7.42225647e-01 6.23569310e-01 -7.67079651e-01 9.41618860e-01 -7.32328475e-01 3.49963866e-02 5.17525136e-01 3.17478657e-01 -6.58076257e-02 4.33715165e-01 -1.43026483e+00 1.02356708e+00 3.97135049e-01 5.44435918e-01 -5.56331038e-01 -9.22371209e-01 -5.92098475e-01 5.17683208e-01 4.08789590e-02 -6.58811867e-01 7.60992050e-01 -9.72299159e-01 -1.14342988e+00 3.84580642e-01 2.91826963e-01 -4.46642250e-01 6.91887796e-01 -4.01116312e-01 -1.03410654e-01 -4.25059378e-01 -1.05863802e-01 3.91900212e-01 6.25007451e-01 -1.10125518e+00 -2.34844491e-01 -4.61936474e-01 -2.59832650e-01 -2.58336127e-01 -1.79055586e-01 -1.43920034e-01 7.17348456e-01 -5.28715849e-01 1.81545876e-02 -6.44189119e-01 -6.71655834e-01 2.88470149e-01 -2.00922757e-01 3.49498466e-02 2.93985933e-01 -5.51652074e-01 1.15883589e+00 -2.06906867e+00 8.87335762e-02 2.76271909e-01 1.50050312e-01 1.43616885e-01 4.01214957e-02 6.22898281e-01 -1.80774599e-01 3.50237578e-01 -8.10460389e-01 -1.12402588e-02 3.55557382e-01 3.49700212e-01 -6.55317962e-01 6.00617826e-01 4.56230879e-01 8.62078428e-01 -6.88847125e-01 -7.42206126e-02 3.71223271e-01 8.34532201e-01 -3.90866965e-01 1.45513281e-01 -5.03194332e-01 4.82906789e-01 -2.71190971e-01 4.43087161e-01 7.10354984e-01 3.91084701e-01 -1.08952664e-01 2.53969818e-01 -4.58281368e-01 -1.72369719e-01 -1.02355564e+00 9.62830961e-01 -5.46749651e-01 3.99874151e-01 -7.73954717e-03 -9.29864049e-01 1.33992445e+00 1.08554259e-01 2.80864090e-01 -3.48612547e-01 -5.56922890e-03 4.05329049e-01 -1.88283771e-02 -3.38551283e-01 2.85896540e-01 -9.67663765e-01 -1.99019805e-01 2.07701370e-01 -3.74524295e-01 -1.02017380e-01 -1.96423158e-01 -5.82815886e-01 6.52218878e-01 4.95605081e-01 4.66796339e-01 -4.25719917e-01 2.12423220e-01 -1.75333753e-01 6.44471109e-01 8.49558353e-01 -2.34656289e-01 5.52018106e-01 1.06867647e+00 -7.42579460e-01 -1.20596409e+00 -1.38411784e+00 -7.03687727e-01 1.03182745e+00 -3.83657634e-01 2.07052901e-01 -2.95859307e-01 -1.20410006e-02 4.65408921e-01 1.08394575e+00 -7.72007704e-01 -3.95286947e-01 -2.72404552e-01 -1.33138573e+00 8.26073527e-01 5.96745491e-01 -1.79503765e-02 -9.95888174e-01 -8.55171919e-01 3.19082737e-01 2.69148678e-01 -3.61871690e-01 3.57193828e-01 7.94273794e-01 -9.30279374e-01 -6.18925333e-01 -5.90032279e-01 1.60301477e-02 1.23525038e-01 -5.29422522e-01 1.30806518e+00 -3.18864316e-01 -1.53862715e-01 -1.52984068e-01 -7.02460632e-02 -5.86831689e-01 -4.67945844e-01 -1.36860028e-01 1.18520953e-01 -9.16192234e-02 4.57804024e-01 -1.17948115e+00 -8.31988633e-01 -2.49151681e-02 -1.16324437e+00 -3.22560132e-01 7.15152502e-01 1.01541984e+00 5.30468941e-01 -3.38395447e-01 1.17633533e+00 -5.63213170e-01 6.72791719e-01 -9.96436119e-01 -6.08346164e-01 1.48523271e-01 -6.27983987e-01 2.24775881e-01 9.86414015e-01 -4.36069012e-01 -1.00458837e+00 1.46598056e-01 -3.00276130e-01 -8.35477039e-02 -5.42023182e-01 9.56286192e-01 2.82422323e-02 4.34827477e-01 9.63798106e-01 1.06432410e-02 2.07509249e-02 -4.08708900e-01 4.78879958e-01 7.45747805e-01 5.79558134e-01 -6.78606451e-01 6.58757091e-01 4.26734835e-01 5.45536578e-01 -8.38611722e-01 -4.83330041e-01 -1.59957372e-02 -7.60780573e-01 -1.97961003e-01 6.77284062e-01 -7.00048029e-01 -7.28870034e-01 3.87172699e-01 -1.00715041e+00 -3.52631748e-01 -6.67406917e-01 4.74310219e-01 -8.30884993e-01 4.63831238e-02 -1.14088796e-01 -1.35103738e+00 -3.51319939e-01 -7.40715027e-01 1.03806365e+00 1.58900857e-01 -2.42431685e-01 -1.31592810e+00 4.00388718e-01 -5.23351908e-01 7.72739947e-01 1.09091437e+00 1.24038005e+00 -6.14373147e-01 -1.37741730e-01 -3.26150209e-01 -4.58148904e-02 2.94720173e-01 -1.45090921e-02 3.54178667e-01 -1.02423394e+00 1.00771785e-02 -7.70915747e-02 -6.80958256e-02 9.84499753e-01 8.50075603e-01 9.25119400e-01 -1.78281412e-01 -1.21158130e-01 9.18230891e-01 1.54621351e+00 -8.59224796e-02 6.67326212e-01 3.92115682e-01 2.19887450e-01 9.59286988e-01 1.71357870e-01 6.50006413e-01 2.13898029e-02 3.56172651e-01 8.88184071e-01 -2.71861136e-01 5.02522647e-01 -1.12825014e-01 3.94102663e-01 -1.44595116e-01 -1.35190606e-01 -1.43548653e-01 -1.17949092e+00 5.07738352e-01 -1.76739299e+00 -1.27490044e+00 -5.71330905e-01 2.52090931e+00 7.49207616e-01 3.30835208e-02 1.81650713e-01 -2.30432972e-02 7.61535406e-01 1.62382796e-01 -6.88717365e-01 -8.60736787e-01 -2.85698712e-01 2.44298428e-02 6.23605430e-01 2.51321912e-01 -1.04652226e+00 5.28450906e-01 6.55221796e+00 3.78995299e-01 -1.09136450e+00 -3.37896585e-01 7.15101123e-01 -2.35941589e-01 -5.00245392e-01 -3.04050688e-02 -4.55274791e-01 3.92602086e-01 1.50339139e+00 -1.74071148e-01 2.93917418e-01 7.22994208e-01 6.53860509e-01 3.63384597e-02 -1.11269259e+00 3.05016220e-01 -6.39172554e-01 -9.01631236e-01 -1.98033988e-01 2.53616601e-01 4.74870771e-01 1.64284542e-01 4.19069707e-01 2.98792928e-01 5.79853773e-01 -1.47111273e+00 5.87273777e-01 1.02090526e+00 7.81764090e-01 -9.19954062e-01 6.66497231e-01 4.87145543e-01 -7.47362077e-01 -3.93877059e-01 -6.99235559e-01 -4.21780258e-01 2.88838178e-01 9.49983060e-01 -4.74903315e-01 2.32932717e-01 5.83605826e-01 4.88707393e-01 -3.16852361e-01 1.05864668e+00 -1.08954713e-01 6.65782928e-01 -9.00239706e-01 2.31851880e-02 2.97801256e-01 -5.07426083e-01 5.54288566e-01 1.20677495e+00 5.83937824e-01 -2.62255281e-01 -2.14290857e-01 1.36218774e+00 4.93962139e-01 1.09271720e-01 -1.00847161e+00 1.23775072e-01 3.05220872e-01 9.82310951e-01 -5.39130270e-01 5.50836623e-02 -3.09256822e-01 3.04661721e-01 1.78967163e-01 4.40417707e-01 -7.24981129e-01 -2.85486639e-01 9.32832181e-01 6.58593550e-02 1.56741932e-01 -2.62446016e-01 -8.10424507e-01 -8.48404288e-01 -5.42275421e-02 -5.24629653e-01 3.42779249e-01 -5.17209411e-01 -1.59148014e+00 4.60106671e-01 1.94320306e-01 -1.01887167e+00 -5.46680272e-01 -7.80470192e-01 -8.59545112e-01 1.31881940e+00 -1.58966875e+00 -1.18648899e+00 -1.96902335e-01 1.61703825e-01 8.59608874e-04 2.41132528e-01 8.85574400e-01 -1.43752426e-01 -2.80819982e-01 1.60627291e-01 1.00559163e+00 -3.57578874e-01 4.51810747e-01 -1.54963827e+00 5.22887707e-01 4.88638818e-01 -3.08783412e-01 6.31964803e-01 9.44858909e-01 -8.55237424e-01 -1.07889724e+00 -1.14194512e+00 6.42285645e-01 -3.72452170e-01 7.60673821e-01 -2.48554245e-01 -1.12633181e+00 5.18418431e-01 -4.06084239e-01 -4.19346429e-02 9.95564103e-01 4.17360246e-01 -3.20979536e-01 -1.68297604e-01 -1.19938564e+00 6.90406024e-01 5.99300146e-01 -3.37725788e-01 -6.61298692e-01 6.50227070e-02 2.77794242e-01 2.97236592e-01 -1.06947362e+00 3.87532473e-01 9.74509180e-01 -1.11709917e+00 1.08932638e+00 -9.34854507e-01 7.42249250e-01 -1.90927655e-01 -3.10618162e-01 -1.46520495e+00 -1.98156774e-01 -2.32309401e-01 1.22243628e-01 1.16948366e+00 2.98253596e-01 -8.75106454e-01 3.42282027e-01 8.47435355e-01 -2.84216776e-02 -7.24472880e-01 -1.35981083e+00 -8.01100910e-01 9.81887460e-01 -5.71301103e-01 8.64733398e-01 5.87740123e-01 -2.51808226e-01 -9.72522497e-02 -4.03143167e-01 2.39939630e-01 7.78247952e-01 5.56296185e-02 4.88198489e-01 -1.74750769e+00 -2.19532430e-01 -6.37645483e-01 -4.37171549e-01 -2.79442489e-01 2.39703193e-01 -8.39960277e-01 1.05023339e-01 -1.21568727e+00 -1.87455922e-01 -5.08797884e-01 -4.07927871e-01 4.07963514e-01 -2.32696384e-01 1.12232789e-01 -1.18384868e-01 1.58496052e-01 2.83570349e-01 9.13507998e-01 5.91117561e-01 1.10531583e-01 -4.16823655e-01 8.52560401e-02 -6.35815978e-01 7.56579280e-01 9.03949142e-01 -5.95615268e-01 -1.41433522e-01 -1.30433097e-01 5.25919437e-01 1.93797946e-01 7.64959514e-01 -8.87819886e-01 -2.11061031e-01 -7.01534629e-01 7.30641603e-01 -2.65444309e-01 9.62116048e-02 -9.72785294e-01 5.42575538e-01 5.13347626e-01 -4.22698587e-01 -1.06871054e-01 1.59859717e-01 6.83434784e-01 -1.49350185e-02 -1.43690526e-01 7.26784825e-01 -1.10902702e-02 -1.39888257e-01 1.91440597e-01 -5.03550828e-01 -2.96224326e-01 9.19237494e-01 -3.26258510e-01 -3.50582689e-01 -4.72782880e-01 -7.93260694e-01 2.10746989e-01 5.02050757e-01 8.86695683e-02 3.97756815e-01 -9.96423185e-01 -1.10143530e+00 1.36226356e-01 -1.51804328e-01 -2.59502698e-02 8.88699889e-02 6.95055366e-01 -4.68849301e-01 -1.24648899e-01 -4.54978526e-01 -6.52916431e-01 -5.37005007e-01 5.11905134e-01 6.84686303e-01 -3.40287499e-02 -6.44322217e-01 2.87611276e-01 3.11907381e-01 -5.61350942e-01 -2.75818825e-01 -2.64619946e-01 -2.86409020e-01 3.73346210e-01 3.51754189e-01 3.73091906e-01 -3.30280155e-01 -5.72812974e-01 -2.38310456e-01 3.18525642e-01 5.18660843e-01 -7.08440319e-02 1.96216214e+00 2.23895852e-02 -1.84008017e-01 8.85107577e-01 8.63285720e-01 -3.73164922e-01 -1.67401958e+00 3.23758841e-01 2.47289822e-01 -2.70436317e-01 1.01685174e-01 -9.44209099e-01 -7.25172460e-01 1.13585329e+00 6.93160534e-01 4.28003132e-01 1.09450388e+00 -3.37652087e-01 1.77176699e-01 4.51930135e-01 -1.62344798e-01 -8.50717545e-01 -6.87715411e-01 3.94358397e-01 1.17638183e+00 -1.00192368e+00 -4.80212308e-02 1.87629029e-01 -1.39819697e-01 1.29772663e+00 1.84945837e-01 -3.16448241e-01 5.82606196e-01 4.63920265e-01 4.38059419e-02 -6.06794655e-02 -7.21978903e-01 -2.47667149e-01 -3.07066049e-02 8.23477209e-01 4.84004170e-01 3.21650863e-01 -4.07658994e-01 7.51651824e-01 -2.19509140e-01 -6.70123473e-02 5.25967598e-01 5.04690528e-01 -6.19873822e-01 -7.34526157e-01 -6.30715072e-01 6.57121599e-01 -5.89416802e-01 -3.78655374e-01 -2.63280123e-01 5.95903456e-01 -7.98110440e-02 6.04956746e-01 2.05951765e-01 -6.68527558e-02 3.05425525e-01 4.33411539e-01 -2.29523644e-01 -1.83224306e-01 -7.15595543e-01 -8.00781399e-02 -1.50386959e-01 -1.44698009e-01 2.11310349e-02 -8.03711116e-01 -8.17613661e-01 -4.78024155e-01 -2.28524968e-01 4.83774394e-02 8.82781267e-01 7.50408649e-01 2.44874582e-01 3.64555448e-01 5.19169211e-01 -9.82431293e-01 -1.14179754e+00 -9.50104237e-01 -7.11640954e-01 2.45590016e-01 5.93003690e-01 -6.67923033e-01 -7.01816320e-01 -2.34026730e-01]
[7.034133434295654, 3.6631884574890137]
0912b7d7-412f-4e6e-b8ff-82fee4a0fe5d
improving-generalizability-of-graph-anomaly
2209.10168
null
https://arxiv.org/abs/2209.10168v1
https://arxiv.org/pdf/2209.10168v1.pdf
Improving Generalizability of Graph Anomaly Detection Models via Data Augmentation
Graph anomaly detection (GAD) is a vital task since even a few anomalies can pose huge threats to benign users. Recent semi-supervised GAD methods, which can effectively leverage the available labels as prior knowledge, have achieved superior performances than unsupervised methods. In practice, people usually need to identify anomalies on new (sub)graphs to secure their business, but they may lack labels to train an effective detection model. One natural idea is to directly adopt a trained GAD model to the new (sub)graph for testing. However, we find that existing semi-supervised GAD methods suffer from poor generalization issue, i.e., well-trained models could not perform well on an unseen area (i.e., not accessible in training) of the same graph. It may cause great troubles. In this paper, we base on the phenomenon and propose a general and novel research problem of generalized graph anomaly detection that aims to effectively identify anomalies on both the training-domain graph and unseen testing graph to eliminate potential dangers. Nevertheless, it is a challenging task since only limited labels are available, and the normal background may differ between training and testing data. Accordingly, we propose a data augmentation method named \textit{AugAN} (\uline{Aug}mentation for \uline{A}nomaly and \uline{N}ormal distributions) to enrich training data and boost the generalizability of GAD models. Experiments verify the effectiveness of our method in improving model generalizability.
['Long-Kai Huang', 'Fu-Lai Chung', 'Ninghao Liu', 'Xiao Huang', 'Shuang Zhou']
2022-09-21
null
null
null
null
['graph-anomaly-detection']
['graphs']
[ 2.01437756e-01 1.12497933e-01 -1.94702744e-01 -2.00957075e-01 -2.56201237e-01 -6.07878506e-01 4.51969951e-01 4.60455149e-01 1.33862086e-02 5.58336973e-01 -4.02757883e-01 -5.70949733e-01 -1.47159351e-02 -1.07256877e+00 -6.39440656e-01 -6.37265384e-01 -2.03841925e-03 2.66016096e-01 4.68311101e-01 -2.68530399e-01 -2.05646157e-02 4.75274026e-01 -1.12517929e+00 -1.40272379e-01 1.35388052e+00 9.58749175e-01 -3.35658878e-01 2.72886515e-01 -2.66258400e-02 5.36544204e-01 -6.99263453e-01 -5.23756862e-01 3.50799501e-01 -5.87007642e-01 -6.11392498e-01 2.88211286e-01 3.70144010e-01 -2.76565790e-01 -6.29880726e-01 1.58351386e+00 2.48455703e-01 7.37409815e-02 6.51992381e-01 -1.57854235e+00 -7.17532516e-01 2.78851628e-01 -7.91874647e-01 5.59636295e-01 1.66705251e-01 1.03540323e-03 9.53269482e-01 -5.87852299e-01 1.54541060e-01 8.60175788e-01 5.40114343e-01 6.50865436e-01 -9.53450024e-01 -8.24616194e-01 5.68225980e-01 5.98025583e-02 -1.43694806e+00 -1.71564132e-01 1.12856996e+00 -1.78600535e-01 1.72186002e-01 5.12796938e-01 3.88891459e-01 1.26852560e+00 -1.24674596e-01 8.18597317e-01 9.66536164e-01 -2.10468501e-01 1.16779953e-01 2.47383267e-01 3.83027464e-01 8.36468220e-01 7.16090739e-01 -6.19720668e-02 -2.21692532e-01 -3.33072871e-01 4.93689895e-01 2.66359776e-01 -2.18447566e-01 -2.03474820e-01 -6.40830159e-01 7.49761581e-01 5.74170470e-01 4.02081013e-01 -2.59025007e-01 -3.53210360e-01 3.34689200e-01 2.43244216e-01 5.32750309e-01 2.52477586e-01 -1.32300183e-01 1.64224893e-01 -4.73862857e-01 -2.60727018e-01 6.07371092e-01 9.02906299e-01 7.68056571e-01 2.93785006e-01 9.59863663e-02 6.89807773e-01 2.76355445e-01 3.49369794e-01 4.42984492e-01 5.96262775e-02 3.46013665e-01 1.00418615e+00 -3.56962204e-01 -1.55576885e+00 -3.01477879e-01 -7.90600657e-01 -1.07915306e+00 -1.84462503e-01 3.83837342e-01 -1.96433917e-01 -1.06605744e+00 1.49965012e+00 4.55374271e-01 4.85880613e-01 -2.07294270e-01 5.84931493e-01 7.81604469e-01 5.49354553e-01 1.30722225e-01 -2.75421232e-01 9.65186179e-01 -8.33825588e-01 -6.34451687e-01 -5.54415524e-01 9.39921856e-01 -3.04750085e-01 1.17067564e+00 5.85776448e-01 -4.35293257e-01 -4.01414484e-01 -1.13795710e+00 5.47018528e-01 -4.39680904e-01 -1.94071680e-01 6.86218262e-01 9.74650860e-01 -7.20157921e-01 3.47147435e-01 -7.17470467e-01 -3.25531751e-01 5.72412968e-01 3.13655615e-01 -3.33123565e-01 -4.27661389e-01 -1.23811579e+00 3.73437881e-01 5.80544889e-01 2.31208950e-01 -1.02119470e+00 5.51398669e-04 -1.03435993e+00 -2.66562670e-01 8.83973956e-01 1.19453482e-01 7.01463819e-01 -9.49742794e-01 -7.18300760e-01 7.07874596e-01 1.51351944e-01 -4.00376469e-01 1.96590140e-01 -5.94262592e-02 -1.09584987e+00 1.76407043e-02 1.83409467e-01 -1.68983694e-02 1.08858514e+00 -1.24865437e+00 -5.85928798e-01 -5.95786631e-01 5.42640053e-02 8.41455162e-02 -1.00236154e+00 -2.09005430e-01 -5.81777215e-01 -7.95546532e-01 3.91359627e-01 -7.90229678e-01 -1.85272515e-01 -5.24038196e-01 -7.99992740e-01 -2.57544935e-01 1.28056669e+00 -8.07161808e-01 1.72149050e+00 -2.32363534e+00 -3.68084401e-01 7.85466969e-01 5.66420972e-01 6.93399251e-01 3.41082662e-02 1.07608818e-01 -2.24699304e-01 3.24176908e-01 -3.66421372e-01 1.09403409e-01 -3.86224151e-01 4.91603047e-01 -2.92188436e-01 5.31909049e-01 2.14454293e-01 6.64195955e-01 -9.32953894e-01 -3.82547617e-01 -1.33408576e-01 -5.34063019e-02 -3.75047088e-01 2.06156209e-01 -1.28281847e-01 5.57008445e-01 -8.52815628e-01 1.00553989e+00 7.43619263e-01 -3.05206448e-01 3.31910588e-02 1.69918016e-01 4.22620326e-01 4.64928858e-02 -1.25654888e+00 1.01898038e+00 2.88514644e-02 1.51630700e-01 -1.81750044e-01 -1.35813582e+00 1.19060171e+00 3.34337689e-02 3.31931025e-01 -6.32272840e-01 3.31902807e-03 2.77376235e-01 1.92127019e-01 -3.33632201e-01 2.04593539e-02 1.52321413e-01 -2.38765441e-02 4.14417028e-01 -1.76756084e-01 4.11229700e-01 1.21339776e-01 3.75045985e-01 1.31684208e+00 -3.58495295e-01 2.87893474e-01 -4.88810763e-02 7.86974669e-01 -1.87234282e-01 7.41890073e-01 7.77020574e-01 -3.82453173e-01 3.61273557e-01 8.03804576e-01 -3.30261379e-01 -5.42031109e-01 -9.83771086e-01 -9.64173675e-02 1.05863726e+00 3.57619047e-01 -6.24638498e-01 -6.43179715e-01 -1.53652084e+00 -1.16443060e-01 6.26806855e-01 -5.41576087e-01 -7.68294752e-01 -3.83415312e-01 -1.03156006e+00 6.61748946e-01 4.85975355e-01 7.24002242e-01 -9.79484737e-01 2.19353929e-01 -4.73295674e-02 -6.18233420e-02 -1.08816624e+00 -4.02355105e-01 -1.36201203e-01 -7.19823539e-01 -1.23482990e+00 -2.02531278e-01 -7.30600178e-01 1.25590920e+00 3.80766422e-01 9.22674656e-01 7.01996744e-01 -5.78402989e-02 3.08852971e-01 -4.92964178e-01 -6.23737156e-01 -4.73537534e-01 8.17459673e-02 3.69452864e-01 5.19090950e-01 5.63276291e-01 -5.64111590e-01 -4.55308020e-01 5.95998645e-01 -1.10830975e+00 -5.35860181e-01 5.68773448e-01 6.47924721e-01 5.68792939e-01 6.98522747e-01 8.36364746e-01 -1.36369061e+00 7.14209259e-01 -6.27431035e-01 -5.01428902e-01 1.24384478e-01 -8.28060269e-01 -1.81339622e-01 7.61221230e-01 -4.46681589e-01 -6.86683416e-01 -1.63793728e-01 -2.00962707e-01 -4.42677110e-01 -2.56887496e-01 6.53752506e-01 -6.67650700e-01 -2.81575590e-01 7.79373348e-01 2.64986783e-01 -9.24594775e-02 -3.60482603e-01 -1.00217529e-01 5.31015098e-01 4.59340841e-01 -4.77448106e-01 1.34262240e+00 3.19938600e-01 7.94791505e-02 -8.39549839e-01 -9.36325788e-01 -5.82507312e-01 -4.62419063e-01 -1.57977909e-01 4.91585493e-01 -6.42497361e-01 -2.33704060e-01 6.38747036e-01 -5.62319994e-01 -3.74913216e-03 -7.08276778e-02 2.22019851e-01 2.69922972e-01 8.48136008e-01 -2.40419507e-01 -8.42066586e-01 -4.47229408e-02 -7.39481568e-01 6.91229641e-01 3.65688384e-01 1.10831961e-01 -1.16358161e+00 -1.69455498e-01 1.95284367e-01 1.03362858e-01 4.17693108e-01 8.38101029e-01 -1.39972579e+00 -3.50420684e-01 -7.42271364e-01 -2.64455140e-01 6.94900095e-01 4.73075479e-01 -1.50567159e-01 -9.38094676e-01 -5.34115911e-01 2.26875052e-01 -1.74898669e-01 6.63100421e-01 -1.94305092e-01 1.54542315e+00 -5.31222701e-01 -3.96855444e-01 2.97489196e-01 9.05250430e-01 2.15169653e-01 6.16690695e-01 1.48254171e-01 1.14837372e+00 3.05063188e-01 6.98262334e-01 2.35945672e-01 9.96598601e-02 3.05281729e-01 6.56377017e-01 -1.96206957e-01 2.60115415e-01 -4.37693357e-01 4.54794019e-01 7.13401496e-01 2.47018831e-03 -3.96212727e-01 -1.04673028e+00 3.96581799e-01 -1.66802251e+00 -5.02482176e-01 -2.94967771e-01 2.47813559e+00 6.10132158e-01 6.67369068e-01 2.83083767e-01 4.25193548e-01 8.86868119e-01 1.06449150e-01 -6.09699070e-01 -1.47511661e-02 -5.03645688e-02 1.65287092e-01 2.37701818e-01 1.44405663e-01 -1.32760990e+00 8.03301871e-01 4.89141989e+00 1.02009118e+00 -9.54926550e-01 1.38413748e-02 8.11452210e-01 4.75862116e-01 -2.09985361e-01 9.14025903e-02 -8.33896458e-01 5.53237140e-01 7.73572505e-01 -1.08857378e-01 1.09679334e-01 1.00252175e+00 -2.14678958e-01 1.67089045e-01 -9.18647945e-01 8.22154701e-01 4.60697472e-01 -6.44171059e-01 1.86466143e-01 2.99458265e-01 6.33876622e-01 -3.29092383e-01 2.05006331e-01 6.06829464e-01 6.44615516e-02 -9.45246220e-01 -1.32332712e-01 2.03962639e-01 7.07762897e-01 -8.14244092e-01 8.30864727e-01 7.65918016e-01 -1.12860906e+00 -7.23913684e-02 -3.01662773e-01 2.31998600e-02 -3.19996774e-01 6.37029409e-01 -7.24354267e-01 7.58938611e-01 5.65433681e-01 7.96472073e-01 -1.12695634e+00 9.76498842e-01 -5.27466059e-01 9.53721821e-01 -2.83945799e-01 1.71993911e-01 2.43075803e-01 -3.00632596e-01 7.91334867e-01 6.91559017e-01 2.19319150e-01 1.59332454e-01 6.15214586e-01 3.90380293e-01 -2.15492353e-01 2.89046615e-01 -9.43551302e-01 -2.67085791e-01 1.56783968e-01 1.25219703e+00 -8.60557079e-01 -9.78278890e-02 -5.92825770e-01 8.75235200e-01 3.00156683e-01 3.94787222e-01 -8.30522537e-01 -4.27415788e-01 1.08792253e-01 4.02978212e-01 -1.43119961e-01 -8.77595916e-02 -2.77310666e-02 -1.31323206e+00 2.39863142e-01 -1.03405046e+00 1.00631464e+00 -2.94250965e-01 -1.57831895e+00 6.05345666e-01 -1.31824851e-01 -1.42603910e+00 -1.86872892e-02 -6.27459764e-01 -8.98579419e-01 5.20328701e-01 -1.13096583e+00 -1.06948030e+00 -5.03994346e-01 8.98274958e-01 1.36984900e-01 -4.12958235e-01 6.29788220e-01 4.64202166e-01 -1.06778646e+00 9.95914876e-01 -2.43956119e-01 5.35697460e-01 6.04957283e-01 -1.20678413e+00 2.08113670e-01 1.51364481e+00 4.02265310e-01 5.81029713e-01 4.02248561e-01 -1.04044318e+00 -1.10974407e+00 -1.39724565e+00 3.69658023e-01 -5.82010269e-01 6.88782215e-01 -4.26344931e-01 -1.31933248e+00 8.59713495e-01 -3.97738606e-01 3.86414766e-01 6.17709935e-01 2.95118093e-01 -2.43187502e-01 -1.57087967e-01 -1.20203447e+00 7.18705714e-01 1.15129280e+00 -4.58328456e-01 -2.81915456e-01 5.76093197e-01 6.11310840e-01 -2.56729990e-01 -4.41481292e-01 6.91058815e-01 -3.10108989e-01 -8.00908267e-01 7.03818798e-01 -8.23252022e-01 -1.01511553e-01 -3.39009047e-01 1.47782862e-01 -1.07470989e+00 -1.00587517e-01 -7.48831272e-01 -5.87035656e-01 1.41744590e+00 3.31992149e-01 -1.02230239e+00 9.52997923e-01 5.57658017e-01 -1.98631331e-01 -5.86225331e-01 -7.13733792e-01 -8.85618508e-01 -2.76275486e-01 -5.02029955e-01 5.39491832e-01 1.31547117e+00 -1.26454711e-01 4.24697936e-01 -3.90714139e-01 6.48042381e-01 6.06760442e-01 -2.70369142e-01 8.73971164e-01 -1.52494395e+00 -1.34892911e-01 -1.63183913e-01 -7.34203041e-01 -8.48638058e-01 9.77697521e-02 -9.86176670e-01 -3.02910209e-01 -1.12977159e+00 -2.26812046e-02 -6.92252338e-01 -6.87701404e-01 8.18228900e-01 -5.27402163e-01 3.55159998e-01 -3.78764451e-01 2.28860959e-01 -6.77963436e-01 4.91033167e-01 1.13348556e+00 -8.80737454e-02 -3.62546712e-01 5.04274368e-01 -9.91352975e-01 9.64072168e-01 8.22587967e-01 -4.19152319e-01 -6.54024124e-01 5.57522289e-02 2.33394444e-01 -4.76588905e-01 4.22598600e-01 -1.10520935e+00 1.26711622e-01 -3.74655263e-03 3.51657927e-01 -3.01710218e-01 -1.95600480e-01 -7.84526229e-01 -3.33607972e-01 3.57487708e-01 1.60607770e-01 5.84885962e-02 1.83530152e-01 1.05615020e+00 -2.48128086e-01 -3.27734321e-01 5.46363592e-01 9.55321267e-02 -8.86521399e-01 9.07864809e-01 -8.63760412e-02 1.63792968e-01 1.20344079e+00 -1.60488576e-01 -3.70094359e-01 -4.97256428e-01 -8.11612844e-01 4.66179281e-01 1.91983819e-01 4.48792219e-01 6.91762030e-01 -1.47404444e+00 -5.69571495e-01 5.94817817e-01 3.97222579e-01 1.98972255e-01 2.49699563e-01 6.78067267e-01 -2.94114321e-01 -9.93046090e-02 -3.53347361e-02 -5.25891304e-01 -1.07994914e+00 8.97346973e-01 2.22094908e-01 -3.14546078e-01 -5.55245578e-01 7.82583117e-01 4.10826802e-01 -4.17056918e-01 2.25827023e-01 2.62191534e-01 -3.62656504e-01 -1.52638465e-01 3.66280556e-01 1.94993630e-01 1.62361920e-01 -7.08836973e-01 -3.73848587e-01 5.41720651e-02 -5.58279932e-01 4.96061713e-01 8.35648060e-01 4.17943820e-02 -2.34865516e-01 1.83858320e-01 7.98546970e-01 3.47474843e-01 -8.39100480e-01 -5.31203210e-01 1.22205079e-01 -5.92494309e-01 -1.56537622e-01 -4.53936666e-01 -1.18156207e+00 7.93039322e-01 5.17391622e-01 7.63463199e-01 1.27051663e+00 -2.45239004e-03 7.80953884e-01 4.79008734e-01 2.58733958e-01 -9.38631833e-01 4.64574218e-01 2.94259548e-01 5.15111446e-01 -1.37715757e+00 3.12160607e-02 -8.11515033e-01 -6.59653246e-01 7.83258617e-01 1.18914187e+00 -2.17051934e-02 7.97515869e-01 -3.50368738e-01 -9.93974954e-02 -4.75333273e-01 -6.32578433e-02 -2.03921497e-01 5.56406140e-01 6.84791267e-01 2.45879684e-02 -4.83986251e-02 -1.28217399e-01 7.44293153e-01 4.84287813e-02 -7.00354397e-01 4.87848043e-01 9.55810726e-01 -3.52349669e-01 -1.14228404e+00 -2.90961653e-01 8.79862964e-01 -5.05784512e-01 3.70926298e-02 -6.46343231e-01 7.86050498e-01 1.84578374e-01 1.01504230e+00 -2.60996222e-01 -6.37887359e-01 4.60286707e-01 4.61015888e-02 -7.06413388e-02 -9.23532367e-01 -1.34445086e-01 -2.05448177e-02 4.86046597e-02 -2.75283664e-01 -7.09492539e-04 -4.83622164e-01 -1.17013514e+00 -2.28726491e-01 -5.42533696e-01 3.43922675e-01 -3.98104824e-02 1.10799718e+00 1.64016619e-01 4.48966891e-01 9.33732092e-01 3.11258342e-02 -4.69071835e-01 -8.14903557e-01 -8.84285867e-01 5.88023782e-01 1.70907184e-01 -6.94228053e-01 -5.33292353e-01 -3.33929658e-01]
[6.63387393951416, 5.774518013000488]
4546730e-24ca-4413-b35c-97844c49b2e3
palm-pre-training-an-autoencoding
2004.07159
null
https://arxiv.org/abs/2004.07159v2
https://arxiv.org/pdf/2004.07159v2.pdf
PALM: Pre-training an Autoencoding&Autoregressive Language Model for Context-conditioned Generation
Self-supervised pre-training, such as BERT, MASS and BART, has emerged as a powerful technique for natural language understanding and generation. Existing pre-training techniques employ autoencoding and/or autoregressive objectives to train Transformer-based models by recovering original word tokens from corrupted text with some masked tokens. The training goals of existing techniques are often inconsistent with the goals of many language generation tasks, such as generative question answering and conversational response generation, for producing new text given context. This work presents PALM with a novel scheme that jointly pre-trains an autoencoding and autoregressive language model on a large unlabeled corpus, specifically designed for generating new text conditioned on context. The new scheme alleviates the mismatch introduced by the existing denoising scheme between pre-training and fine-tuning where generation is more than reconstructing original text. An extensive set of experiments show that PALM achieves new state-of-the-art results on a variety of language generation benchmarks covering generative question answering (Rank 1 on the official MARCO leaderboard), abstractive summarization on CNN/DailyMail as well as Gigaword, question generation on SQuAD, and conversational response generation on Cornell Movie Dialogues.
['Chenliang Li', 'Ming Yan', 'Fei Huang', 'Bin Bi', 'Luo Si', 'Chen Wu', 'Wei Wang', 'Songfang Huang']
2020-04-14
null
null
null
null
['generative-question-answering', 'conversational-response-generation']
['natural-language-processing', 'natural-language-processing']
[ 5.56415737e-01 6.47843003e-01 2.52120495e-01 -5.15659332e-01 -1.49661040e+00 -5.44911504e-01 1.05116868e+00 2.02303808e-02 -2.64263839e-01 1.11311555e+00 9.73568320e-01 -2.69633561e-01 2.20734000e-01 -8.90752792e-01 -7.62750268e-01 -4.63263661e-01 4.08544511e-01 1.01132941e+00 -3.13283086e-01 -8.11268687e-01 9.04730633e-02 -3.59720945e-01 -1.14120877e+00 7.61502326e-01 9.84350801e-01 7.53460348e-01 2.02667654e-01 1.24979365e+00 -4.66761142e-01 1.35076153e+00 -1.16921854e+00 -7.34064996e-01 -1.10329464e-01 -1.10511506e+00 -1.37770593e+00 1.46431088e-01 3.31013411e-01 -4.81717050e-01 -5.21333098e-01 5.81172824e-01 7.81493485e-01 3.83796781e-01 5.84658563e-01 -1.01201153e+00 -1.05584872e+00 1.29241753e+00 1.78630725e-02 2.72632509e-01 5.52154541e-01 2.94748515e-01 1.15963078e+00 -7.94984818e-01 6.90588713e-01 1.49190617e+00 4.40734744e-01 9.09981430e-01 -1.25587368e+00 -2.74641305e-01 -1.12514034e-01 6.82568774e-02 -7.71845162e-01 -7.24901259e-01 8.07137072e-01 -2.32101958e-02 1.18126988e+00 3.64084065e-01 2.78415471e-01 1.69316220e+00 1.98513433e-01 1.06506431e+00 8.26918066e-01 -3.30684274e-01 1.59507558e-01 1.07897846e-02 4.35728617e-02 3.98614377e-01 -3.09431374e-01 -1.86602265e-01 -6.14289343e-01 -3.72855216e-01 2.84144521e-01 -4.83554512e-01 -2.51879424e-01 4.34005678e-01 -1.39676547e+00 1.15814137e+00 1.25481695e-01 1.15654558e-01 -5.31928062e-01 2.32785881e-01 6.09235346e-01 6.15696371e-01 8.39434505e-01 7.82505214e-01 -1.52039781e-01 -2.78894514e-01 -9.96613443e-01 7.15635061e-01 1.08362567e+00 1.07337046e+00 6.15836501e-01 6.58090293e-01 -8.54349434e-01 9.76467431e-01 -2.81090867e-02 4.25221622e-01 9.97613311e-01 -9.03006136e-01 7.88367033e-01 2.71539211e-01 4.29554880e-02 -5.48636258e-01 1.04295015e-01 -4.26895291e-01 -1.28409386e+00 -5.71266055e-01 1.62937343e-01 -6.86824441e-01 -8.77657413e-01 1.59000015e+00 1.70805842e-01 1.43814292e-02 7.05587208e-01 5.86885929e-01 1.33923292e+00 1.18387210e+00 -2.79142916e-01 -2.82728761e-01 1.18655455e+00 -1.42010558e+00 -8.00986767e-01 -3.87058169e-01 3.32972944e-01 -8.27093959e-01 1.01776862e+00 3.18915963e-01 -1.50489080e+00 -6.98720276e-01 -6.51541710e-01 -5.11673868e-01 2.49775145e-02 -8.03419948e-02 4.35621053e-01 2.55182177e-01 -1.18371332e+00 3.61050189e-01 -4.25275117e-01 -1.34516165e-01 4.51245345e-02 -6.16681837e-02 -1.90771878e-01 -3.84940393e-02 -1.47994363e+00 8.08082998e-01 4.34081793e-01 1.92407504e-01 -1.08225417e+00 -6.63718343e-01 -9.60264146e-01 1.56769991e-01 3.52904052e-01 -1.23446500e+00 1.88134146e+00 -8.94928753e-01 -2.01056886e+00 5.66121101e-01 -2.17903033e-01 -1.13073456e+00 5.11443496e-01 -2.76417166e-01 -1.84169069e-01 1.24049045e-01 2.02501357e-01 8.10999095e-01 1.02440119e+00 -1.07363701e+00 -4.23772991e-01 -3.19116637e-02 6.95470124e-02 4.07911569e-01 2.51592923e-04 -2.36538589e-01 7.12474138e-02 -9.05325592e-01 -3.63417178e-01 -4.89340693e-01 -3.74382377e-01 -1.01075482e+00 -6.32616460e-01 -6.11403823e-01 7.80936003e-01 -9.43518221e-01 8.80060077e-01 -1.52580416e+00 2.97242075e-01 -3.62089306e-01 2.35691518e-02 1.75521478e-01 -5.75114965e-01 9.30417478e-01 -3.58036309e-02 -1.12802228e-02 -2.76766598e-01 -7.76051164e-01 1.38687611e-01 3.09804946e-01 -1.17077613e+00 -2.11996347e-01 5.62866032e-01 1.21974838e+00 -1.12596750e+00 -2.67691761e-01 -1.54100761e-01 3.26598227e-01 -6.38034940e-01 7.27829576e-01 -7.90731490e-01 4.88523692e-01 -2.65468061e-01 9.39943939e-02 1.30027324e-01 -2.81488448e-01 -1.19394287e-02 2.76672810e-01 2.26512045e-01 9.30980504e-01 -7.65681505e-01 1.92115951e+00 -6.81457162e-01 5.90478361e-01 4.76964191e-02 -1.10360456e+00 1.13217413e+00 7.25803852e-01 2.83994712e-02 -4.71334219e-01 1.53904902e-02 -7.00787874e-03 -2.14291036e-01 -5.54008782e-01 1.22502065e+00 -3.06109935e-01 -3.10397208e-01 8.16675246e-01 5.70303500e-01 -7.35611916e-01 4.52731639e-01 6.75580382e-01 1.17719603e+00 -1.67835795e-04 1.21719532e-01 2.63276640e-02 4.66412485e-01 1.51503071e-01 1.49196267e-01 9.89399850e-01 4.82360631e-01 8.70947659e-01 5.79535723e-01 -1.55844510e-01 -9.98510838e-01 -8.71433556e-01 5.16235411e-01 1.29124808e+00 -3.67494076e-01 -4.66513544e-01 -8.63453329e-01 -5.53505540e-01 -4.26083982e-01 1.22206306e+00 -5.41881919e-01 -4.11419004e-01 -7.54212558e-01 -6.23234808e-01 8.60557199e-01 4.22262251e-01 5.27339697e-01 -1.51843691e+00 -2.98687428e-01 7.48133600e-01 -9.24898088e-01 -1.15380001e+00 -6.43005192e-01 -2.41105929e-02 -7.89261699e-01 -5.80114424e-01 -8.55246842e-01 -8.79746497e-01 4.44898933e-01 1.08410910e-01 1.58251965e+00 -1.51336923e-01 4.41492461e-02 4.15136009e-01 -5.56268871e-01 -3.06352288e-01 -1.19676280e+00 5.58686256e-01 -3.85978043e-01 1.15501292e-01 -2.84779459e-01 -6.16594315e-01 -3.86213750e-01 -1.45384476e-01 -1.12494504e+00 4.36009198e-01 6.05339587e-01 1.39074850e+00 3.25826198e-01 -5.05377412e-01 1.20309627e+00 -1.09716666e+00 1.42260540e+00 -6.14262104e-01 -1.42503023e-01 3.03908110e-01 -2.79955119e-01 3.88953716e-01 9.06083107e-01 -3.74413759e-01 -1.42717719e+00 -4.86348778e-01 -3.95372301e-01 -7.43183568e-02 -7.90358707e-02 5.78095734e-01 1.54443130e-01 8.38682711e-01 9.49568391e-01 6.43023014e-01 -9.88280922e-02 -3.49076658e-01 7.84157634e-01 7.49640763e-01 1.05179787e+00 -6.23127520e-01 6.83308244e-01 4.49259244e-02 -6.00301325e-01 -6.30896986e-01 -9.73565638e-01 -4.01001036e-01 -8.06119293e-02 7.39615560e-02 7.59358346e-01 -9.51821327e-01 -2.74596632e-01 4.45596129e-01 -1.64133155e+00 -4.92535353e-01 -8.33252072e-01 6.84663747e-03 -7.45369077e-01 3.91220033e-01 -7.94659376e-01 -6.68187618e-01 -1.29357266e+00 -7.44476795e-01 1.23444629e+00 1.38560772e-01 -4.25288528e-01 -9.96263623e-01 3.72012705e-01 9.16767716e-01 7.14411557e-01 1.82122707e-01 7.46141970e-01 -8.59292686e-01 -3.89915019e-01 -9.49431807e-02 2.57734567e-01 6.16188288e-01 3.49859186e-02 -4.26669389e-01 -8.99092555e-01 -2.17506945e-01 2.48457626e-01 -8.12740624e-01 9.59884584e-01 -8.82215053e-03 8.91134381e-01 -8.79850030e-01 2.89823204e-01 1.78513169e-01 8.45503092e-01 -1.57616675e-01 7.30722368e-01 -2.66524941e-01 4.24155742e-01 5.69804847e-01 2.40449011e-01 5.00855923e-01 6.81764305e-01 3.31495672e-01 2.37665161e-01 5.38240932e-02 -1.23783499e-01 -5.62241733e-01 6.05644166e-01 1.10623825e+00 1.59288555e-01 -7.45102525e-01 -6.45355463e-01 7.94462979e-01 -1.88239133e+00 -1.22732210e+00 -3.57124470e-02 1.74386644e+00 1.38972783e+00 -6.41234592e-02 5.92187643e-02 -1.79288968e-01 6.09288096e-01 4.56731021e-01 -4.59810168e-01 -5.30965149e-01 -2.21241787e-01 6.37833595e-01 -1.13791198e-01 7.45991051e-01 -7.08378553e-01 1.14533007e+00 5.98948383e+00 7.72387087e-01 -8.66127372e-01 3.17129850e-01 7.61537433e-01 -1.44404590e-01 -5.73725045e-01 -5.52355275e-02 -6.66408420e-01 2.50092894e-01 1.17380786e+00 -6.80902004e-01 4.64889556e-01 7.22097933e-01 2.60987431e-01 1.67236999e-01 -1.15435135e+00 8.20683956e-01 5.83463669e-01 -1.71119452e+00 3.85685682e-01 -3.77921432e-01 1.01560354e+00 -2.68922206e-02 -6.04890138e-02 8.28934729e-01 8.90614510e-01 -1.17197382e+00 5.89910924e-01 4.53164667e-01 5.84025800e-01 -4.83344346e-01 7.97107041e-01 6.67652786e-01 -6.32384658e-01 1.39063686e-01 -3.60691249e-01 -1.35633603e-01 5.75257361e-01 5.35635591e-01 -1.41821218e+00 7.77290404e-01 1.25742331e-01 3.83955151e-01 -3.95518273e-01 3.81250858e-01 -4.69534457e-01 9.28845823e-01 -4.09624204e-02 -4.63574156e-02 4.32589203e-01 5.04478589e-02 6.43935204e-01 1.21337950e+00 2.04954371e-01 1.22877300e-01 4.38226648e-02 1.03016472e+00 -5.73623359e-01 1.84678528e-02 -4.10032928e-01 -2.30685607e-01 1.25888884e-01 1.29743588e+00 -8.89183134e-02 -8.36031735e-01 1.45650044e-01 1.05241489e+00 2.19908372e-01 4.21368510e-01 -7.08758891e-01 -3.66722703e-01 3.23708765e-02 -4.64115143e-02 1.36672303e-01 -1.38997614e-01 -2.75583938e-02 -1.34722269e+00 1.44941295e-02 -1.32452381e+00 3.66605192e-01 -9.27551270e-01 -1.38910949e+00 9.56787348e-01 5.68016879e-02 -7.33600438e-01 -1.33499765e+00 1.38046041e-01 -9.84445512e-01 9.18559968e-01 -1.34706092e+00 -1.24987888e+00 -2.62053221e-01 5.41896403e-01 1.34811497e+00 -3.75707537e-01 9.16774392e-01 -1.04947403e-01 -2.39871651e-01 4.87305045e-01 1.43825961e-02 1.34364724e-01 8.15670013e-01 -1.46761465e+00 9.34930861e-01 8.58208001e-01 3.66012573e-01 3.07632476e-01 7.64226019e-01 -3.54272246e-01 -1.32770717e+00 -1.26275468e+00 1.21214259e+00 -4.14118886e-01 5.04164636e-01 -5.46873569e-01 -9.04002190e-01 7.61037767e-01 9.83362377e-01 -5.61482072e-01 4.97830272e-01 -2.34631956e-01 -9.17884037e-02 -2.18447186e-02 -7.38704979e-01 6.07247770e-01 7.06101596e-01 -4.80809569e-01 -1.07629764e+00 7.13412404e-01 1.06451428e+00 -6.49304509e-01 -5.72406769e-01 1.74184874e-01 8.86279866e-02 -7.08544254e-01 6.91035092e-01 -9.45000112e-01 1.02440155e+00 2.09530190e-01 1.24052145e-01 -1.66409636e+00 4.65667807e-02 -1.48503435e+00 -2.01190963e-01 1.55982828e+00 4.66152281e-01 -4.73473340e-01 6.36701286e-01 3.00564855e-01 -5.41828811e-01 -5.75003743e-01 -8.42107415e-01 -3.29010695e-01 2.33036220e-01 -1.20422587e-01 6.24254644e-01 6.09994113e-01 -1.17123723e-01 1.23749006e+00 -5.94235241e-01 -4.35690731e-01 2.47877225e-01 1.40242264e-01 1.27000487e+00 -6.93888605e-01 -6.09107912e-01 -2.56982386e-01 2.16178477e-01 -1.46137214e+00 4.33275074e-01 -9.36274052e-01 4.64129061e-01 -1.94043255e+00 -9.80248824e-02 1.93912238e-01 3.54416281e-01 3.80802959e-01 -3.63646746e-01 -4.44754399e-02 7.63528720e-02 -4.02739272e-02 -5.61530650e-01 1.08450389e+00 1.24435234e+00 -4.20239270e-01 -1.83155879e-01 1.38730273e-01 -1.12766635e+00 3.26325446e-01 6.82008445e-01 -3.50316852e-01 -6.45448446e-01 -7.09925115e-01 3.28616887e-01 6.08313978e-01 3.84002209e-01 -6.44448161e-01 3.10764790e-01 1.27849385e-01 -1.04022108e-01 -6.21309459e-01 3.37484717e-01 7.48333288e-03 -1.05734810e-01 2.49370754e-01 -9.92679775e-01 2.23736912e-01 4.85678250e-03 5.88033855e-01 -5.56143761e-01 -2.76049793e-01 4.78890896e-01 -3.84208709e-01 -3.19397092e-01 2.68816818e-02 -4.84352767e-01 7.63784826e-01 2.89142936e-01 2.40653411e-01 -5.82493424e-01 -1.21952903e+00 -5.90874493e-01 4.81521547e-01 -2.75758803e-01 7.06600904e-01 6.96430027e-01 -1.16181815e+00 -1.56640279e+00 -1.29354537e-01 -2.51090348e-01 5.04581332e-01 4.42677796e-01 3.94659519e-01 -2.96992660e-01 4.04029131e-01 2.07646579e-01 -3.08478415e-01 -1.05898333e+00 1.08001254e-01 1.87556252e-01 -1.09674942e+00 -5.45076668e-01 9.22882080e-01 1.70277685e-01 -4.95479047e-01 -2.93242801e-02 -4.23421651e-01 -9.68005061e-02 4.43488806e-02 4.84853208e-01 1.69259042e-01 1.55392721e-01 -3.75868052e-01 3.46301943e-01 -3.46542031e-01 -4.40369099e-01 -5.27146757e-01 1.25380898e+00 -6.84552565e-02 -1.63092718e-01 3.42815489e-01 9.24915493e-01 -2.71139979e-01 -9.20904636e-01 -4.66506749e-01 -2.48876169e-01 1.77493170e-01 -2.88702250e-01 -8.88337672e-01 -6.29416823e-01 9.11185503e-01 -2.42192388e-01 4.19889271e-01 8.35094810e-01 5.19138984e-02 1.34814405e+00 8.27622175e-01 -4.10148390e-02 -9.82342184e-01 6.21434152e-01 1.03568709e+00 1.54594231e+00 -1.06831026e+00 -3.99114579e-01 1.48763489e-02 -9.35524225e-01 9.66502190e-01 6.08268261e-01 -2.50902295e-01 -6.21519610e-02 4.42639850e-02 -6.97539821e-02 5.53337969e-02 -1.39444089e+00 1.10815898e-01 2.31346607e-01 4.24448967e-01 4.89582270e-01 -7.38670677e-02 -5.90794347e-02 7.86866009e-01 -1.04267836e+00 -1.75169706e-01 8.25042129e-01 6.12202823e-01 -2.49496579e-01 -1.04646087e+00 -2.20647156e-01 4.38867390e-01 -4.73812044e-01 -5.16434133e-01 -6.83811367e-01 4.35618579e-01 -4.01604533e-01 1.31504047e+00 1.34561449e-01 -1.24528758e-01 1.92484036e-01 4.24739480e-01 2.67837077e-01 -1.01723039e+00 -1.07581961e+00 5.31680174e-02 5.61570644e-01 -5.55994213e-02 -2.81080514e-01 -2.92937905e-01 -1.01561999e+00 -1.20192200e-01 -3.30606014e-01 6.57675087e-01 3.20917606e-01 1.01166129e+00 5.61798871e-01 6.74612224e-01 6.89146459e-01 -6.02181733e-01 -1.18909228e+00 -1.53923106e+00 -4.91316132e-02 4.26462501e-01 3.13291728e-01 2.40010872e-01 -1.61305726e-01 4.89166349e-01]
[11.934329986572266, 8.935415267944336]
8b283d9d-3aa7-42c0-9c4c-a13bbe85a752
zero-shot-action-recognition-in-videos-a
1909.06423
null
https://arxiv.org/abs/1909.06423v2
https://arxiv.org/pdf/1909.06423v2.pdf
Zero-Shot Action Recognition in Videos: A Survey
Zero-Shot Action Recognition has attracted attention in the last years and many approaches have been proposed for recognition of objects, events and actions in images and videos. There is a demand for methods that can classify instances from classes that are not present in the training of models, especially in the complex problem of automatic video understanding, since collecting, annotating and labeling videos are difficult and laborious tasks. We have identified that there are many methods available in the literature, however, it is difficult to categorize which techniques can be considered state of the art. Despite the existence of some surveys about zero-shot action recognition in still images and experimental protocol, there is no work focused on videos. Therefore, we present a survey of the methods that comprise techniques to perform visual feature extraction and semantic feature extraction as well to learn the mapping between these features considering specifically zero-shot action recognition in videos. We also provide a complete description of datasets, experiments and protocols, presenting open issues and directions for future work, essential for the development of the computer vision research field.
['Valter Estevam', 'Helio Pedrini', 'David Menotti']
2019-09-13
null
null
null
null
['zero-shot-action-recognition', 'action-recognition-in-still-images']
['computer-vision', 'computer-vision']
[ 6.07136965e-01 -3.18180978e-01 -3.90412033e-01 -4.49593753e-01 -3.17090213e-01 -3.71483415e-01 5.61312497e-01 -1.10761523e-01 -3.40721160e-01 5.21664798e-01 2.06211686e-01 2.89496005e-01 -2.32985944e-01 -4.78833675e-01 -4.44813937e-01 -8.43394637e-01 -1.87610537e-01 7.49797076e-02 6.34991765e-01 1.70624822e-01 5.25296330e-01 5.44791162e-01 -2.39913726e+00 5.98130584e-01 2.74004549e-01 9.81714487e-01 1.67816773e-01 7.05430567e-01 -1.13663062e-01 1.37267375e+00 -5.92527807e-01 -1.79216713e-01 1.59421891e-01 -7.67817259e-01 -9.81775641e-01 7.65435874e-01 5.87320685e-01 -4.03434604e-01 -4.09555554e-01 1.00086665e+00 1.86529726e-01 3.29450279e-01 7.32746243e-01 -1.45835435e+00 -4.23934966e-01 2.00409979e-01 -1.67930946e-01 6.97151661e-01 5.98293483e-01 -1.01202220e-01 6.54280305e-01 -8.04759979e-01 7.04967320e-01 9.31074083e-01 2.77291209e-01 6.22723699e-01 -5.54795980e-01 -2.84190714e-01 1.00433312e-01 9.99988496e-01 -1.25340617e+00 -6.44470453e-01 7.42063522e-01 -7.33362854e-01 1.03793478e+00 2.88926840e-01 7.48790801e-01 1.05628860e+00 -6.57524243e-02 1.20868254e+00 8.64028215e-01 -7.54923046e-01 2.86236852e-01 1.23935953e-01 3.07714880e-01 5.72125375e-01 -2.14964803e-02 -1.62562758e-01 -6.35877490e-01 6.91013783e-02 7.19855070e-01 3.17845523e-01 -2.27976799e-01 -7.70489097e-01 -1.16230857e+00 5.98455846e-01 -5.47446646e-02 7.99957395e-01 -4.86042261e-01 -1.85142174e-01 7.17669606e-01 3.44719172e-01 1.87790841e-01 1.32621408e-01 -2.44523436e-01 -6.10240102e-01 -8.98446679e-01 1.52844548e-01 6.09585643e-01 8.66639495e-01 6.63737237e-01 -1.37546705e-03 -1.84433267e-01 7.17828274e-01 -1.26521572e-01 2.35842958e-01 5.68509161e-01 -1.05069065e+00 8.65605995e-02 6.87940538e-01 3.69212627e-02 -1.25793326e+00 -1.52873144e-01 4.69126403e-01 -4.27029729e-01 8.81999657e-02 3.46425772e-01 1.89883158e-01 -7.07384109e-01 1.10111976e+00 1.32099867e-01 5.08781731e-01 1.30914569e-01 9.42000628e-01 1.04128635e+00 4.79807258e-01 1.34205297e-01 -4.98497337e-01 1.34854782e+00 -1.05849230e+00 -1.00662780e+00 -1.70616373e-01 6.06671274e-01 -6.94260299e-01 5.82221746e-01 2.59841710e-01 -8.83856416e-01 -7.21364915e-01 -8.58889878e-01 1.01885386e-01 -5.91632962e-01 2.27434814e-01 6.86420143e-01 5.61279774e-01 -6.72371209e-01 6.21345580e-01 -1.05418694e+00 -1.01845479e+00 4.36039746e-01 2.01291993e-01 -5.52189052e-01 -1.86484173e-01 -9.03631389e-01 9.37202036e-01 5.67060411e-01 2.05416251e-02 -1.07650876e+00 -3.44294347e-02 -9.08070087e-01 -2.11449787e-01 7.72553504e-01 1.43238558e-02 1.01398158e+00 -1.39034498e+00 -1.20670259e+00 1.03732693e+00 -1.15731329e-01 -4.93104994e-01 2.00584993e-01 -1.00673046e-02 -6.06430769e-01 4.98019964e-01 -8.18953961e-02 4.76388782e-01 7.97718048e-01 -8.32081378e-01 -9.02897835e-01 -3.87091547e-01 4.39637423e-01 2.01007530e-01 -4.29814994e-01 4.75128204e-01 -4.08838123e-01 -4.65529710e-01 -2.74061505e-02 -7.36582398e-01 3.28799039e-02 6.47481233e-02 1.22903652e-01 -4.46347356e-01 1.00840878e+00 -5.43736100e-01 1.12272632e+00 -2.32868028e+00 2.20086813e-01 -3.10328335e-01 -2.79484957e-01 6.16541266e-01 -2.41072215e-02 6.79358184e-01 -1.08597487e-01 -2.73532480e-01 -1.50840357e-01 1.08556718e-01 -2.21653163e-01 4.45390373e-01 -2.88319200e-01 5.80939293e-01 5.39392885e-03 6.72460735e-01 -9.76230979e-01 -7.20919192e-01 9.27577019e-01 3.03835958e-01 -1.50659934e-01 2.82151818e-01 7.48703256e-02 4.19471711e-01 -6.40400767e-01 8.56080413e-01 2.38866687e-01 -7.52241444e-03 8.15694928e-02 -1.22843206e-01 -2.11811811e-01 -7.42184371e-02 -1.43714213e+00 1.51057041e+00 9.07201543e-02 7.91766107e-01 -2.89640188e-01 -1.56403875e+00 5.87482452e-01 6.85208440e-01 7.93457091e-01 -4.27596897e-01 2.65888274e-01 5.99396788e-02 -1.37485623e-01 -1.24673593e+00 4.41858798e-01 -9.97057706e-02 1.19588062e-01 2.64826059e-01 4.17626798e-01 2.67002493e-01 7.30254710e-01 -1.92730650e-01 9.79944050e-01 2.24865586e-01 6.54237568e-01 1.44531131e-01 9.66711402e-01 1.37795374e-01 4.31964219e-01 5.33427656e-01 -6.04343772e-01 4.50559884e-01 3.01383764e-01 -8.73850465e-01 -7.76706338e-01 -5.83895564e-01 -8.08961987e-02 1.18983138e+00 2.64442086e-01 -4.09774780e-01 -7.70390093e-01 -7.44732559e-01 -3.54267478e-01 2.67672151e-01 -7.55274534e-01 9.27369967e-02 -4.63857204e-01 -4.40629959e-01 2.44797662e-01 6.82621717e-01 5.80268919e-01 -1.46899390e+00 -9.64131713e-01 -6.54115528e-02 -2.10250065e-01 -1.31973648e+00 -5.53422375e-04 -1.17561646e-01 -9.46427345e-01 -1.54552877e+00 -6.53153419e-01 -9.77847159e-01 7.30356216e-01 6.84208572e-01 8.84615183e-01 1.55481637e-01 -6.71263218e-01 9.05321479e-01 -8.95872056e-01 -3.28238070e-01 -1.98460892e-01 -3.36767197e-01 -4.69232537e-02 3.63370538e-01 9.81958628e-01 -1.87519774e-01 -4.55477446e-01 5.03460050e-01 -1.04837906e+00 -2.01240376e-01 4.68416333e-01 4.29639846e-01 4.78739709e-01 1.64412692e-01 2.81188965e-01 -5.59312165e-01 -2.52935849e-02 -2.65276790e-01 -3.27888370e-01 4.80494380e-01 -1.74146853e-02 -3.47630352e-01 3.81813794e-01 -3.09974462e-01 -9.21353281e-01 4.07993048e-01 7.20889196e-02 -4.72393960e-01 -8.16285610e-01 8.69902447e-02 -1.35687307e-01 -3.17800790e-01 4.18901235e-01 4.90442067e-01 -4.16367985e-02 -4.02816832e-01 3.07355613e-01 9.39162135e-01 2.26628557e-01 -6.26469264e-03 2.27445081e-01 7.46355593e-01 -2.31929615e-01 -1.33937371e+00 -1.01625848e+00 -1.06158853e+00 -1.16385436e+00 -6.88597620e-01 1.24151540e+00 -5.54814219e-01 -4.19545680e-01 6.48726344e-01 -8.72346878e-01 9.42353681e-02 -3.25125784e-01 6.70615673e-01 -1.03355432e+00 6.79837823e-01 -2.69397199e-01 -8.76431584e-01 4.74233069e-02 -1.13300288e+00 8.48422348e-01 2.54392952e-01 -2.65748024e-01 -1.00136399e+00 9.49789509e-02 6.70955420e-01 1.07076220e-01 1.90984756e-01 5.77502131e-01 -5.78185558e-01 -5.87455034e-01 -3.83506030e-01 9.20935869e-02 5.74804366e-01 3.88800234e-01 1.10863782e-01 -9.75150585e-01 -1.71508208e-01 2.11143851e-01 -6.35957122e-01 6.93539739e-01 5.85797071e-01 1.20614278e+00 -7.05271885e-02 -4.35404778e-01 2.18699381e-01 1.34419203e+00 6.03932500e-01 9.46909666e-01 2.24416018e-01 3.73634011e-01 7.59026825e-01 1.03379643e+00 3.73417616e-01 1.67625491e-02 7.85692394e-01 3.78109336e-01 1.93743587e-01 -9.05534178e-02 7.74872079e-02 4.38620895e-01 6.17379546e-01 -7.14931667e-01 -1.30025983e-01 -4.95066792e-01 6.07544422e-01 -1.92361426e+00 -1.59930038e+00 -5.40165342e-02 2.23795795e+00 3.26264143e-01 -2.54833698e-01 2.57240057e-01 4.80645210e-01 8.58352780e-01 4.08598483e-01 -2.02660769e-01 -1.17421538e-01 1.07782103e-01 -1.50667936e-01 2.20825762e-01 1.12683629e-03 -1.66468298e+00 9.56350684e-01 7.15485573e+00 6.02594733e-01 -1.06470954e+00 3.03546656e-02 2.19101712e-01 1.63012132e-01 7.33680487e-01 -5.05842604e-02 -7.21962154e-01 3.05921495e-01 7.77530611e-01 -4.74838503e-02 2.14207947e-01 1.03835070e+00 1.27408549e-01 -3.97460192e-01 -1.22923315e+00 1.21157050e+00 8.85722637e-01 -1.06794786e+00 1.77181270e-02 -2.44300872e-01 6.76050544e-01 -2.50517070e-01 -4.08204943e-01 2.40510449e-01 -3.23447943e-01 -6.76458776e-01 3.35698068e-01 5.37253499e-01 3.31862599e-01 -5.65103531e-01 8.15147638e-01 4.13103014e-01 -1.27262247e+00 -2.45653421e-01 -6.04084849e-01 -3.55067760e-01 2.28121713e-01 -7.73637593e-02 -4.11409855e-01 5.15724957e-01 7.87279904e-01 1.35847485e+00 -5.38396537e-01 1.31994045e+00 -1.11919604e-01 3.74265105e-01 3.37398559e-01 -9.52746421e-02 7.67322630e-02 -2.24982858e-01 3.68035078e-01 9.80680585e-01 2.39058867e-01 2.97573000e-01 5.86335123e-01 1.97546467e-01 4.56054181e-01 2.60907322e-01 -8.18091810e-01 -5.23814619e-01 -9.91002992e-02 1.13424408e+00 -1.11179233e+00 -5.85260510e-01 -8.92619193e-01 1.14812410e+00 -8.43574852e-02 8.11223760e-02 -8.57640982e-01 -3.24320942e-01 6.63288474e-01 5.57152890e-02 4.52782065e-01 -2.73402601e-01 3.56106013e-01 -1.34216130e+00 -6.29145280e-03 -8.07328522e-01 7.09549427e-01 -6.97572708e-01 -1.03709674e+00 4.38177973e-01 3.82470906e-01 -1.71279001e+00 -4.00196373e-01 -8.72843325e-01 -2.61998445e-01 -1.10348715e-02 -1.13970447e+00 -8.57918620e-01 -5.28536558e-01 7.51881838e-01 1.14246881e+00 -2.57712066e-01 9.48497593e-01 4.40631390e-01 -3.73547375e-01 -7.81095624e-02 -2.11385749e-02 4.65272576e-01 5.71551323e-01 -7.01525629e-01 -2.22796902e-01 7.63601661e-01 5.50763845e-01 2.74024904e-01 7.07563043e-01 -4.95629907e-01 -1.45862877e+00 -7.81264901e-01 8.79939497e-01 -4.32539374e-01 6.30006492e-01 -4.55809981e-02 -8.93523216e-01 8.22684944e-01 2.00981140e-01 1.97179645e-01 8.98744106e-01 -2.69123495e-01 4.54415604e-02 2.74011873e-05 -8.75064313e-01 2.74278283e-01 1.07201445e+00 -4.14181113e-01 -9.09920871e-01 5.32002032e-01 -7.39171281e-02 -5.26149347e-02 -6.67229533e-01 3.47221613e-01 5.46925426e-01 -1.00715196e+00 8.99090111e-01 -7.73838699e-01 1.68987244e-01 -4.43378359e-01 -1.44065931e-01 -7.93747544e-01 -2.13050306e-01 -5.83456596e-04 -1.24603733e-01 1.09033525e+00 -2.40570918e-01 -2.22884700e-01 6.94579959e-01 3.80736351e-01 2.66095698e-02 -3.71175468e-01 -8.10195446e-01 -8.92070472e-01 -5.18693089e-01 -4.25621867e-01 -1.14654832e-01 7.90903449e-01 7.71484077e-02 9.97554362e-02 -7.09729373e-01 -1.90851867e-01 4.90411550e-01 1.59187421e-01 7.63108253e-01 -1.12944794e+00 6.73584221e-03 -2.55445659e-01 -1.34784567e+00 -8.12926471e-01 1.97431237e-01 -5.31434536e-01 8.28497410e-02 -1.74637341e+00 5.80667973e-01 2.91147500e-01 -2.87690252e-01 4.50866193e-01 2.30937868e-01 5.81187010e-01 2.26193652e-01 3.56585830e-01 -1.07007790e+00 4.93898809e-01 1.14093053e+00 -1.91972718e-01 1.42703101e-01 8.30911174e-02 -6.98399693e-02 1.03630793e+00 4.65228945e-01 -3.55731905e-01 -5.76159298e-01 -1.86231062e-01 -3.03800672e-01 -3.06418110e-02 3.32667202e-01 -1.41102827e+00 1.67901069e-01 -4.53967512e-01 3.08031529e-01 -6.93612337e-01 5.89053690e-01 -1.06086874e+00 2.14569904e-02 3.39613289e-01 -2.76891798e-01 -2.23168716e-01 -2.12108135e-01 6.33371353e-01 -6.36743486e-01 -6.55631959e-01 8.39028239e-01 -5.28798997e-01 -1.62708354e+00 2.60012418e-01 -8.15685749e-01 -1.03086121e-01 1.78921163e+00 -6.72776222e-01 7.74259642e-02 -2.56729275e-01 -9.65054870e-01 -4.63731401e-02 4.90111738e-01 7.64266193e-01 7.01969981e-01 -1.16477311e+00 -3.56075823e-01 2.46231332e-01 4.00416702e-01 -6.20432973e-01 7.22707152e-01 8.85465205e-01 -6.15941644e-01 5.65416038e-01 -6.62809551e-01 -5.40356338e-01 -1.86015880e+00 1.02600205e+00 1.46794021e-01 2.69869506e-01 -5.82777321e-01 3.53364617e-01 9.62787792e-02 2.03518167e-01 5.31835318e-01 -8.05720836e-02 -8.87046337e-01 1.56106740e-01 1.09106505e+00 6.50165081e-01 -9.67919230e-02 -1.24422932e+00 -5.45283318e-01 7.48149753e-01 1.89770624e-01 3.32670152e-01 1.30654120e+00 -1.87676743e-01 -1.94152653e-01 8.51509809e-01 1.01604891e+00 -6.04216397e-01 -1.02791333e+00 -7.66522484e-04 1.92858547e-01 -8.14551055e-01 -2.46878818e-01 -3.47573519e-01 -1.02815485e+00 9.77961540e-01 8.04065287e-01 4.21968222e-01 1.25539768e+00 1.70651138e-01 3.92428070e-01 4.07280505e-01 7.19794989e-01 -1.34541202e+00 3.39404970e-01 5.69478989e-01 7.56836593e-01 -1.34698308e+00 -1.92010850e-02 -5.75405657e-01 -7.50484645e-01 1.31314898e+00 5.15291274e-01 -3.31385225e-01 6.73141003e-01 5.46476617e-02 2.80688945e-02 -1.93474516e-01 -4.80560064e-01 -4.83089715e-01 2.52682954e-01 7.27676868e-01 4.80653554e-01 -1.80432618e-01 -6.14721775e-01 1.60584971e-01 3.98895979e-01 4.47319001e-01 5.12821019e-01 1.31186771e+00 -7.70795643e-01 -1.14306295e+00 -2.23423854e-01 3.11990470e-01 -5.44651389e-01 4.47834820e-01 -4.51478034e-01 6.98420882e-01 3.16088170e-01 1.14343750e+00 1.73310682e-01 -3.35126787e-01 2.11216003e-01 3.28551680e-01 7.28277385e-01 -7.51373887e-01 -7.42696598e-02 -3.34988385e-01 3.95828262e-02 -7.10845530e-01 -1.29966879e+00 -8.91575813e-01 -1.00747013e+00 2.15282038e-01 -2.38257751e-01 6.50870353e-02 5.10119319e-01 1.21486461e+00 9.88714918e-02 2.83185333e-01 3.59276801e-01 -9.08171952e-01 -9.85503644e-02 -9.83771026e-01 -8.28795314e-01 7.38054514e-01 -2.68109776e-02 -1.09885883e+00 -2.94177264e-01 6.62705421e-01]
[8.219636917114258, 0.48429521918296814]
7402385b-e7da-4c7a-b3ab-de085703670d
tea-program-repair-using-neural-network-based
2107.08262
null
https://arxiv.org/abs/2107.08262v1
https://arxiv.org/pdf/2107.08262v1.pdf
Tea: Program Repair Using Neural Network Based on Program Information Attention Matrix
The advance in machine learning (ML)-driven natural language process (NLP) points a promising direction for automatic bug fixing for software programs, as fixing a buggy program can be transformed to a translation task. While software programs contain much richer information than one-dimensional natural language documents, pioneering work on using ML-driven NLP techniques for automatic program repair only considered a limited set of such information. We hypothesize that more comprehensive information of software programs, if appropriately utilized, can improve the effectiveness of ML-driven NLP approaches in repairing software programs. As the first step towards proving this hypothesis, we propose a unified representation to capture the syntax, data flow, and control flow aspects of software programs, and devise a method to use such a representation to guide the transformer model from NLP in better understanding and fixing buggy programs. Our preliminary experiment confirms that the more comprehensive information of software programs used, the better ML-driven NLP techniques can perform in fixing bugs in these programs.
['Yang Zhang', 'Liang Cheng', 'Chen Wu', 'Wenshuo Wang']
2021-07-17
null
null
null
null
['program-repair', 'program-repair']
['computer-code', 'reasoning']
[ 3.46489362e-02 3.39702368e-01 -5.61781168e-01 -3.40778440e-01 -7.91264713e-01 -7.49326050e-01 1.12272300e-01 4.45798665e-01 3.56644213e-01 4.68746632e-01 2.62376934e-01 -8.18721354e-01 3.13937031e-02 -9.64903593e-01 -9.35856640e-01 1.70535043e-01 -1.58047751e-01 8.47674236e-02 2.23180026e-01 -1.09767981e-01 6.48182392e-01 9.22243968e-02 -1.40369058e+00 3.65260035e-01 8.28309059e-01 -8.58987868e-02 3.08618784e-01 7.43357241e-01 -7.64775872e-01 1.11653411e+00 -6.68140292e-01 -3.69091690e-01 1.35866571e-02 -3.30077499e-01 -1.13430226e+00 -1.58526286e-01 8.19357112e-02 -7.66348243e-02 3.39706004e-01 1.41144300e+00 -3.44153196e-01 -5.11017799e-01 3.86657566e-01 -1.26368332e+00 -7.75467753e-01 1.03503036e+00 -4.51411307e-01 -7.67471865e-02 1.02937078e+00 2.09842667e-01 1.12203670e+00 -6.01372480e-01 6.70656085e-01 1.39977837e+00 6.47492886e-01 8.09993327e-01 -1.11796927e+00 -6.14087917e-02 1.70948226e-02 -1.95337102e-01 -1.13862646e+00 2.68406477e-02 7.95544863e-01 -8.87452841e-01 1.38971448e+00 2.02225178e-01 3.60638112e-01 6.65510356e-01 7.40661919e-01 5.34368932e-01 6.80494189e-01 -1.10890317e+00 2.18454480e-01 4.18869168e-01 8.33771229e-01 1.35049188e+00 5.80696821e-01 -1.17704794e-01 -2.28846952e-01 -7.46510148e-01 3.46763283e-01 -5.90263829e-02 -2.26448059e-01 -2.34950021e-01 -1.01227045e+00 9.21684623e-01 -1.14853166e-01 6.31768644e-01 6.22296929e-02 2.65311867e-01 5.47521591e-01 5.27747571e-01 2.79266685e-01 1.13940334e+00 -8.08051825e-01 -4.23600495e-01 -7.96018720e-01 3.05357277e-01 1.21128547e+00 1.08262765e+00 9.84874249e-01 1.46938711e-02 1.45474583e-01 3.06280941e-01 6.73648000e-01 3.85747612e-01 3.30611438e-01 -9.17597115e-01 7.15411723e-01 1.30720508e+00 2.48022918e-02 -1.07496071e+00 -6.46482557e-02 -1.69333126e-02 8.74177590e-02 2.07519576e-01 1.42717108e-01 1.03232943e-01 -4.32884455e-01 1.52662575e+00 -6.02293983e-02 -5.18740356e-01 -1.09101832e-02 4.70473558e-01 1.93843767e-01 7.65156627e-01 -2.73644656e-01 -3.44936192e-01 1.35556436e+00 -6.02888465e-01 -4.26150382e-01 -4.79223162e-01 1.43082750e+00 -5.59166968e-01 1.42441952e+00 3.48749757e-01 -8.18956375e-01 -2.62166619e-01 -1.27326035e+00 -6.72762394e-02 -2.86726624e-01 7.75061846e-02 8.07729006e-01 8.12426507e-01 -1.13126850e+00 4.52350557e-01 -8.98588061e-01 -2.77884394e-01 -4.03740220e-02 2.61734813e-01 -2.92369306e-01 -3.00832421e-01 -6.72138453e-01 8.45611751e-01 3.31923395e-01 -2.03884333e-01 -7.90963292e-01 -6.53368890e-01 -1.27711737e+00 2.92219430e-01 6.57048643e-01 -6.40250683e-01 1.48180461e+00 -5.88808358e-01 -9.10689771e-01 6.53240383e-01 -5.64480245e-01 -4.25778180e-01 -2.89801955e-01 -1.03608802e-01 -1.24109956e-02 -2.81396598e-01 1.98971972e-01 1.70597598e-01 5.49005866e-01 -1.25274491e+00 -4.42235559e-01 -4.55064148e-01 4.55989003e-01 -7.27370083e-01 -4.72410530e-01 6.22866929e-01 -5.27977161e-02 -1.88619152e-01 -7.08634853e-02 -8.03718626e-01 -2.19174162e-01 -3.36464465e-01 -4.26181883e-01 -4.01857674e-01 4.91298020e-01 -7.71478832e-01 1.62302101e+00 -2.04535341e+00 4.84721780e-01 1.62764505e-01 4.84247506e-01 8.65184143e-02 -2.87358671e-01 6.78808331e-01 -2.10063025e-01 6.88362479e-01 -1.76903248e-01 3.05388253e-02 4.24880803e-01 1.54900968e-01 -5.07602870e-01 1.42988026e-01 4.20392185e-01 7.91555643e-01 -9.90245044e-01 -3.66229564e-01 -2.06769109e-01 -1.78011924e-01 -9.13431704e-01 3.78358476e-02 -1.01254857e+00 -4.24473770e-02 -7.01801836e-01 9.89163101e-01 3.51835817e-01 3.13292406e-02 3.31678689e-02 4.65659142e-01 -1.38019830e-01 4.03191745e-01 -7.15394557e-01 1.52490139e+00 -6.99760795e-01 7.40292907e-01 -2.27940947e-01 -5.02786398e-01 1.15267205e+00 2.18240902e-01 1.23293456e-02 -2.50899851e-01 -2.15753391e-01 -1.86267793e-02 -5.36120012e-02 -1.07244861e+00 3.36397827e-01 -1.03782713e-02 -6.65820718e-01 8.67859960e-01 -4.37282287e-02 -3.17319363e-01 4.58700538e-01 1.54377490e-01 1.55264413e+00 2.30969563e-01 5.25608301e-01 -1.85558602e-01 6.19535744e-01 6.07997715e-01 7.77422965e-01 6.66963339e-01 1.90029651e-01 1.62233084e-01 1.25007808e+00 -7.41739452e-01 -8.55867684e-01 -3.90494198e-01 4.41399157e-01 8.66353452e-01 -4.47980821e-01 -1.11919129e+00 -1.10462987e+00 -1.01398349e+00 -2.19654247e-01 1.12877285e+00 -3.38615984e-01 -3.62930149e-01 -7.34750211e-01 -4.65680122e-01 6.61011815e-01 4.16931272e-01 -2.21896172e-01 -1.05719972e+00 -4.61685181e-01 5.16676493e-02 1.77785382e-02 -5.50871253e-01 -2.70690352e-01 5.03874235e-02 -9.40276802e-01 -1.38350010e+00 7.76466280e-02 -9.81572092e-01 1.05481434e+00 4.04284112e-02 1.18811941e+00 6.39051378e-01 -3.30426216e-01 2.52800465e-01 -5.40461361e-01 -3.65193933e-01 -1.23025477e+00 -5.58117367e-02 -1.94380566e-01 -8.57734442e-01 7.14014113e-01 -3.35307628e-01 5.33139884e-01 -1.52875513e-01 -8.84963036e-01 -3.86455059e-01 3.95025969e-01 4.92086619e-01 6.50224239e-02 5.17612755e-01 6.11176267e-02 -1.09128201e+00 1.08501053e+00 -3.25975269e-01 -9.41024840e-01 4.94385511e-01 -5.55828989e-01 5.98963380e-01 8.26296806e-01 -2.01471806e-01 -8.65401745e-01 2.00919539e-01 -2.18139425e-01 -4.18335199e-02 -2.14449555e-01 1.12457335e+00 -3.35141361e-01 -2.09892124e-01 9.57904637e-01 1.38316393e-01 -3.39928687e-01 -4.50486213e-01 2.39418939e-01 3.14374655e-01 2.40306616e-01 -1.09546387e+00 1.13105512e+00 -2.98315048e-01 -2.32967213e-01 -4.80751306e-01 -3.87289345e-01 -2.01173529e-01 -5.25062203e-01 1.54511347e-01 5.78239202e-01 -3.26539069e-01 -3.89657378e-01 -1.57400817e-01 -1.56060374e+00 1.00278109e-01 -1.39369518e-01 2.83328928e-02 -4.59697187e-01 4.78914469e-01 -4.37082887e-01 -8.97369862e-01 -1.43510252e-01 -1.46313477e+00 1.10137415e+00 -1.90344956e-02 -7.64232159e-01 -8.88362348e-01 6.23800337e-01 3.18797618e-01 1.34888545e-01 4.98402119e-02 2.04547882e+00 -6.51479781e-01 -8.58144701e-01 -6.25594795e-01 1.08398177e-01 2.96825260e-01 3.85839164e-01 4.99225855e-01 -4.88373339e-01 5.64608127e-02 5.00527680e-01 -3.18654953e-03 2.33833268e-02 4.03850488e-02 7.86077261e-01 -4.99178439e-01 -3.52802485e-01 8.56730267e-02 1.73315644e+00 7.39809349e-02 6.16531432e-01 3.46157849e-01 7.18605101e-01 7.81379342e-01 7.28584528e-01 8.62454697e-02 4.22511965e-01 2.15092272e-01 3.59941006e-01 4.67594206e-01 1.60172522e-01 -4.39426214e-01 7.31067359e-01 8.26407909e-01 3.65751296e-01 1.41505122e-01 -1.49436796e+00 7.06223726e-01 -1.62677252e+00 -6.41080081e-01 -2.74445355e-01 1.97742116e+00 1.07387888e+00 2.34693587e-01 -8.42137635e-02 2.24230915e-01 3.27073187e-01 -1.94722936e-01 -7.05968589e-02 -7.33761489e-01 4.86747622e-01 -4.69527096e-02 4.56570834e-03 7.39228964e-01 -6.87169731e-01 8.50549579e-01 6.19195223e+00 2.91967452e-01 -8.80789459e-01 -1.08967580e-01 -9.91040990e-02 4.31427896e-01 -7.47991621e-01 5.70525765e-01 -8.68780196e-01 1.37572005e-01 1.17240810e+00 -5.39853811e-01 4.81647938e-01 1.24459600e+00 3.83237720e-01 -1.14309900e-01 -1.67338610e+00 3.08231413e-01 1.37670502e-01 -1.45300055e+00 2.96916068e-02 -5.99639043e-02 5.12841880e-01 -3.52929860e-01 -3.81965637e-01 4.85587806e-01 2.12497205e-01 -8.06178570e-01 5.18729627e-01 6.05049193e-01 -7.50029599e-03 -7.38944888e-01 8.99773598e-01 9.19153571e-01 -8.12737525e-01 -2.89644688e-01 -5.16993344e-01 -2.28039891e-01 -2.99625069e-01 7.66201556e-01 -1.23576391e+00 3.57576102e-01 2.62157977e-01 4.03549880e-01 -1.02972269e+00 9.59218800e-01 -5.06712914e-01 4.71977502e-01 1.83800533e-01 -3.94199044e-01 -1.61725163e-01 9.12021026e-02 5.77958822e-01 1.31832385e+00 2.31441975e-01 -2.61000395e-01 2.52810240e-01 1.42778277e+00 2.58408636e-01 -4.36220132e-02 -1.02698898e+00 -6.82585478e-01 1.99197724e-01 8.53774250e-01 -6.21689796e-01 -4.83798608e-02 -5.92871070e-01 2.86018103e-01 2.97707051e-01 1.37509495e-01 -5.95253289e-01 -3.04017812e-01 3.74730110e-01 1.87851936e-01 -9.98666286e-02 -6.28076792e-01 -5.96015990e-01 -1.28678465e+00 5.97776771e-01 -1.35226274e+00 2.14357097e-02 -8.89699996e-01 -8.95554841e-01 6.07719958e-01 2.29401663e-01 -1.00142229e+00 -5.84672272e-01 -5.61039388e-01 -7.78920889e-01 7.56296992e-01 -1.13698864e+00 -8.70095015e-01 1.74879551e-01 -2.49514505e-02 6.81150079e-01 -1.91577256e-01 1.14655447e+00 -4.29974832e-02 -5.59422374e-01 3.19462359e-01 -5.10778666e-01 3.60142477e-02 4.49566692e-01 -1.29383349e+00 5.82073092e-01 1.36011326e+00 1.36070326e-01 1.56068468e+00 8.98821890e-01 -9.20130849e-01 -2.22473097e+00 -1.07094133e+00 1.19216442e+00 -8.20210576e-01 9.80436862e-01 -3.41202497e-01 -1.07047272e+00 8.76232743e-01 1.94367357e-02 -4.40678954e-01 5.17744899e-01 2.10513398e-01 -6.01197660e-01 2.28605837e-01 -1.11060822e+00 6.18611097e-01 4.04574603e-01 -9.17355835e-01 -1.04667294e+00 6.94265306e-01 1.11123276e+00 -1.47393510e-01 -8.62971783e-01 2.20127013e-02 -5.27655073e-02 -5.50383747e-01 5.01715243e-01 -6.83683336e-01 7.86810815e-01 -7.03412116e-01 -1.90027758e-01 -1.13866317e+00 1.66916810e-02 -7.01921046e-01 -1.14744738e-01 1.30995047e+00 8.28347325e-01 -4.44271147e-01 7.98174620e-01 9.33878481e-01 -9.86508355e-02 -5.86183488e-01 -3.20474476e-01 -5.21792352e-01 2.71873400e-02 -7.43018389e-01 4.21310514e-01 8.22364271e-01 7.38476992e-01 7.28892535e-02 -2.81486381e-03 4.08627361e-01 3.06546718e-01 2.92974323e-01 7.46028662e-01 -1.29679656e+00 -5.76339543e-01 -4.01522256e-02 -4.32102084e-01 -5.33998966e-01 7.85185575e-01 -8.28458905e-01 4.77633327e-01 -1.37837064e+00 3.34894031e-01 -1.13770120e-01 4.68888462e-01 9.60070252e-01 -9.46981311e-02 -5.89850426e-01 -1.15174755e-01 2.74922233e-02 -2.57113636e-01 -7.51989111e-02 7.62779176e-01 -4.51919675e-01 -2.39079624e-01 1.95478573e-01 -8.30027759e-01 9.37009156e-01 5.35873055e-01 -1.07724059e+00 -4.84394193e-01 -6.54684126e-01 7.46857047e-01 4.34860319e-01 1.56312250e-02 -6.17892802e-01 3.87961894e-01 -3.63464653e-01 -3.77181262e-01 -7.44647682e-02 -3.52330148e-01 -6.91876709e-01 -1.52363643e-01 6.21599019e-01 -3.76795411e-01 2.22940683e-01 3.68273348e-01 2.83579707e-01 -3.76084685e-01 -1.19438469e+00 3.47046196e-01 -4.60955203e-01 -8.06726933e-01 -1.91944867e-01 -7.68352509e-01 -2.51098514e-01 8.58505905e-01 1.92136113e-02 -5.89331925e-01 1.25653252e-01 -4.72928792e-01 -3.82327884e-02 9.73632216e-01 6.60670519e-01 6.55379593e-01 -7.16787636e-01 -3.68085206e-01 4.43410426e-01 2.25997016e-01 -2.93971807e-01 -5.15817940e-01 4.77228463e-01 -7.77625859e-01 6.44755125e-01 -1.11045815e-01 -3.93806040e-01 -1.45972133e+00 7.30059743e-01 1.08237565e-01 -2.71182090e-01 -4.45771754e-01 5.91837585e-01 -1.91733748e-01 -6.11935019e-01 2.02912331e-01 -8.32675874e-01 4.60594893e-02 -5.62251508e-01 7.59216905e-01 3.00325546e-03 2.65096873e-01 2.87274048e-02 -5.74396670e-01 6.21526062e-01 -2.10692421e-01 1.28513604e-01 1.43951035e+00 3.11729312e-01 -9.86711919e-01 4.42307234e-01 8.65323067e-01 4.26673859e-01 -4.82331812e-01 6.32647350e-02 8.92452419e-01 -4.32334632e-01 -1.23131335e-01 -5.66103101e-01 -5.17381549e-01 8.08347523e-01 -5.97982593e-02 4.83682781e-01 8.73450518e-01 1.41662762e-01 3.50636214e-01 9.13902402e-01 9.14763927e-01 -3.41311455e-01 -1.19583018e-01 6.66877866e-01 6.93963885e-01 -9.05319393e-01 -9.36404392e-02 -7.15122640e-01 -1.49124950e-01 1.52606440e+00 6.86522067e-01 -9.63152796e-02 1.74465045e-01 9.32600141e-01 -2.63502687e-01 -4.77181196e-01 -8.46306562e-01 4.35323149e-01 7.30523700e-03 6.58181608e-01 6.38179541e-01 -1.88201606e-01 -1.77078798e-01 7.42204785e-01 -1.93622947e-01 2.59994388e-01 1.10686004e+00 1.56834435e+00 -6.59416616e-01 -1.61278737e+00 -6.43164396e-01 1.30207002e-01 -3.94011497e-01 -2.13639915e-01 -8.61650229e-01 7.43053555e-01 6.11533821e-02 9.90067184e-01 -6.23378634e-01 -4.05680984e-01 3.30678135e-01 3.32368672e-01 7.31479943e-01 -1.59736764e+00 -5.08192301e-01 -3.68646741e-01 3.09332043e-01 -6.95148706e-01 -2.05619503e-02 -4.06433731e-01 -1.33334994e+00 -1.30760849e-01 -4.54106212e-01 4.05924201e-01 7.61677563e-01 1.27634287e+00 1.88979387e-01 4.17705923e-01 3.51356834e-01 -3.96311104e-01 -5.78546166e-01 -5.12934148e-01 -1.36473358e-01 -3.44084650e-01 5.12296677e-01 -1.67297855e-01 -3.86947662e-01 5.17443776e-01]
[7.692233085632324, 7.700230598449707]
d8f765ef-2de3-490b-ab09-96f6acb4fb6d
an-experimental-study-of-the-effects-of
2007.15066
null
https://arxiv.org/abs/2007.15066v1
https://arxiv.org/pdf/2007.15066v1.pdf
An Experimental Study of The Effects of Position Bias on Emotion CauseExtraction
Emotion Cause Extraction (ECE) aims to identify emotion causes from a document after annotating the emotion keywords. Some baselines have been proposed to address this problem, such as rule-based, commonsense based and machine learning methods. We show, however, that a simple random selection approach toward ECE that does not require observing the text achieves similar performance compared to the baselines. We utilized only position information relative to the emotion cause to accomplish this goal. Since position information alone without observing the text resulted in higher F-measure, we therefore uncovered a bias in the ECE single genre Sina-news benchmark. Further analysis showed that an imbalance of emotional cause location exists in the benchmark, with a majority of cause clauses immediately preceding the central emotion clause. We examine the bias from a linguistic perspective, and show that high accuracy rate of current state-of-art deep learning models that utilize location information is only evident in datasets that contain such position biases. The accuracy drastically reduced when a dataset with balanced location distribution is introduced. We therefore conclude that it is the innate bias in this benchmark that caused high accuracy rate of these deep learning models in ECE. We hope that the case study in this paper presents both a cautionary lesson, as well as a template for further studies, in interpreting the superior fit of deep learning models without checking for bias.
['Jiayuan Ding', 'Mayank Kejriwal']
2020-07-16
null
null
null
null
['emotion-cause-extraction']
['natural-language-processing']
[ 1.19287163e-01 3.22331876e-01 -2.53290564e-01 -3.81651044e-01 -6.79640234e-01 -8.51976871e-01 8.95374060e-01 2.42183000e-01 -4.42165554e-01 8.41423273e-01 7.05886185e-01 -2.88205773e-01 -3.57823759e-01 -5.52429676e-01 -7.94531286e-01 -6.11408710e-01 5.76737560e-02 5.80769628e-02 -4.45726097e-01 -4.19791877e-01 5.29991150e-01 2.29372278e-01 -1.60928977e+00 7.09991753e-01 5.99421501e-01 1.00406659e+00 -1.75010294e-01 3.56871933e-01 -3.30414653e-01 1.46380258e+00 -1.08217299e+00 -6.06407344e-01 -7.42724240e-02 -4.72292274e-01 -1.10149968e+00 -2.99346685e-01 3.46198797e-01 -4.57766317e-02 2.65022498e-02 8.43744576e-01 7.21195877e-01 -1.12130873e-01 9.86906052e-01 -1.46688926e+00 -8.59525323e-01 1.13963008e+00 -4.41214204e-01 3.88129681e-01 5.14689624e-01 -1.86208054e-01 1.38061607e+00 -6.86074972e-01 8.93618643e-01 1.17183399e+00 8.04565251e-01 6.28462195e-01 -7.10853219e-01 -7.30183959e-01 3.82103860e-01 4.77384120e-01 -1.00860465e+00 -2.39681318e-01 1.07610703e+00 -2.93618560e-01 1.37866974e+00 3.97927642e-01 5.66077471e-01 1.77534747e+00 4.77617085e-01 9.01187837e-01 1.55221248e+00 -6.33410454e-01 2.55733579e-01 1.48595676e-01 3.47114652e-01 2.42801815e-01 1.32959470e-01 -3.91276889e-02 -7.19578445e-01 -1.72062919e-01 -8.52596164e-02 -5.09144187e-01 -3.48776937e-01 4.21646386e-01 -9.87953961e-01 9.93739963e-01 4.03111547e-01 6.41864836e-01 -6.22686803e-01 1.56321019e-01 8.14218044e-01 1.84272960e-01 5.59994757e-01 7.51754820e-01 -1.02919447e+00 -4.29663181e-01 -8.91499162e-01 5.21487534e-01 9.55915749e-01 6.68597221e-01 6.61193579e-02 5.60344346e-02 -3.00383389e-01 7.62830317e-01 -2.27376688e-02 1.70124233e-01 6.76963210e-01 -7.53190696e-01 3.04232419e-01 7.29294121e-01 -1.10413596e-01 -1.43080914e+00 -4.81373668e-01 -6.45927012e-01 -7.29136288e-01 -8.48384723e-02 3.30357075e-01 -5.17653227e-01 -7.26526618e-01 2.00195718e+00 8.87672678e-02 -3.42194289e-01 2.51360893e-01 9.44625139e-01 1.14681053e+00 4.56428379e-01 9.04941708e-02 -3.10387731e-01 1.49342310e+00 -5.04804790e-01 -1.33721209e+00 -2.88541526e-01 7.59210050e-01 -8.02514136e-01 1.30869210e+00 7.65456557e-01 -5.29584527e-01 -3.84786204e-02 -1.21774590e+00 -1.52422592e-01 -9.49765801e-01 8.56389329e-02 8.52257669e-01 4.20031011e-01 -5.92606544e-01 4.11084175e-01 -4.15184684e-02 -3.64818871e-01 3.08675647e-01 1.58208951e-01 -8.50179642e-02 3.68702829e-01 -1.75245821e+00 1.15044498e+00 4.21405107e-01 1.46912094e-02 -6.04948938e-01 -6.18761182e-01 -7.01437831e-01 -3.69530031e-03 4.95356470e-01 -1.11201443e-01 1.23322999e+00 -1.38807487e+00 -9.12147820e-01 1.13055122e+00 -2.20897287e-01 -5.28367460e-01 2.65175670e-01 -6.01317883e-01 -5.74845314e-01 -1.59433916e-01 8.35134014e-02 4.84816343e-01 5.83020151e-01 -1.50370336e+00 -5.80349207e-01 -2.35786468e-01 3.05349994e-02 -1.02221288e-01 -2.48303473e-01 3.76432151e-01 2.31940880e-01 -7.53924727e-01 -2.36572608e-01 -8.19338262e-01 3.57720077e-01 -7.48546839e-01 -5.82399070e-01 -8.19668949e-01 9.22307968e-01 -6.01902544e-01 1.59765959e+00 -1.82474613e+00 -2.00169474e-01 -7.43373483e-02 2.17335865e-01 -2.48107746e-01 2.28845030e-01 4.39154446e-01 -5.89187682e-01 7.38353372e-01 5.19846417e-02 6.93366230e-02 2.97832608e-01 1.78230777e-01 -7.20287561e-01 2.49026850e-01 2.87227094e-01 6.93031192e-01 -8.39132428e-01 -5.82327008e-01 -2.58594245e-01 3.58327329e-01 -3.99158061e-01 -8.76656547e-02 -1.92202687e-01 -2.60236472e-01 4.48394596e-04 6.41207516e-01 3.80399555e-01 -2.21593976e-02 6.55323267e-02 -4.92362320e-01 -1.39544755e-01 8.13112617e-01 -8.97777736e-01 1.07649946e+00 -3.91622037e-01 1.06790066e+00 -1.61993027e-01 -7.86329925e-01 7.83437788e-01 5.46960831e-01 2.25056782e-01 -6.77326500e-01 4.96197492e-01 3.38628918e-01 5.06943405e-01 -6.75237536e-01 4.71167088e-01 -4.22360778e-01 -4.28991467e-01 2.08136648e-01 3.33866477e-03 -1.36189181e-02 3.16972621e-02 2.65925020e-01 1.10901880e+00 7.32865930e-02 5.40060639e-01 -3.94570827e-01 1.79270566e-01 2.95355827e-01 8.27813029e-01 6.36789858e-01 -3.88338864e-01 4.28504914e-01 9.21267748e-01 -4.78641063e-01 -7.53424585e-01 -3.54695797e-01 -2.24020988e-01 1.06297207e+00 -2.72415042e-01 -5.37914038e-01 -7.14749813e-01 -1.00398362e+00 -2.48073906e-01 1.39445162e+00 -1.21484375e+00 -1.49795964e-01 -4.51506585e-01 -9.90487218e-01 7.77008593e-01 3.50197256e-01 5.13421237e-01 -1.33268833e+00 -9.70785201e-01 1.48026526e-01 -4.74938691e-01 -9.82427716e-01 2.21335575e-01 9.30735588e-01 -3.97662044e-01 -1.05538297e+00 -1.42734453e-01 -4.87032264e-01 9.13202763e-02 -4.98659641e-01 1.34150600e+00 -9.97330621e-02 -2.67136712e-02 1.60947785e-01 -6.51978016e-01 -1.10877204e+00 -3.45830858e-01 3.48866917e-02 -8.83806646e-02 -2.69603819e-01 1.12956500e+00 -2.29608521e-01 -4.40327227e-01 -3.26734602e-01 -7.50211596e-01 -1.52678043e-01 4.69836235e-01 7.16302991e-01 1.59387887e-01 2.00130507e-01 8.68599832e-01 -8.64578426e-01 9.81295884e-01 -8.03761661e-01 2.29794815e-01 9.04860161e-03 -7.25124717e-01 -4.65568118e-02 6.14083171e-01 -4.58942145e-01 -1.16712809e+00 -4.07398164e-01 -8.27537850e-02 -8.97403359e-02 -4.50620443e-01 6.68030143e-01 -2.06867889e-01 7.09128499e-01 7.51570344e-01 -2.75157504e-02 -4.04908806e-01 -2.84945250e-01 1.57340750e-01 8.26474905e-01 3.68448794e-01 -6.93388402e-01 1.72940537e-01 2.92616904e-01 -2.05754548e-01 -3.48073512e-01 -1.38651240e+00 -8.80214721e-02 -2.59987593e-01 -4.29551065e-01 8.84914875e-01 -6.85717881e-01 -6.01567209e-01 1.62179098e-01 -1.52581406e+00 -1.58683494e-01 5.21295257e-02 2.67101258e-01 -2.78951883e-01 -2.68252134e-01 -5.69535196e-01 -9.61856484e-01 -3.33103955e-01 -6.42031074e-01 1.00719655e+00 2.95150951e-02 -1.15541673e+00 -9.83146012e-01 -8.03254172e-02 1.86743379e-01 1.51241645e-01 8.11815321e-01 1.26750231e+00 -1.30593741e+00 2.37433180e-01 -5.82052991e-02 -7.87094906e-02 1.49100438e-01 3.05158496e-01 2.18496770e-01 -1.35943902e+00 4.03364480e-01 2.78632879e-01 -6.06513739e-01 6.93808854e-01 1.49800524e-01 1.08313179e+00 -7.00279474e-01 -1.16288766e-01 8.72534290e-02 1.61535454e+00 4.89800870e-01 4.40618992e-01 7.56641567e-01 6.31638765e-01 9.37979102e-01 6.69904232e-01 4.37748045e-01 3.18628609e-01 3.30202222e-01 5.71261346e-01 -1.51049048e-01 -1.72220498e-01 -5.74159101e-02 4.06117707e-01 5.66206515e-01 -5.45212300e-03 -4.08116877e-01 -1.09909225e+00 9.16570127e-01 -1.61500490e+00 -1.04791641e+00 -3.99303496e-01 1.49603546e+00 1.20044112e+00 1.79072142e-01 -1.10043995e-01 6.43244267e-01 4.82486486e-01 1.80309772e-01 -1.66725010e-01 -1.05367446e+00 -4.55994874e-01 -8.29619020e-02 2.12428227e-01 3.02107364e-01 -1.01613867e+00 8.33204031e-01 6.66158152e+00 9.87065911e-01 -1.26043940e+00 1.74219310e-01 7.13458061e-01 -3.38461637e-01 -4.72891897e-01 -1.99807674e-01 -6.37048721e-01 5.46277642e-01 8.01119566e-01 1.44105285e-01 9.06266123e-02 8.48509550e-01 3.52650911e-01 -1.57357380e-01 -1.21449065e+00 9.16288376e-01 3.32505286e-01 -1.03129721e+00 -1.16132312e-01 1.40609175e-01 6.37879431e-01 -1.36744469e-01 -1.12580016e-01 5.46687782e-01 1.54980212e-01 -1.12914002e+00 1.00837588e+00 1.57938808e-01 2.56046712e-01 -9.59111691e-01 1.11306703e+00 1.68942183e-01 -3.51446360e-01 -1.34235933e-01 -2.67047733e-02 -5.62956095e-01 -1.72578275e-01 8.02534044e-01 -8.26155782e-01 2.39686966e-01 9.88300681e-01 3.34046870e-01 -6.85697794e-01 3.16973209e-01 -4.26586568e-01 1.09595454e+00 -1.65162474e-01 -4.10531193e-01 3.98778796e-01 4.05692577e-01 7.00631797e-01 1.67876327e+00 1.35702297e-01 -7.10696951e-02 -4.03886974e-01 1.08076739e+00 -9.37359855e-02 2.86867261e-01 -8.03705394e-01 5.60946390e-02 3.73115033e-01 1.26395988e+00 -8.55507731e-01 -3.65422517e-01 -2.41016716e-01 8.27785075e-01 3.68895799e-01 2.72119999e-01 -1.14968145e+00 -3.35814416e-01 3.86467069e-01 -2.40573555e-01 1.43935919e-01 3.49379689e-01 -8.58558357e-01 -7.63071775e-01 1.93630084e-02 -1.18372655e+00 3.97359759e-01 -9.35941398e-01 -1.43725097e+00 5.81663191e-01 2.50038952e-01 -7.49249816e-01 -3.29297870e-01 -7.90725291e-01 -6.80476010e-01 3.52677464e-01 -1.27830088e+00 -9.53432620e-01 2.77277315e-04 3.48723829e-01 5.52748024e-01 -9.40998271e-02 8.23502421e-01 3.45032029e-02 -5.67285419e-01 5.70560753e-01 -5.28102756e-01 3.23923469e-01 9.00004864e-01 -1.63763249e+00 -5.20372331e-01 8.02671671e-01 1.29527807e-01 7.33542085e-01 1.27035546e+00 -7.93351889e-01 -1.08509266e+00 -6.12636030e-01 1.46208489e+00 -7.56921589e-01 6.16754651e-01 -3.01126570e-01 -6.55697763e-01 6.10056579e-01 1.00950789e+00 -3.78086030e-01 9.57156003e-01 5.76883078e-01 -4.52695966e-01 4.50823680e-02 -1.17488205e+00 6.83073819e-01 7.17591703e-01 -2.65796542e-01 -1.28233969e+00 1.72331512e-01 8.24779868e-01 -1.97816998e-01 -6.40052319e-01 4.49907631e-01 3.99271578e-01 -1.05525458e+00 3.28739643e-01 -7.23732769e-01 1.27878046e+00 -6.08627535e-02 -4.37531203e-01 -1.35257328e+00 -2.05710813e-01 -4.33203042e-01 -2.25047156e-01 1.97675407e+00 7.64286160e-01 -2.78907210e-01 2.73031116e-01 7.31629789e-01 -3.72970365e-02 -1.00494695e+00 -6.13523066e-01 -6.08523428e-01 3.72822016e-01 -8.70525956e-01 5.57319343e-01 1.42438102e+00 7.23284781e-02 6.47329271e-01 -1.35803744e-01 -1.33745238e-01 1.62825227e-01 3.26068811e-02 2.03618512e-01 -9.16345596e-01 1.01245075e-01 -6.34677112e-01 8.38636681e-02 -6.24699034e-02 3.97260547e-01 -8.20233941e-01 9.37576592e-02 -1.47008991e+00 3.75062376e-01 -1.26910470e-02 -4.65107709e-01 7.47560918e-01 -3.52411449e-01 1.79773569e-02 1.35602459e-01 -1.84691355e-01 -3.35298687e-01 3.76469374e-01 9.91044819e-01 8.84649679e-02 4.26699175e-03 -6.80014789e-01 -1.21172488e+00 1.16847992e+00 9.45522189e-01 -7.36579299e-01 -3.00198644e-01 -1.60742342e-01 9.44589376e-01 -7.70268083e-01 5.37153065e-01 -5.69678128e-01 -2.15217639e-02 -2.14288503e-01 6.70282602e-01 -5.05670369e-01 -1.52507544e-01 -9.30085123e-01 -3.54259163e-01 1.38076574e-01 -6.06395662e-01 1.87174365e-01 4.87144709e-01 8.52040723e-02 -4.21933979e-01 -3.19355398e-01 2.46697843e-01 -1.97347373e-01 -7.61773109e-01 -5.40780306e-01 -6.40373290e-01 3.78656328e-01 7.23873258e-01 -6.50789067e-02 -4.08133149e-01 -4.97308314e-01 -3.83101642e-01 -2.28501290e-01 -3.56227383e-02 6.49713099e-01 3.21728796e-01 -1.42544556e+00 -6.86734915e-01 -5.49183369e-01 6.51484728e-02 -2.84539551e-01 -6.32820129e-02 9.23814118e-01 -3.23825888e-03 4.88550991e-01 8.48370790e-02 -1.06100060e-01 -1.32885671e+00 6.81795299e-01 3.41725707e-01 -3.10235351e-01 -2.83778638e-01 7.97207415e-01 -5.72822541e-02 -3.23607445e-01 3.84267479e-01 -5.90157688e-01 -4.12363261e-01 7.27648854e-01 2.82061458e-01 2.41211802e-01 1.58864453e-01 -5.85358918e-01 -7.70592451e-01 2.77496248e-01 -2.32241694e-02 -2.21122861e-01 1.26756966e+00 -6.36596307e-02 -3.07874233e-01 8.47751856e-01 1.27292168e+00 3.32535326e-01 -5.82223654e-01 3.02667916e-01 2.81003565e-01 4.25782762e-02 3.94107014e-01 -1.62558794e+00 -8.19225550e-01 6.36819303e-01 3.79903197e-01 4.88328189e-01 1.23186839e+00 1.26663357e-01 4.50604439e-01 1.82831451e-01 -9.98785533e-03 -1.54195225e+00 -1.54997796e-01 7.45150208e-01 1.46604526e+00 -1.35166836e+00 -1.46959852e-02 -1.95851371e-01 -8.55680108e-01 1.09950793e+00 8.87148261e-01 -1.70897454e-01 4.08486754e-01 7.07868278e-01 3.20013314e-01 -7.69961953e-01 -1.15863919e+00 -2.03472093e-01 4.15728033e-01 2.88862467e-01 1.05920660e+00 2.78523508e-02 -8.10135484e-01 1.44914556e+00 -7.44865119e-01 -1.41282335e-01 4.81231242e-01 8.74957383e-01 2.89852656e-02 -6.69445395e-01 -5.49629748e-01 4.15161014e-01 -1.30945992e+00 -3.60574603e-01 -1.28862119e+00 1.31043530e+00 6.38509989e-01 1.16599226e+00 9.60010942e-03 -4.80173290e-01 8.88767466e-02 6.39779925e-01 2.99376965e-01 -2.92572230e-01 -9.60171998e-01 1.36901811e-01 5.47344923e-01 -5.22437274e-01 -6.58704162e-01 -5.87831914e-01 -1.48586333e+00 -1.59993768e-02 -2.03913406e-01 7.72629119e-03 6.65406168e-01 1.32937503e+00 2.45085463e-01 8.98080647e-01 2.77653456e-01 -3.68587554e-01 -2.14963362e-01 -1.22794688e+00 -3.71001482e-01 5.89719713e-01 3.83747131e-01 -6.81750774e-01 -8.98090839e-01 6.47902563e-02]
[12.621248245239258, 6.224817276000977]
ad7e6d5f-c83c-41bf-9ba6-f8a6003e87ee
code-prediction-by-feeding-trees-to
2003.13848
null
https://arxiv.org/abs/2003.13848v1
https://arxiv.org/pdf/2003.13848v1.pdf
Code Prediction by Feeding Trees to Transformers
In this paper, we describe how to leverage \emph{Transformer}, a recent neural architecture for learning from sequential data (such as text), for code completion. As in the realm of natural language processing, Transformers surpass the prediction accuracy achievable by RNNs; we provide an experimental confirmation of this over a Python dataset. Furthermore, we show that the way to obtain even better accuracy from Transformers is to expose the syntactic structure of code, which is easily recovered by parsing, to the neural network. This works significantly better than presenting the code as a linear token sequence, which is how Transformers were originally intended to be used. To accomplish this, we propose a novel enhancement to the self-attention mechanism of the Transformer. We enable the mechanism to learn weights---that is, how much to focus on each preceding token in the input---not only on the basis of a token's value, but also on the basis of the spatial relationships, as in their positions in the abstract syntax tree, between each pair of tokens. We provide comprehensive experimental evaluation of our proposal, along with alternative design choices, on a standard Python dataset, as well as on a Python corpus internal to Facebook.
['Satish Chandra', 'Jinman Zhao', 'Yuchi Tian', 'Seohyun Kim']
2020-03-30
null
null
null
null
['type-prediction', 'value-prediction']
['computer-code', 'computer-code']
[ 3.46316904e-01 4.12819475e-01 3.64056341e-02 -3.86541069e-01 -5.11878371e-01 -7.87186384e-01 3.13369185e-01 3.96238834e-01 -4.89144713e-01 2.39722759e-01 7.80264974e-01 -7.16789544e-01 2.17474416e-01 -7.99992144e-01 -1.08059919e+00 -1.78399757e-01 -2.02427860e-02 2.47795209e-02 -1.26722768e-01 -1.02661364e-01 3.02126139e-01 8.10557827e-02 -1.37198627e+00 5.93354881e-01 4.99685049e-01 5.62140822e-01 3.97700250e-01 5.44385791e-01 -3.82605135e-01 1.26364934e+00 -4.96801585e-01 -5.08219182e-01 8.62866864e-02 -1.60508603e-01 -1.23639214e+00 -3.00088465e-01 4.14817750e-01 -5.71259439e-01 -2.94973522e-01 7.28655696e-01 -3.68440934e-02 -5.97859062e-02 4.99262840e-01 -5.92427611e-01 -8.25222313e-01 1.32604110e+00 -1.40843242e-01 8.26970711e-02 4.33545381e-01 9.02744830e-02 1.62480664e+00 -8.35120916e-01 7.06550837e-01 7.89470971e-01 8.84289324e-01 5.96777201e-01 -1.18660307e+00 -3.51730257e-01 2.29559988e-01 -1.78143546e-01 -8.95848632e-01 -6.64216757e-01 6.79081142e-01 -6.27458274e-01 1.40884078e+00 1.66040007e-02 5.05028009e-01 9.98059630e-01 1.42636523e-01 9.77321267e-01 5.29403329e-01 -6.22659922e-01 6.49390370e-02 8.33226740e-03 4.95093107e-01 7.91292846e-01 -6.48394749e-02 -2.61470556e-01 -5.09448469e-01 3.47039066e-02 5.65811217e-01 6.08084165e-02 -1.31228685e-01 -2.21942052e-01 -1.03681755e+00 5.57098627e-01 5.56929410e-01 4.12540019e-01 -2.17729717e-01 3.94708157e-01 4.87922013e-01 2.16618195e-01 1.72285303e-01 4.19606060e-01 -6.97282135e-01 -4.85053837e-01 -7.87518144e-01 4.79175858e-02 9.03880537e-01 9.62539673e-01 7.22277403e-01 3.08251716e-02 -1.19634047e-01 4.51669604e-01 2.75042087e-01 5.20707034e-02 6.81877077e-01 -7.74074614e-01 8.20108712e-01 6.02326810e-01 -8.50003660e-02 -6.73811615e-01 -3.01781565e-01 -4.04775143e-01 -5.14258504e-01 -1.47979064e-02 7.63396144e-01 -2.07147196e-01 -4.89220530e-01 2.02343512e+00 -1.69161811e-01 -2.59581842e-02 9.82741266e-02 4.09298003e-01 5.76352239e-01 5.94170988e-01 2.31549650e-01 2.41411462e-01 1.41073930e+00 -7.64914453e-01 -2.74341494e-01 -4.19350684e-01 1.01752102e+00 -3.92799526e-01 1.32701266e+00 2.25966275e-01 -1.12636232e+00 -4.58442986e-01 -8.69917214e-01 -2.91382372e-01 -2.60597438e-01 2.31428891e-01 7.40421355e-01 4.20314908e-01 -1.13263357e+00 9.55092192e-01 -8.79890323e-01 -3.26673061e-01 3.54285955e-01 1.97704971e-01 -4.39389825e-01 2.53843576e-01 -9.17383790e-01 7.80439436e-01 1.88504204e-01 5.36090583e-02 -6.10599935e-01 -7.11535692e-01 -1.06953406e+00 6.25278950e-01 -1.48114702e-02 -3.19121569e-01 1.60399055e+00 -1.23699832e+00 -1.20339346e+00 8.08918536e-01 -5.08940279e-01 -6.71845734e-01 6.52150884e-02 -2.05686316e-01 1.30797669e-01 -1.41843602e-01 1.90560520e-01 5.46354175e-01 6.16226375e-01 -9.47812676e-01 -4.56061721e-01 -3.43720227e-01 4.19797987e-01 1.06624458e-02 -5.74943185e-01 1.57415092e-01 -3.80206198e-01 -4.71633524e-01 -1.07712261e-01 -8.86699438e-01 -1.89802021e-01 -4.90175821e-02 -4.61592406e-01 -4.01818246e-01 3.08783233e-01 -1.03628814e+00 1.32222629e+00 -2.47805238e+00 1.39153227e-01 1.48506254e-01 4.69923228e-01 1.39568105e-01 -6.76641837e-02 4.65543985e-01 -4.06746447e-01 2.80412614e-01 -4.18537349e-01 -6.59286022e-01 2.03506246e-01 1.70329645e-01 -6.64110661e-01 2.92576075e-01 2.94651598e-01 9.62543428e-01 -7.15039968e-01 -1.07863523e-01 -9.14916247e-02 2.33825654e-01 -8.48030329e-01 4.54629213e-02 -2.68046021e-01 -6.24863654e-02 -3.35672468e-01 2.96021640e-01 2.91849554e-01 -2.83786535e-01 4.84384447e-01 8.64617676e-02 -2.38438576e-01 9.66264427e-01 -7.15244234e-01 1.83829939e+00 -7.17093945e-01 8.64407778e-01 7.31834620e-02 -1.04793715e+00 5.58407128e-01 3.14322174e-01 2.04751536e-01 -7.28579164e-01 -2.03561932e-01 -6.59561008e-02 -6.74126744e-02 -5.70354044e-01 6.17230713e-01 4.49720770e-02 -3.62919927e-01 8.24573040e-01 -3.68763390e-03 3.87961537e-01 2.04184547e-01 4.93190467e-01 1.28580427e+00 4.95811939e-01 1.16321608e-01 -1.58441946e-01 1.82562634e-01 -1.20727055e-01 3.31158489e-01 8.23389769e-01 7.31517747e-02 3.75371575e-01 1.06844366e+00 -5.80780387e-01 -1.28992748e+00 -8.70582998e-01 5.60537912e-02 1.32011688e+00 -3.81132483e-01 -7.49645054e-01 -1.04274774e+00 -7.26315498e-01 3.11933532e-02 8.01651359e-01 -8.01087797e-01 2.86165322e-03 -8.81001234e-01 -3.45428437e-01 6.82312608e-01 8.53789270e-01 5.93222789e-02 -1.27478826e+00 -7.60690808e-01 3.20987821e-01 -1.44638345e-01 -9.82911527e-01 -4.78710175e-01 7.63057888e-01 -7.51500845e-01 -8.34237397e-01 -5.19752130e-02 -8.94068062e-01 7.60380507e-01 -7.94000477e-02 1.19855714e+00 3.97378623e-01 -6.78522885e-02 2.31425494e-01 -2.43139774e-01 -1.52418017e-01 -4.17862564e-01 3.00382376e-01 -3.06912243e-01 -2.23005921e-01 3.24013740e-01 -7.95040846e-01 -1.77146316e-01 -3.53833407e-01 -9.51120496e-01 1.57354861e-01 5.48227251e-01 6.22736514e-01 8.62586051e-02 -9.65400264e-02 1.01194128e-01 -1.10249782e+00 5.63849986e-01 -5.24600923e-01 -4.80965495e-01 8.67251381e-02 -2.70321548e-01 4.90015239e-01 9.91070867e-01 -1.48087487e-01 -7.91397631e-01 3.72032166e-01 -5.12516439e-01 2.48431079e-02 -1.29878789e-01 7.22415149e-01 1.15010090e-01 3.67478639e-01 5.37733257e-01 4.95721400e-01 6.56161308e-02 -6.75384104e-01 4.91897672e-01 6.14044547e-01 5.43119133e-01 -8.89434040e-01 6.78706586e-01 3.03406894e-01 -4.39021438e-01 -4.73914623e-01 -6.95276976e-01 -2.15898946e-01 -7.49050677e-01 3.19762945e-01 8.06633234e-01 -8.76180291e-01 -8.13260674e-01 2.44759038e-01 -1.51662505e+00 -6.33434355e-01 -4.02204573e-01 1.76990807e-01 -2.83610612e-01 3.37397307e-01 -9.93604004e-01 -4.47692811e-01 -1.76040307e-01 -1.01497579e+00 7.86561072e-01 3.30388807e-02 -4.41998959e-01 -1.07083702e+00 -1.46546751e-01 1.45735189e-01 5.01168728e-01 -2.30813473e-01 1.40782917e+00 -9.12557364e-01 -5.39966524e-01 1.07750343e-02 -2.25899428e-01 3.65974456e-01 -1.32953331e-01 1.63991332e-01 -9.39045906e-01 -1.75437964e-02 3.96578424e-02 -1.06752358e-01 8.31710458e-01 -3.69648659e-03 1.18372750e+00 -6.13374472e-01 -2.91524172e-01 5.22367835e-01 1.38846076e+00 -2.33452186e-01 5.25288761e-01 3.44864368e-01 6.87196314e-01 5.12272894e-01 -2.82417417e-01 5.06542444e-01 7.28673875e-01 4.61194992e-01 5.59327841e-01 6.24121279e-02 -2.04994395e-01 -6.43343985e-01 5.12097061e-01 8.92703831e-01 3.37187231e-01 4.98959944e-02 -1.03997552e+00 6.84138417e-01 -1.61311233e+00 -9.66617942e-01 -7.95284659e-02 2.05077815e+00 9.45150912e-01 2.79141068e-01 -7.00993165e-02 -3.09706130e-03 3.03165048e-01 1.04195029e-01 -2.37324461e-01 -7.25338340e-01 1.96369782e-01 4.02276903e-01 4.61844504e-01 6.16492271e-01 -9.16872263e-01 8.77676904e-01 6.69290733e+00 4.74911481e-01 -1.11321414e+00 9.21694636e-02 3.84312481e-01 3.15434970e-02 -5.55768788e-01 2.68767625e-01 -7.66609192e-01 5.43012500e-01 1.11599278e+00 2.99033280e-02 7.43341148e-01 9.63915050e-01 3.24368216e-02 -2.62236781e-02 -1.53433192e+00 4.37193274e-01 -1.37269989e-01 -1.46926761e+00 -1.24693751e-01 -7.49391913e-02 1.13230936e-01 6.59207404e-02 4.19944078e-02 5.68638742e-01 4.85987693e-01 -1.01240277e+00 1.08244228e+00 1.36185080e-01 5.21844745e-01 -4.21731174e-01 4.42152619e-01 5.33299744e-01 -1.21974421e+00 -3.49777669e-01 -2.74212122e-01 -5.82064390e-01 -3.31744671e-01 3.64583671e-01 -9.52608705e-01 3.06763202e-01 6.11915350e-01 8.95172894e-01 -6.94765747e-01 5.70634544e-01 -5.88496566e-01 6.89926326e-01 -1.18982829e-01 -2.05045149e-01 2.82811373e-01 1.79362908e-01 1.05086304e-01 1.46632409e+00 1.44952789e-01 -9.35655311e-02 -1.56934306e-01 1.05848360e+00 -4.04559076e-01 8.73258933e-02 -7.29578078e-01 -1.22903667e-01 5.22442818e-01 1.09847367e+00 -4.10165012e-01 -3.72873366e-01 -7.06886470e-01 7.74664819e-01 8.64424169e-01 3.90028715e-01 -6.71513319e-01 -6.03085458e-01 5.66115081e-01 1.69926077e-01 6.01722479e-01 -3.04309458e-01 -4.30845529e-01 -1.08992708e+00 4.16397333e-01 -8.25559914e-01 8.51056073e-03 -8.44448626e-01 -8.33892465e-01 5.01793265e-01 -2.33444557e-01 -8.86773348e-01 -2.92889446e-01 -6.45994246e-01 -7.31685996e-01 1.03988361e+00 -1.27801716e+00 -8.29620838e-01 2.31606066e-01 3.96043986e-01 4.47229981e-01 1.35554180e-01 8.93168569e-01 2.71939397e-01 -5.26480913e-01 5.77542067e-01 1.06468409e-01 9.02929723e-01 2.23031431e-01 -1.26329446e+00 8.33490729e-01 1.02440786e+00 3.45473826e-01 1.14198554e+00 4.51533228e-01 -3.04312468e-01 -1.52682173e+00 -7.81480014e-01 1.29475904e+00 -7.01343477e-01 9.26163137e-01 -7.06991494e-01 -8.58911455e-01 1.19573927e+00 3.85237753e-01 -8.49111229e-02 5.61927557e-01 3.80047023e-01 -7.14791358e-01 -4.49795872e-02 -7.08793938e-01 5.83776116e-01 8.15641284e-01 -8.52015138e-01 -8.60675573e-01 8.37371871e-02 9.88418162e-01 -4.45542127e-01 -5.90201199e-01 -8.27102736e-02 6.97837830e-01 -9.52588916e-01 6.99565589e-01 -7.38070607e-01 8.99629653e-01 -5.80016188e-02 -2.37608641e-01 -1.14709651e+00 -2.90320247e-01 -5.06585121e-01 3.26235965e-02 1.27835751e+00 6.66098833e-01 -3.42698574e-01 9.58139539e-01 8.02878499e-01 -3.53554994e-01 -7.23177075e-01 -9.07617986e-01 -2.81069905e-01 3.13055754e-01 -7.03216910e-01 6.25142574e-01 8.57590616e-01 4.48097229e-01 4.28919554e-01 -1.71367437e-01 5.37984148e-02 3.12008798e-01 1.38945421e-02 6.02588594e-01 -1.08928823e+00 -5.57765007e-01 -6.08058512e-01 -5.42296432e-02 -1.39072108e+00 5.09555578e-01 -1.15383220e+00 2.82150894e-01 -1.50672483e+00 3.09581637e-01 -4.79305655e-01 -1.68827370e-01 9.53277111e-01 2.06521135e-02 -1.37323141e-01 3.27961951e-01 2.68664241e-01 -4.22493130e-01 1.27578110e-01 7.05689847e-01 -8.49614516e-02 3.48041393e-02 -1.53852850e-01 -1.00324869e+00 7.54475772e-01 5.91779411e-01 -6.29663050e-01 -1.68605268e-01 -9.92844462e-01 5.24214149e-01 3.08176070e-01 2.41763711e-01 -8.04139912e-01 3.65773618e-01 2.59738356e-01 2.97468781e-01 -1.75648078e-01 4.62077260e-02 -8.69127095e-01 -2.66738057e-01 3.85495424e-01 -8.40194941e-01 2.34950915e-01 3.67386341e-01 2.97756582e-01 9.31768417e-02 -5.80717981e-01 2.80064881e-01 -3.49688888e-01 -5.38984954e-01 -9.96511951e-02 -6.85249925e-01 7.99036585e-03 4.76385742e-01 -8.81218687e-02 -4.22143102e-01 -1.98399469e-01 -6.91840231e-01 -2.29925010e-03 5.18631756e-01 2.12391615e-01 4.44549739e-01 -9.76040483e-01 -3.77639711e-01 4.22825158e-01 -9.57282484e-02 -6.01903759e-02 -1.18356034e-01 7.70995259e-01 -3.01969796e-01 3.90641958e-01 -7.30961785e-02 -2.71798491e-01 -1.06636655e+00 5.31494141e-01 3.35199952e-01 -4.15894687e-01 -6.72670245e-01 6.96220756e-01 1.09547362e-01 -5.55594921e-01 4.43184197e-01 -8.18969488e-01 -5.70336394e-02 -1.08994968e-01 3.87134582e-01 -4.48881745e-01 2.11675972e-01 -3.14761192e-01 -3.62633824e-01 3.83686244e-01 -1.85212389e-01 -4.82381321e-02 1.55630517e+00 2.81172432e-02 -3.77055258e-01 3.50125223e-01 1.18851316e+00 4.99502540e-01 -1.34274316e+00 -5.66625834e-01 1.02141209e-01 -1.28107041e-01 -3.67665201e-01 -6.88686430e-01 -8.01420689e-01 1.00364792e+00 -5.10309078e-02 4.66755450e-01 6.65193439e-01 1.53145716e-02 5.97705841e-01 6.12095952e-01 2.62115002e-02 -8.33831191e-01 2.16041282e-02 8.09166133e-01 3.49110037e-01 -8.33240330e-01 -3.02197129e-01 4.76987995e-02 -4.11196917e-01 1.27940631e+00 3.29225451e-01 -1.51970968e-01 4.95485157e-01 5.37123859e-01 -2.73146719e-01 4.24737819e-02 -9.49422479e-01 1.82193935e-01 -2.28522226e-01 4.94018614e-01 8.73100698e-01 -1.84943587e-01 2.33272225e-01 8.05233419e-01 -3.06142569e-01 -8.39289129e-02 7.40029156e-01 1.09613192e+00 -4.45986956e-01 -1.07206440e+00 -1.04210950e-01 4.64040011e-01 -6.49481237e-01 -6.69319630e-01 -2.12513313e-01 6.38365805e-01 4.56452668e-02 6.75952017e-01 3.26081425e-01 -2.97933221e-01 2.39521220e-01 2.48521551e-01 3.20495576e-01 -8.67405474e-01 -9.99803722e-01 -3.57125729e-01 1.03527695e-01 -4.92936671e-01 -5.75229637e-02 -6.95684075e-01 -1.54431355e+00 -3.75250131e-01 1.55964926e-01 1.27414420e-01 7.80169070e-01 1.17766428e+00 2.23490641e-01 5.41829050e-01 4.61621702e-01 -7.83444703e-01 -7.27378845e-01 -7.62690723e-01 -3.88247937e-01 4.72825021e-01 5.13437808e-01 5.90649918e-02 -3.45619589e-01 2.03767434e-01]
[9.947339057922363, 8.502829551696777]
1ebe8bf6-61e2-4124-949b-e896753ca80e
universal-lesion-detection-by-learning-from
2005.13753
null
https://arxiv.org/abs/2005.13753v1
https://arxiv.org/pdf/2005.13753v1.pdf
Universal Lesion Detection by Learning from Multiple Heterogeneously Labeled Datasets
Lesion detection is an important problem within medical imaging analysis. Most previous work focuses on detecting and segmenting a specialized category of lesions (e.g., lung nodules). However, in clinical practice, radiologists are responsible for finding all possible types of anomalies. The task of universal lesion detection (ULD) was proposed to address this challenge by detecting a large variety of lesions from the whole body. There are multiple heterogeneously labeled datasets with varying label completeness: DeepLesion, the largest dataset of 32,735 annotated lesions of various types, but with even more missing annotation instances; and several fully-labeled single-type lesion datasets, such as LUNA for lung nodules and LiTS for liver tumors. In this work, we propose a novel framework to leverage all these datasets together to improve the performance of ULD. First, we learn a multi-head multi-task lesion detector using all datasets and generate lesion proposals on DeepLesion. Second, missing annotations in DeepLesion are retrieved by a new method of embedding matching that exploits clinical prior knowledge. Last, we discover suspicious but unannotated lesions using knowledge transfer from single-type lesion detectors. In this way, reliable positive and negative regions are obtained from partially-labeled and unlabeled images, which are effectively utilized to train ULD. To assess the clinically realistic protocol of 3D volumetric ULD, we fully annotated 1071 CT sub-volumes in DeepLesion. Our method outperforms the current state-of-the-art approach by 29% in the metric of average sensitivity.
['Dakai Jin', 'Jinzheng Cai', 'Ke Yan', 'Le Lu', 'Jing Xiao', 'Adam P. Harrison']
2020-05-28
null
null
null
null
['medical-object-detection']
['computer-vision']
[ 3.91325280e-02 2.86964297e-01 -6.76008344e-01 -1.48897067e-01 -1.44368005e+00 -6.52730048e-01 3.66936594e-01 2.16415361e-01 -1.90466076e-01 3.96086633e-01 3.39036793e-01 -3.35201979e-01 1.05949841e-01 -6.42331481e-01 -6.63072586e-01 -7.54001439e-01 1.30437650e-02 7.41014004e-01 6.93629682e-01 4.38663065e-01 -4.95611906e-01 6.03146851e-01 -7.38208175e-01 6.60906792e-01 6.46362722e-01 9.78386939e-01 4.27620679e-01 5.56931019e-01 1.10723332e-01 6.86849177e-01 -8.33393037e-02 -3.70380938e-01 3.00460249e-01 -3.68760049e-01 -6.81865156e-01 1.94705606e-01 5.67765355e-01 -4.93754566e-01 -6.50342584e-01 9.61956859e-01 8.09031188e-01 -4.67705280e-01 1.09950411e+00 -8.49057436e-01 -2.81292588e-01 6.44965947e-01 -6.50989711e-01 4.03068513e-01 -6.54022098e-02 4.50429559e-01 1.06734228e+00 -1.10883403e+00 6.64280653e-01 8.02865565e-01 9.36364770e-01 6.38819754e-01 -8.51624012e-01 -5.35263896e-01 -3.90625209e-01 -2.06854083e-02 -1.35648644e+00 -5.49712591e-02 1.37677610e-01 -6.47877932e-01 4.74558890e-01 5.45392394e-01 3.14289212e-01 1.09646058e+00 4.46146995e-01 9.58724618e-01 7.51190960e-01 -1.18783921e-01 -1.82398483e-01 3.32400024e-01 -3.57663408e-02 1.23675931e+00 6.86493576e-01 -3.51414904e-02 8.08293670e-02 -4.44076300e-01 7.85083473e-01 3.85020077e-01 -3.97269636e-01 -6.90546811e-01 -1.66308439e+00 8.26918900e-01 7.86952674e-01 3.82444084e-01 -5.64850569e-01 -3.90200154e-03 7.37545133e-01 -4.31876630e-01 1.49065986e-01 2.75842041e-01 -1.53204352e-01 6.74259424e-01 -7.52456725e-01 -1.82374656e-01 8.02769661e-01 6.53010666e-01 1.98577866e-01 -3.55101496e-01 -8.32238972e-01 7.24045813e-01 3.70400518e-01 4.77667660e-01 9.75745976e-01 -2.69739211e-01 3.66330594e-01 8.84053767e-01 -2.64274329e-01 -4.14287418e-01 -8.53668034e-01 -7.17333734e-01 -1.04126501e+00 -1.75909758e-01 5.40897548e-01 -9.71514657e-02 -1.35209858e+00 1.29185534e+00 5.63860238e-01 1.76501587e-01 -8.20483044e-02 9.89385307e-01 1.34560168e+00 2.82863855e-01 2.97220320e-01 5.47715575e-02 1.86986244e+00 -1.19708073e+00 -3.24185580e-01 -1.86486557e-01 8.45453382e-01 -6.06894851e-01 8.39777648e-01 -1.93867907e-01 -8.57499182e-01 -2.77776748e-01 -7.06735551e-01 2.46409729e-01 -4.23128493e-02 7.24010289e-01 5.45019627e-01 7.02017725e-01 -6.31275356e-01 1.49669841e-01 -1.04188371e+00 -4.50150937e-01 7.46502638e-01 2.19731018e-01 -3.15512657e-01 -4.33325976e-01 -8.31739247e-01 1.12104774e+00 5.66251874e-01 -2.00006053e-01 -1.42908609e+00 -1.12100136e+00 -8.09474409e-01 7.34395683e-02 8.48411143e-01 -7.12884724e-01 1.41296756e+00 -2.48309523e-01 -8.36635053e-01 1.14138830e+00 5.91917560e-02 -4.41029519e-01 8.26404572e-01 3.60454828e-01 -3.51411641e-01 2.67337918e-01 3.95414948e-01 5.90394020e-01 5.77957690e-01 -9.48846161e-01 -4.63393390e-01 -2.29476541e-01 -4.05748457e-01 1.66196242e-01 -4.25507016e-02 -2.24606425e-01 -7.11155474e-01 -6.42253041e-01 1.22636996e-01 -1.13321280e+00 -4.98310864e-01 4.62887615e-01 -6.99020803e-01 -2.74848878e-01 6.88021779e-01 -8.45405102e-01 9.69451427e-01 -2.15509629e+00 -2.08327666e-01 2.19966635e-01 5.57432890e-01 2.00953737e-01 8.72018114e-02 -2.93762475e-01 -1.58374831e-01 9.58031043e-02 -2.88301200e-01 -7.58909795e-04 -1.15847223e-01 2.25171983e-01 2.44785994e-01 5.48602641e-01 1.36903554e-01 1.54202902e+00 -8.55205059e-01 -1.05964160e+00 3.87181371e-01 1.15936942e-01 -3.62138897e-01 4.17634100e-02 -3.11153922e-02 5.27375400e-01 -6.94058359e-01 1.06185114e+00 3.74331832e-01 -8.54319274e-01 1.46120533e-01 -5.79784453e-01 2.69783378e-01 -1.02128059e-01 -1.02612388e+00 1.67733502e+00 -5.60202837e-01 1.96685433e-01 -1.76755846e-01 -6.76640868e-01 3.71488243e-01 6.85848057e-01 9.70743418e-01 -3.05309802e-01 1.66056558e-01 6.00476742e-01 3.14111233e-01 -7.33450830e-01 -1.60646513e-01 1.54631129e-02 -5.65159647e-03 4.62459952e-01 2.98802778e-02 -1.79179922e-01 3.11974697e-02 1.65412784e-01 1.57069623e+00 -4.97545809e-01 7.73843944e-01 -1.07723735e-01 4.62317526e-01 1.83822542e-01 4.16552782e-01 9.31159437e-01 -6.52223706e-01 7.41347075e-01 2.78015316e-01 -3.07706267e-01 -9.20706391e-01 -1.29409778e+00 -6.54140294e-01 7.59694636e-01 -2.91753188e-02 -1.00511983e-01 -3.32891464e-01 -1.31258488e+00 3.00388992e-01 3.67909640e-01 -7.49698937e-01 -2.20401362e-02 -5.90731323e-01 -1.15101433e+00 6.80801272e-01 5.64509034e-01 3.71663034e-01 -1.12987912e+00 -8.19830418e-01 2.05152601e-01 -1.66451931e-01 -1.06958127e+00 -5.89793205e-01 2.68718719e-01 -7.25891113e-01 -1.58372855e+00 -1.21657407e+00 -7.94019163e-01 7.79303908e-01 1.02714881e-01 1.28129458e+00 -4.72880453e-02 -1.08897221e+00 5.15816748e-01 4.53210436e-02 -1.77304208e-01 -8.27327490e-01 1.04865000e-01 -3.38664263e-01 -6.23608707e-03 8.03067833e-02 1.31875455e-01 -7.83227921e-01 5.58233976e-01 -8.93542111e-01 -3.06385895e-03 1.35494435e+00 9.33008015e-01 8.70707273e-01 -2.48076320e-01 5.51828861e-01 -1.06675267e+00 2.88817048e-01 -8.22508574e-01 -1.35887578e-01 6.41720295e-01 -1.47307768e-01 6.32136595e-03 2.36700311e-01 -3.70766521e-01 -9.85259235e-01 5.06610513e-01 -3.28987204e-02 -5.05000532e-01 -2.67406344e-01 3.29090267e-01 3.24536771e-01 -2.26014659e-01 1.03075075e+00 1.97808534e-01 -2.47214496e-01 -1.58152312e-01 2.66347021e-01 5.10730267e-01 6.57343566e-01 -2.14312911e-01 6.16266847e-01 4.40326601e-01 2.44350612e-01 -4.69516039e-01 -1.00361967e+00 -9.42612827e-01 -6.93488419e-01 -1.18685767e-01 9.42398012e-01 -8.84732127e-01 -3.37681621e-01 4.13024724e-02 -7.78868437e-01 -1.76510081e-01 -6.21044695e-01 7.81002939e-01 -2.76616722e-01 2.28498578e-01 -8.11685801e-01 -1.72246724e-01 -5.87224662e-01 -1.45789897e+00 1.38170862e+00 -1.75071895e-04 -5.45663275e-02 -8.31324220e-01 -2.95581818e-02 -6.61195628e-03 3.78349036e-01 3.40362608e-01 8.05621386e-01 -1.02596927e+00 -7.07866490e-01 -4.07984257e-01 -5.86986840e-01 6.60122260e-02 3.50047886e-01 -3.10965061e-01 -6.58547640e-01 -3.65896553e-01 -3.06629688e-01 -5.05573988e-01 1.26983619e+00 7.84779072e-01 1.43165040e+00 2.22377643e-01 -8.83617878e-01 5.35690725e-01 1.29232717e+00 7.36014044e-04 1.04362173e-02 -5.80571480e-02 7.16411471e-01 1.72903761e-01 2.25141644e-01 3.27532291e-01 1.45016283e-01 4.40796971e-01 6.80999875e-01 -2.55207330e-01 -6.49109066e-01 -8.01282451e-02 -6.38014898e-02 4.63284045e-01 1.66324899e-01 -2.93862611e-01 -1.30096471e+00 7.95570612e-01 -1.57750833e+00 -5.79442561e-01 -1.60890520e-01 1.79775953e+00 8.04572582e-01 -1.64657162e-04 -1.58531219e-01 -3.55407447e-01 9.19588983e-01 2.43183356e-02 -9.45971787e-01 4.21293944e-01 1.92913145e-01 -5.47534786e-03 8.97689939e-01 -3.12238485e-02 -1.56710577e+00 4.57121640e-01 5.98973370e+00 8.62350225e-01 -1.01620436e+00 5.62276125e-01 6.03320479e-01 2.72789057e-02 -8.70262086e-02 -5.22485614e-01 -8.80401313e-01 3.34209353e-01 4.64157164e-01 -1.45062044e-01 -3.10145080e-01 9.52384651e-01 -2.09483355e-01 -1.65996805e-01 -1.30087078e+00 9.34898555e-01 3.21769983e-01 -1.37424779e+00 -1.24509826e-01 -3.70116159e-03 8.84438336e-01 4.88787442e-01 8.54864623e-03 5.13794899e-01 1.68248519e-01 -1.01283562e+00 1.09510623e-01 3.14552426e-01 1.16150737e+00 -1.01677351e-01 8.90742600e-01 3.78393531e-01 -1.25080037e+00 8.81407112e-02 -2.17861339e-01 1.04129827e+00 1.19619608e-01 6.59304798e-01 -1.70559585e+00 6.23973012e-01 4.21590358e-01 6.46610260e-01 -9.29929912e-01 1.61935878e+00 -2.02155754e-01 7.83234537e-01 -4.95151430e-01 2.80399531e-01 2.15837464e-01 5.50161958e-01 4.13610876e-01 1.39532340e+00 5.40186584e-01 2.05152437e-01 5.21133423e-01 7.58490205e-01 -4.16637689e-01 1.50038391e-01 -5.38862646e-01 3.10715497e-01 2.71216810e-01 1.70134377e+00 -8.75215828e-01 -5.30407250e-01 -5.53560317e-01 8.55974853e-01 -1.89465582e-01 -1.23163469e-01 -1.11812043e+00 2.98954219e-01 -1.57297328e-01 1.52756900e-01 9.15595442e-02 3.34393799e-01 7.92732760e-02 -1.24318588e+00 -1.60617560e-01 -4.56663966e-01 8.72299910e-01 -4.12528276e-01 -1.54260135e+00 4.77410883e-01 -2.36934945e-01 -1.35268700e+00 -2.18900993e-01 -7.20938921e-01 -6.82355464e-01 5.75381339e-01 -1.65528405e+00 -1.32011819e+00 -7.58841336e-01 5.79111218e-01 7.04197764e-01 -1.60980731e-01 7.93726802e-01 5.30682266e-01 -4.72974122e-01 7.29900479e-01 2.49211378e-02 3.10501158e-01 1.05487835e+00 -1.31031668e+00 -9.77086741e-03 4.81931597e-01 -2.67216051e-03 -1.38737917e-01 -1.19782545e-01 -7.72615492e-01 -1.20117402e+00 -1.72748363e+00 3.29550743e-01 -5.79260767e-01 6.06158137e-01 1.33702129e-01 -9.52784002e-01 1.00170791e+00 -2.57760584e-01 9.42503691e-01 6.13292992e-01 -7.08767056e-01 -4.84665185e-02 3.01610172e-01 -1.31592834e+00 2.76472837e-01 8.53500724e-01 -2.76835710e-01 -4.72273856e-01 8.89294326e-01 5.36126733e-01 -9.24326301e-01 -1.00974798e+00 7.81993449e-01 3.00056010e-01 -4.30444092e-01 1.22108757e+00 -4.57931966e-01 3.21853310e-01 -9.41670835e-02 -7.06387833e-02 -1.14215994e+00 -3.47036093e-01 2.28483319e-01 -1.98192433e-01 5.43233454e-01 5.31638086e-01 -5.47217309e-01 1.02862215e+00 2.91062713e-01 -5.69931448e-01 -8.42662871e-01 -9.47662473e-01 -5.57089388e-01 5.06083146e-02 -1.58990920e-01 1.11977145e-01 9.10525084e-01 -5.62638521e-01 -7.63814524e-02 1.74032860e-02 3.51471871e-01 7.43238330e-01 1.40957579e-01 3.15907568e-01 -1.07914436e+00 -3.42012316e-01 -4.04072106e-01 -2.69699574e-01 -7.43184030e-01 -1.00750998e-01 -1.60120499e+00 -8.67784396e-02 -1.70210731e+00 8.12402844e-01 -5.22793889e-01 -5.51589608e-01 6.13243282e-01 -2.16638863e-01 4.56504226e-01 -4.68383841e-02 4.17993695e-01 -6.74275339e-01 1.56048894e-01 1.67619550e+00 -4.12002504e-01 2.10383311e-01 2.58317143e-01 -2.93365419e-01 9.54931438e-01 5.35104394e-01 -5.92833698e-01 5.55097200e-02 -1.54136464e-01 -2.87367105e-01 3.59205693e-01 8.34180295e-01 -1.06988299e+00 3.20658147e-01 3.64278927e-02 6.08265698e-01 -9.50103641e-01 1.30186483e-01 -8.70625794e-01 -1.38759077e-01 9.63754594e-01 -1.94586903e-01 -4.34490860e-01 1.37511745e-01 6.63718641e-01 -1.36734754e-01 -4.30916399e-01 9.13920581e-01 -4.99895364e-01 -8.33052158e-01 7.11761355e-01 -5.33645749e-02 1.43845156e-01 1.69572055e+00 1.27916545e-01 -3.47732812e-01 2.07711115e-01 -1.07803333e+00 4.40116048e-01 -5.62808402e-02 3.29314083e-01 4.90203857e-01 -1.43181419e+00 -1.08828676e+00 9.48139280e-02 3.65223050e-01 3.27054501e-01 4.17566448e-01 1.17546594e+00 -4.89142030e-01 7.63053954e-01 3.30952555e-02 -1.04898489e+00 -1.05587685e+00 4.05611217e-01 7.71991253e-01 -7.55880594e-01 -7.24133372e-01 8.21472943e-01 6.08397007e-01 -6.01844668e-01 1.75447017e-01 -5.76467454e-01 -4.90445197e-02 1.46849137e-02 2.67396033e-01 3.18890586e-02 3.56848300e-01 -4.19701189e-01 -5.59193671e-01 4.51264292e-01 -2.25782171e-01 4.29191858e-01 9.86036241e-01 3.58989775e-01 2.05560520e-01 1.29318923e-01 1.12341523e+00 -6.66168556e-02 -9.59811151e-01 -6.40374303e-01 -9.34844464e-03 -1.83333829e-01 4.78026271e-02 -1.03379607e+00 -1.18013787e+00 7.11328030e-01 9.75315630e-01 -2.41608322e-01 7.58353174e-01 6.43755019e-01 7.90439487e-01 2.33728051e-01 1.39109373e-01 -2.59527981e-01 3.11617464e-01 4.89763953e-02 7.99780250e-01 -1.84270799e+00 1.89276841e-02 -6.12690091e-01 -7.91758716e-01 1.10340250e+00 7.69048393e-01 1.08594425e-01 5.46791792e-01 2.32440338e-01 -1.41882971e-01 -4.18497056e-01 -5.75114727e-01 -3.07383806e-01 6.82181656e-01 1.64086074e-01 3.79399300e-01 5.28042376e-01 -1.35808423e-01 6.67149067e-01 4.62541550e-01 -1.72138587e-01 3.11977893e-01 6.91203296e-01 -5.24742782e-01 -7.48422146e-01 -5.07516444e-01 1.09988844e+00 -5.91723561e-01 4.20584604e-02 -8.16083401e-02 1.05513024e+00 2.33767778e-01 1.24821991e-01 -2.29626894e-01 2.52986699e-01 4.16180909e-01 4.97286133e-02 3.14348459e-01 -1.01885498e+00 -5.63969314e-01 4.76944493e-03 -1.99761838e-01 -3.77025098e-01 -1.94338020e-02 -7.00791955e-01 -1.32946908e+00 4.23645556e-01 -4.04223889e-01 -1.82026342e-01 4.94879574e-01 6.10791981e-01 5.11332974e-03 9.64401960e-01 4.55809176e-01 -7.18507826e-01 -9.28681970e-01 -8.37021530e-01 -3.57347429e-01 5.06001592e-01 2.64595866e-01 -7.69089699e-01 -8.08148310e-02 -2.23860964e-01]
[15.234066009521484, -2.2379093170166016]
b79cf410-5b41-4cf7-ada9-d78e4e929393
ontoprotein-protein-pretraining-with-gene-1
2201.11147
null
https://arxiv.org/abs/2201.11147v6
https://arxiv.org/pdf/2201.11147v6.pdf
OntoProtein: Protein Pretraining With Gene Ontology Embedding
Self-supervised protein language models have proved their effectiveness in learning the proteins representations. With the increasing computational power, current protein language models pre-trained with millions of diverse sequences can advance the parameter scale from million-level to billion-level and achieve remarkable improvement. However, those prevailing approaches rarely consider incorporating knowledge graphs (KGs), which can provide rich structured knowledge facts for better protein representations. We argue that informative biology knowledge in KGs can enhance protein representation with external knowledge. In this work, we propose OntoProtein, the first general framework that makes use of structure in GO (Gene Ontology) into protein pre-training models. We construct a novel large-scale knowledge graph that consists of GO and its related proteins, and gene annotation texts or protein sequences describe all nodes in the graph. We propose novel contrastive learning with knowledge-aware negative sampling to jointly optimize the knowledge graph and protein embedding during pre-training. Experimental results show that OntoProtein can surpass state-of-the-art methods with pre-trained protein language models in TAPE benchmark and yield better performance compared with baselines in protein-protein interaction and protein function prediction. Code and datasets are available in https://github.com/zjunlp/OntoProtein.
['Huajun Chen', 'Qiang Zhang', 'Jiazhang Lian', 'Shumin Deng', 'Haosen Hong', 'Siyuan Cheng', 'Xiaozhuan Liang', 'Zhen Bi', 'Ningyu Zhang']
2022-01-23
ontoprotein-protein-pretraining-with-gene
https://openreview.net/forum?id=yfe1VMYAXa4
https://openreview.net/pdf?id=yfe1VMYAXa4
iclr-2022-4
['ontology-embedding', 'protein-function-prediction']
['knowledge-base', 'medical']
[ 1.86188549e-01 3.70409191e-01 -6.10384285e-01 -4.15678889e-01 -4.94392335e-01 -5.39413750e-01 8.48336071e-02 5.15549481e-01 -2.15471461e-01 1.30732012e+00 2.50191480e-01 -3.16285551e-01 1.14207894e-01 -7.47328222e-01 -1.37054598e+00 -9.21907246e-01 -3.21032964e-02 6.50043786e-01 2.34265372e-01 -1.96817562e-01 -1.60337746e-01 1.73744321e-01 -1.07985771e+00 3.85914057e-01 9.30485606e-01 4.48030949e-01 4.61468935e-01 4.16223526e-01 -5.42638481e-01 7.03373909e-01 -8.98392349e-02 -5.76422393e-01 -2.75133222e-01 -2.13431209e-01 -9.15213645e-01 -4.26429123e-01 -1.79345105e-02 3.19113851e-01 -4.06404793e-01 1.06367874e+00 6.19631529e-01 -1.09820165e-01 6.51742697e-01 -1.02065361e+00 -1.32835722e+00 7.78816700e-01 -5.07826746e-01 2.01633051e-01 2.94555962e-01 4.17040467e-01 1.37650359e+00 -1.01351440e+00 1.07889092e+00 1.29856241e+00 6.32536650e-01 6.15109086e-01 -1.38310015e+00 -4.40113872e-01 1.51750147e-01 4.16716695e-01 -1.40794218e+00 -4.72190194e-02 4.94998872e-01 -3.87556255e-01 1.91071796e+00 -1.36478588e-01 6.13885164e-01 1.16419768e+00 2.17584774e-01 7.22641051e-01 4.90455002e-01 -4.29271668e-01 -1.10186599e-01 -2.43879221e-02 5.90264201e-01 1.05971158e+00 4.19760853e-01 -3.34109634e-01 -7.91115105e-01 -7.55374491e-01 3.55453610e-01 1.17847845e-01 -5.25739551e-01 -8.27282786e-01 -1.07357097e+00 8.37412298e-01 4.72028494e-01 2.64361918e-01 -5.61800957e-01 1.13408275e-01 5.29434621e-01 2.84100473e-01 6.11673415e-01 5.18403769e-01 -1.12458432e+00 2.38322064e-01 -1.82370678e-01 1.03106909e-01 1.09760702e+00 1.01331317e+00 1.02222705e+00 -3.34279120e-01 -5.96843734e-02 8.90266180e-01 3.20690811e-01 2.05026180e-01 7.71657050e-01 -3.08319509e-01 2.51856834e-01 8.06391954e-01 -5.44361621e-02 -6.82522178e-01 -3.23225379e-01 -4.79404211e-01 -5.45029640e-01 -4.30519849e-01 1.70559332e-01 2.29031816e-01 -7.83281922e-01 1.67339551e+00 4.81860757e-01 3.11568975e-01 5.59427083e-01 6.12586260e-01 1.03072631e+00 7.62147605e-01 7.46402621e-01 -1.05930187e-01 1.58973670e+00 -1.19991457e+00 -7.26866841e-01 -4.49164882e-02 1.08250225e+00 -3.30911011e-01 1.23700106e+00 1.48685342e-02 -7.16289878e-01 -3.78867060e-01 -9.69394982e-01 -2.57412046e-01 -9.43425417e-01 8.12474713e-02 1.04442358e+00 2.09797457e-01 -9.37369108e-01 8.13653052e-01 -7.16539919e-01 -5.65288424e-01 6.15875721e-01 5.14108539e-01 -7.06146955e-01 -3.21070790e-01 -1.63201344e+00 1.02940595e+00 1.08325374e+00 -3.69612336e-01 -6.14605606e-01 -9.52460468e-01 -9.13748264e-01 2.21258909e-01 4.60704118e-01 -9.06273961e-01 7.53692389e-01 -5.63136220e-01 -1.20040381e+00 1.23289287e+00 -4.63461727e-01 -6.29793048e-01 -3.31057131e-01 -5.25543809e-01 -2.67517209e-01 1.98926985e-01 -4.39166501e-02 7.92955995e-01 1.67010710e-01 -9.20807719e-01 1.06337676e-02 -4.02866006e-01 5.13614863e-02 1.88851744e-01 -4.37113315e-01 -1.87028021e-01 -6.83749795e-01 -3.10236216e-01 -2.30225563e-01 -6.27792776e-01 -4.90338653e-01 -3.41201462e-02 -4.46948349e-01 -6.92036211e-01 3.29602331e-01 -6.32050931e-01 1.10002553e+00 -1.73319852e+00 3.61401558e-01 8.06547999e-02 4.37095344e-01 4.89778340e-01 -3.46665144e-01 7.71771014e-01 -3.46288472e-01 9.69574451e-02 -1.84652731e-01 1.79559156e-01 -4.88403626e-02 5.07099688e-01 -1.47338599e-01 3.46245915e-01 5.23388088e-01 1.50779521e+00 -1.14962745e+00 -4.58141863e-01 -4.70916033e-02 7.97271371e-01 -4.75379020e-01 1.30443335e-01 -7.90969670e-01 1.95788041e-01 -8.49728107e-01 8.50417495e-01 5.65222323e-01 -1.22478521e+00 6.44451916e-01 -4.05996591e-01 4.64080811e-01 2.40782782e-01 -3.53136897e-01 1.70395851e+00 2.37766594e-01 -1.29059300e-01 -3.80503476e-01 -1.29373109e+00 9.10579979e-01 2.39409879e-01 4.35996085e-01 -6.16708957e-02 -1.90756768e-01 1.77768558e-01 -8.75046775e-02 -6.31169140e-01 3.27183958e-03 6.68913871e-02 4.10751253e-01 -2.67542098e-02 6.52704358e-01 3.08794081e-01 2.32800618e-01 3.97623271e-01 1.23039854e+00 6.86893821e-01 8.48829448e-01 -2.71719664e-01 7.60519683e-01 2.38178939e-01 6.59934938e-01 3.33546430e-01 1.22641334e-02 -5.25671057e-02 6.09058321e-01 -5.13051748e-01 -9.52111125e-01 -8.10004890e-01 2.40506604e-02 1.51408958e+00 -7.17632398e-02 -6.94385111e-01 -6.09099507e-01 -9.12201583e-01 3.10716748e-01 3.70375574e-01 -4.74566638e-01 -4.14509952e-01 -4.42657918e-01 -1.30730128e+00 5.95370173e-01 4.10428554e-01 -3.45401615e-02 -1.19595087e+00 4.78627741e-01 4.24995512e-01 -9.45072398e-02 -8.70647192e-01 -4.04359400e-01 6.62075460e-01 -8.53298426e-01 -1.28580260e+00 -8.63701105e-01 -1.16434038e+00 8.20033491e-01 4.81447540e-02 1.49555147e+00 -7.36077577e-02 -4.26862359e-01 1.05867244e-01 -4.18857038e-01 -3.65575612e-01 -3.07985544e-01 1.28233671e-01 9.83621180e-02 -3.62419695e-01 1.19374466e+00 -7.78846800e-01 -4.70480531e-01 -1.19171008e-01 -8.70937109e-01 1.63114414e-01 7.60573804e-01 1.08294225e+00 1.16425610e+00 -5.20357847e-01 1.00830340e+00 -1.31593883e+00 5.09151459e-01 -8.92083943e-01 -3.61784339e-01 6.93811715e-01 -5.07592380e-01 5.76767802e-01 6.27568722e-01 -4.54846889e-01 -6.70785129e-01 4.18504268e-01 -2.88290262e-01 -2.68588245e-01 1.38191814e-02 8.86422455e-01 -3.14372033e-01 -2.17241660e-01 7.63325036e-01 6.12574577e-01 -6.33758605e-02 -6.82636559e-01 7.33670771e-01 3.89193892e-01 3.47595990e-01 -6.77263379e-01 3.70971650e-01 1.96611732e-01 -6.59390986e-02 -5.03825068e-01 -9.82183099e-01 -8.48104775e-01 -7.05795944e-01 4.87111598e-01 7.22420573e-01 -1.23199379e+00 -9.57194984e-01 4.01411541e-02 -1.06107330e+00 -6.55743256e-02 -9.85043272e-02 3.85706753e-01 -6.02065861e-01 7.11010158e-01 -8.53052914e-01 -3.38543445e-01 -7.11632788e-01 -8.20053637e-01 1.08067000e+00 5.86449318e-02 -5.43361902e-02 -1.14142120e+00 2.09288046e-01 3.53314757e-01 9.25696418e-02 6.72585890e-02 1.25624907e+00 -1.16605544e+00 -5.56905031e-01 -7.03829015e-03 -2.19996750e-01 2.28095576e-01 1.83252349e-01 -6.01870894e-01 -7.44173169e-01 -2.49844402e-01 -5.60495615e-01 -8.40246499e-01 1.18440735e+00 1.10235922e-01 1.11237502e+00 -3.12709719e-01 -9.41287935e-01 5.20561755e-01 1.54674494e+00 -2.10402608e-01 4.85929251e-01 1.84509248e-01 9.84579265e-01 5.82268894e-01 5.82792401e-01 1.45929158e-01 3.13469738e-01 5.42369425e-01 1.99300483e-01 -7.66837597e-02 6.34165630e-02 -5.99184275e-01 3.96317631e-01 1.03750646e+00 -6.51204139e-02 -5.43725014e-01 -9.09070432e-01 6.26267791e-01 -2.09146762e+00 -6.22541130e-01 4.21509473e-03 1.62505102e+00 1.58907998e+00 -1.82820395e-01 -3.03477257e-01 -5.95197380e-01 6.13690555e-01 -1.10536456e-01 -1.02214193e+00 1.15350522e-01 -5.32842517e-01 2.20649570e-01 6.12342477e-01 4.97308105e-01 -1.12166274e+00 1.52335286e+00 5.62112045e+00 1.15672195e+00 -6.87240124e-01 5.39473891e-02 5.01454055e-01 2.93946743e-01 -2.82180429e-01 -1.25459626e-01 -1.24220264e+00 3.02628696e-01 1.08323050e+00 -3.41509402e-01 6.10447600e-02 1.16407633e+00 -9.74774957e-02 5.18338203e-01 -1.13977218e+00 8.02386403e-01 4.98607475e-03 -1.70816696e+00 4.29065973e-01 -1.20888986e-01 5.73295712e-01 5.62289000e-01 -4.02323186e-01 4.36126024e-01 7.24302590e-01 -1.09995294e+00 -2.83706516e-01 7.99711585e-01 6.22777104e-01 -4.10153508e-01 9.11654711e-01 2.88872927e-01 -1.17606699e+00 5.67704856e-01 -9.78969216e-01 3.07890356e-01 8.12575445e-02 8.08045208e-01 -1.16302061e+00 7.70481348e-01 5.48123479e-01 1.10057306e+00 -4.07374054e-01 5.54243386e-01 -3.05401862e-01 7.74455607e-01 -1.66377649e-01 -2.96693414e-01 5.84011190e-02 -1.66458458e-01 1.86888173e-01 1.46486306e+00 -1.00604489e-01 3.46138209e-01 5.58150947e-01 9.01612401e-01 -4.75995153e-01 7.32911348e-01 -6.93829596e-01 -7.48262525e-01 2.46731713e-01 1.09642732e+00 -3.03508878e-01 -6.98912501e-01 -6.60999179e-01 1.07418430e+00 1.03394628e+00 5.13861179e-01 -5.74510574e-01 -2.37274989e-01 1.04368401e+00 -1.42043561e-01 3.38878512e-01 1.23450860e-01 4.51117843e-01 -1.22341168e+00 -1.40350208e-01 -7.78634250e-01 5.19759595e-01 -8.27409804e-01 -1.90656304e+00 3.20218444e-01 -4.38108981e-01 -6.45400226e-01 1.06998324e-01 -1.04014885e+00 1.57154873e-01 1.00264657e+00 -1.80819774e+00 -1.43919873e+00 3.57748032e-01 3.32367480e-01 2.73971558e-01 -3.62833053e-01 1.35706961e+00 2.08698884e-01 -4.02378112e-01 6.12205923e-01 5.37127316e-01 -1.18418008e-01 7.26020455e-01 -1.16834617e+00 6.87330961e-01 1.98168773e-02 4.96376120e-02 1.16597569e+00 5.89420080e-01 -1.07470417e+00 -1.61446202e+00 -1.32663417e+00 1.16863084e+00 -6.44044697e-01 7.78211772e-01 -3.89264345e-01 -1.53467202e+00 7.05409706e-01 -1.18038602e-01 2.55041689e-01 1.18785965e+00 1.43971309e-01 -6.19746923e-01 4.31557506e-01 -1.00073600e+00 3.98997813e-01 1.42499518e+00 -6.70420468e-01 -6.46377206e-01 8.71213198e-01 1.36755753e+00 -7.94299245e-02 -1.22007442e+00 5.92857003e-01 2.89130121e-01 -1.29772440e-01 1.36950421e+00 -1.43153048e+00 3.46447289e-01 -3.50238442e-01 -1.05808228e-02 -1.07740557e+00 -5.34994423e-01 -3.94874185e-01 -5.12292743e-01 8.80278111e-01 7.30076313e-01 -8.40903461e-01 9.48734701e-01 -6.47442564e-02 -1.74972683e-01 -1.18879855e+00 -4.58270907e-01 -6.09322369e-01 1.03762761e-01 1.73986793e-01 4.38026398e-01 1.25331402e+00 4.97514457e-01 6.56894326e-01 -3.89945835e-01 1.06182948e-01 3.78719866e-01 3.74437869e-02 5.66782713e-01 -1.26315999e+00 -6.59637690e-01 -3.59872170e-02 -5.45737028e-01 -1.18925261e+00 4.93149340e-01 -1.42264950e+00 -1.67936742e-01 -1.66591573e+00 8.02841544e-01 1.54273286e-01 -8.60550344e-01 9.35929835e-01 -7.83086896e-01 -8.21212605e-02 -5.72643042e-01 1.77137956e-01 -1.04417074e+00 6.09640360e-01 1.17452753e+00 -2.34619915e-01 1.29561320e-01 -6.14084184e-01 -8.97496223e-01 5.79041064e-01 8.37523460e-01 -4.03479487e-01 -4.06997561e-01 8.88319612e-02 3.33414465e-01 -3.14174473e-01 1.37379929e-01 -4.39712524e-01 1.73309892e-01 -6.18484095e-02 3.21472794e-01 -4.33802694e-01 3.78584534e-01 -4.78652000e-01 2.31167331e-01 5.77235043e-01 -5.26016295e-01 -3.35435659e-01 3.06991130e-01 1.19571602e+00 -1.49429709e-01 -8.25292766e-02 4.85058308e-01 -4.44421917e-01 -7.62431443e-01 5.08128583e-01 -9.41568986e-03 -4.86430041e-02 7.19318330e-01 1.37196124e-01 -4.50132728e-01 1.13361970e-01 -1.02096772e+00 4.25888717e-01 4.69411820e-01 3.64059597e-01 4.97349054e-01 -1.04144192e+00 -6.27615690e-01 -1.09264396e-01 5.92431664e-01 -5.15288413e-01 1.92858711e-01 5.02835572e-01 -4.96129692e-01 9.16247904e-01 7.79483002e-03 -3.29226911e-01 -1.50684607e+00 8.29826057e-01 6.13653138e-02 -7.33135819e-01 -5.88611484e-01 1.25960934e+00 7.33699918e-01 -8.05735648e-01 1.39101997e-01 -3.40602696e-02 -5.21557987e-01 -3.96161675e-01 5.14647007e-01 -3.21460128e-01 -1.12923816e-01 -4.85757113e-01 -4.07305390e-01 4.96808708e-01 -4.10657942e-01 7.97777057e-01 1.51351357e+00 2.31603771e-01 -5.40666103e-01 3.43165845e-01 1.33677483e+00 -3.34061325e-01 -9.49882329e-01 -8.13409805e-01 3.56780320e-01 8.56284201e-02 -5.47042549e-01 -1.02484858e+00 -5.79777479e-01 7.17840970e-01 3.97790939e-01 -5.94484031e-01 5.84899664e-01 4.75949585e-01 7.73649573e-01 1.06045043e+00 5.80412984e-01 -6.62410319e-01 -1.01941181e-02 7.13407516e-01 6.03296399e-01 -1.36639762e+00 -2.63546668e-02 -8.88850451e-01 -6.15044355e-01 8.43714237e-01 6.16218626e-01 7.96216652e-02 5.22482395e-01 1.09293535e-01 -1.20653145e-01 -7.10758448e-01 -1.09256852e+00 -3.76888961e-01 3.51172358e-01 7.66486049e-01 9.87345219e-01 1.71738669e-01 -6.00407958e-01 1.15006340e+00 1.67323485e-01 2.32733399e-01 -2.39371583e-01 8.97555590e-01 -7.58816779e-01 -1.49574924e+00 1.87581778e-01 5.51811099e-01 -7.36049473e-01 -6.71424091e-01 -7.01357305e-01 3.63765836e-01 4.14764434e-02 4.48498577e-01 -7.10234761e-01 -1.84578836e-01 4.80951481e-02 8.65086198e-01 3.71881813e-01 -9.53699827e-01 -2.79516548e-01 -2.11500734e-01 2.62613893e-01 -4.06341612e-01 -4.11220849e-01 -1.34124696e-01 -1.66118205e+00 -2.38896403e-02 -6.18817151e-01 5.87357938e-01 2.34843910e-01 6.15232706e-01 9.43231106e-01 6.31618261e-01 -2.88094491e-01 -4.41866517e-01 -4.53811437e-01 -1.12876916e+00 -5.14926910e-01 4.18867022e-01 -1.82649985e-01 -6.13491297e-01 -3.05287410e-02 5.41249573e-01]
[4.898274898529053, 5.759952545166016]
afe344f8-d255-47c8-91ac-a1c40975a592
merge-and-label-a-novel-neural-network
1907.00464
null
https://arxiv.org/abs/1907.00464v1
https://arxiv.org/pdf/1907.00464v1.pdf
Merge and Label: A novel neural network architecture for nested NER
Named entity recognition (NER) is one of the best studied tasks in natural language processing. However, most approaches are not capable of handling nested structures which are common in many applications. In this paper we introduce a novel neural network architecture that first merges tokens and/or entities into entities forming nested structures, and then labels each of them independently. Unlike previous work, our merge and label approach predicts real-valued instead of discrete segmentation structures, which allow it to combine word and nested entity embeddings while maintaining differentiability. %which smoothly groups entities into single vectors across multiple levels. We evaluate our approach using the ACE 2005 Corpus, where it achieves state-of-the-art F1 of 74.6, further improved with contextual embeddings (BERT) to 82.4, an overall improvement of close to 8 F1 points over previous approaches trained on the same data. Additionally we compare it against BiLSTM-CRFs, the dominant approach for flat NER structures, demonstrating that its ability to predict nested structures does not impact performance in simpler cases.
['Andreas Vlachos', 'Joseph Fisher']
2019-06-30
merge-and-label-a-novel-neural-network-1
https://aclanthology.org/P19-1585
https://aclanthology.org/P19-1585.pdf
acl-2019-7
['nested-named-entity-recognition', 'nested-mention-recognition']
['natural-language-processing', 'natural-language-processing']
[-1.61374232e-03 3.44248772e-01 -1.31584421e-01 -4.36598927e-01 -7.19527245e-01 -8.94225121e-01 6.47045910e-01 7.28008807e-01 -1.09951115e+00 7.20264018e-01 3.33047479e-01 -4.67373788e-01 1.53507248e-01 -8.45407426e-01 -6.09755099e-01 -3.25565249e-01 -2.15924814e-01 5.22974908e-01 3.45914781e-01 7.75091127e-02 -4.48461436e-03 6.27706528e-01 -1.24795687e+00 2.11792827e-01 8.99075866e-01 6.74686790e-01 -1.22101158e-02 6.74488008e-01 -6.05788708e-01 7.29846835e-01 -6.76494002e-01 -5.58679879e-01 4.72088493e-02 2.05610707e-01 -1.09228051e+00 -3.06454808e-01 5.93200684e-01 -1.09715663e-01 1.72086470e-02 7.01640368e-01 2.82656580e-01 2.41794869e-01 7.56428063e-01 -6.54801011e-01 -7.90479481e-01 9.00031209e-01 -3.39132875e-01 1.56876683e-01 3.21824178e-02 -4.50638622e-01 1.46846950e+00 -9.04798925e-01 7.55674660e-01 9.91704166e-01 1.05696607e+00 5.28917491e-01 -1.37601006e+00 -3.56122047e-01 1.42434582e-01 -2.43285790e-01 -1.30968106e+00 -2.16988295e-01 5.96088581e-02 -3.05635899e-01 1.45473111e+00 -1.73158385e-02 1.48852631e-01 7.28202045e-01 6.02843538e-02 8.59158218e-01 8.71646702e-01 -6.27478719e-01 5.97918108e-02 9.31667835e-02 6.19936526e-01 3.82580757e-01 4.43581581e-01 -1.61830917e-01 5.43047376e-02 2.58216411e-02 6.21367395e-01 -9.06202495e-02 1.72104165e-01 -9.26021039e-02 -1.33069837e+00 8.29809248e-01 6.62656546e-01 8.91946971e-01 -4.70070213e-01 1.39209121e-01 3.90240878e-01 2.80669499e-02 3.88211638e-01 7.13307023e-01 -8.53797197e-01 -2.22027704e-01 -1.24538577e+00 1.62564099e-01 9.87881184e-01 8.56470048e-01 7.39346027e-01 8.92987624e-02 -3.71695220e-01 8.20320427e-01 1.05511300e-01 1.23834006e-01 4.32975233e-01 -6.98190272e-01 4.85297322e-01 7.11554229e-01 1.73504800e-01 -5.62655509e-01 -6.53937042e-01 -4.29520220e-01 -6.48874462e-01 -6.68389350e-02 6.22478187e-01 -4.64803249e-01 -1.25927937e+00 1.68829274e+00 2.09336057e-01 3.32809836e-02 1.51688159e-01 3.37267578e-01 8.97974193e-01 8.50154221e-01 6.56227171e-01 1.94430515e-01 1.50625288e+00 -9.91217852e-01 -6.97795331e-01 -2.58715659e-01 9.32797611e-01 -6.88825011e-01 6.49251401e-01 2.06693798e-01 -9.67137635e-01 -6.37086034e-01 -7.77983844e-01 -4.09226328e-01 -1.00451577e+00 2.81228095e-01 4.57373023e-01 8.60632420e-01 -1.24867129e+00 7.44531333e-01 -8.96001756e-01 -4.22918737e-01 2.19818741e-01 6.32203579e-01 -6.01140440e-01 1.35864273e-01 -1.22335815e+00 1.19823086e+00 8.16468894e-01 7.00996146e-02 -4.50523943e-01 -7.43911088e-01 -1.12948430e+00 3.61716390e-01 2.34627217e-01 -2.03048423e-01 1.34122694e+00 -3.88665259e-01 -1.00418615e+00 8.70929360e-01 -3.96237493e-01 -6.66398466e-01 1.39105707e-01 -5.17495751e-01 -6.11866355e-01 -2.62745172e-01 1.81521624e-01 1.18757284e+00 5.47796935e-02 -1.19313622e+00 -8.72692227e-01 -3.23915407e-02 1.31304607e-01 1.50682312e-02 -5.28742254e-01 2.15249985e-01 -1.79520428e-01 -4.72531140e-01 -2.54986703e-01 -6.16710305e-01 -5.13710976e-01 -4.76428211e-01 -5.33489645e-01 -7.45263875e-01 3.21226329e-01 -6.63588107e-01 1.46499622e+00 -2.00209808e+00 -1.99771851e-01 -3.28711122e-02 2.53811002e-01 5.49649775e-01 -2.46413335e-01 5.53257525e-01 -2.15652585e-01 7.06785679e-01 -3.53498667e-01 -6.10186696e-01 3.51687253e-01 4.07330036e-01 1.77222572e-03 1.63369700e-01 8.41173232e-01 1.05619228e+00 -7.04748929e-01 -4.52891141e-01 9.51718315e-02 6.80083573e-01 -3.76275748e-01 1.90637037e-01 -4.38000299e-02 1.29150236e-02 -2.06688434e-01 2.78801471e-01 6.22108281e-01 -6.36169873e-03 3.29710662e-01 1.02636293e-01 -6.21209264e-01 6.14789903e-01 -1.35198498e+00 1.26594377e+00 -5.70304394e-01 5.77330112e-01 -4.04847190e-02 -8.88357162e-01 1.01484227e+00 5.21955192e-01 2.02597976e-01 -5.11376917e-01 4.67380770e-02 2.90429622e-01 -4.25386615e-02 -1.72526062e-01 8.69949758e-01 -1.55948192e-01 -4.32358801e-01 3.87164727e-02 4.47782904e-01 2.86622256e-01 4.99991536e-01 -4.32298109e-02 1.17182040e+00 -4.19870764e-02 5.07040024e-01 -3.13353777e-01 3.39740992e-01 -2.59636249e-02 7.30622351e-01 8.17379355e-01 -2.06859022e-01 5.73581398e-01 5.77783227e-01 -4.56025630e-01 -1.22004271e+00 -1.28250527e+00 -3.75929087e-01 1.45385158e+00 -2.05633953e-01 -3.78974885e-01 -8.06437969e-01 -8.64922345e-01 -1.64367333e-01 9.43519652e-01 -7.67514765e-01 4.84156638e-01 -9.86644864e-01 -6.24007761e-01 9.60334003e-01 8.55074644e-01 2.33457237e-01 -1.49336421e+00 -4.21759367e-01 6.19109988e-01 1.20057836e-02 -1.19794941e+00 -2.56316483e-01 7.37621129e-01 -8.18687320e-01 -7.17535675e-01 -7.99129546e-01 -1.29982722e+00 3.98387372e-01 -3.56796831e-01 1.45521498e+00 -8.35531875e-02 -6.36428893e-02 -1.50693534e-02 -5.13653934e-01 -3.52070272e-01 -3.57345521e-01 6.19578421e-01 -1.97380677e-01 -2.88320810e-01 6.84039712e-01 -2.63043672e-01 -1.48133174e-01 -7.39965495e-03 -1.12428045e+00 -4.45527673e-01 8.45452547e-01 9.25260305e-01 4.87765819e-01 -1.64354607e-01 5.44387758e-01 -1.22708011e+00 4.27927524e-01 -4.33098465e-01 -2.73707956e-01 2.89730042e-01 -5.66220105e-01 2.07436979e-01 7.54793704e-01 -2.99613297e-01 -1.14685750e+00 1.33001924e-01 -5.72797418e-01 1.53237402e-01 -9.99783635e-01 5.05412221e-01 -1.03286438e-01 3.74612480e-01 5.35211623e-01 -8.76939222e-02 -7.32013881e-01 -6.89480305e-01 6.85211301e-01 6.89497948e-01 5.05800784e-01 -5.43075979e-01 5.30186415e-01 1.72305331e-02 -3.67585748e-01 -8.91236067e-01 -9.05508220e-01 -8.39633048e-01 -1.10390425e+00 3.52206200e-01 1.23643279e+00 -7.44815767e-01 -3.75688493e-01 1.07036389e-01 -1.45731175e+00 -1.95202813e-01 -4.83939886e-01 4.12027508e-01 -1.99380964e-02 3.60250533e-01 -1.18107617e+00 -7.46079147e-01 -3.13373625e-01 -7.38007486e-01 9.95795667e-01 4.75596637e-01 -2.78924346e-01 -1.24050641e+00 2.84266979e-01 9.79842246e-02 3.81396770e-01 2.81796575e-01 8.27237964e-01 -1.46555734e+00 -9.12935212e-02 -2.21420005e-01 -3.77101094e-01 3.79915088e-01 -1.57704316e-02 -7.66011998e-02 -1.00337934e+00 -1.02591254e-01 -4.29532766e-01 -1.32454693e-01 9.53976870e-01 1.43562943e-01 5.18293560e-01 -1.43503621e-01 -4.22962904e-01 1.22804061e-01 1.54928780e+00 2.46915624e-01 7.80165672e-01 5.29110610e-01 8.14512551e-01 8.12172234e-01 2.48063758e-01 6.84258863e-02 4.13457453e-01 2.20085770e-01 2.10130870e-01 -4.61828440e-01 1.28286794e-01 -2.06161901e-01 3.50216329e-01 9.48080063e-01 1.42572328e-01 -4.88269866e-01 -1.20698118e+00 1.11265683e+00 -1.51038218e+00 -7.43488550e-01 -5.02198875e-01 1.83807290e+00 8.98976684e-01 2.70054281e-01 -2.21361555e-02 -1.07789181e-01 9.29668307e-01 1.05195999e-01 -1.83778524e-01 -9.14651036e-01 -4.64300439e-02 9.12605405e-01 7.24705100e-01 3.24235588e-01 -1.54368627e+00 1.22733068e+00 6.63799286e+00 8.14709246e-01 -7.54573107e-01 7.29557350e-02 5.24655223e-01 5.75481474e-01 -1.70870662e-01 -4.00382653e-02 -1.21617866e+00 2.26842016e-02 1.41252077e+00 1.76356107e-01 -1.69395223e-01 7.86189914e-01 -5.16613610e-02 1.74345583e-01 -1.08233106e+00 3.17430139e-01 -2.54546016e-01 -1.01520407e+00 -9.29420963e-02 1.07797340e-01 7.73479581e-01 2.86777735e-01 -2.85569966e-01 7.40908682e-01 7.18475461e-01 -1.30530131e+00 5.41760385e-01 1.91630319e-01 5.47538459e-01 -7.39135683e-01 1.13934064e+00 3.26460332e-01 -1.41843474e+00 1.32482558e-01 -3.54103565e-01 -1.72471479e-02 4.21064794e-01 5.30435741e-01 -9.96773839e-01 7.04345822e-01 6.15178704e-01 3.78151774e-01 -5.78886330e-01 1.22061718e+00 -2.27095142e-01 8.92306149e-01 -5.01708746e-01 -1.90942794e-01 6.40212953e-01 9.33599621e-02 1.77962333e-01 1.99094570e+00 1.03211120e-01 -2.87709087e-01 2.80074567e-01 8.17315459e-01 -2.65277803e-01 3.63470018e-01 -4.05674607e-01 -2.16918960e-01 5.21424711e-01 1.41476965e+00 -8.62964094e-01 -3.85446131e-01 -5.56572616e-01 8.08492601e-01 7.35076487e-01 1.43360391e-01 -7.60485947e-01 -9.59274352e-01 3.99066687e-01 -1.05124235e-01 9.41954851e-01 -5.21028697e-01 -3.27657759e-01 -1.01808178e+00 -6.65637255e-02 -3.85565251e-01 4.89935279e-01 -2.71023363e-01 -1.47210264e+00 8.94308686e-01 -1.56628668e-01 -8.20804536e-01 -1.92767471e-01 -9.46375489e-01 -6.31742120e-01 8.01531971e-01 -1.62816656e+00 -1.23223555e+00 2.47396439e-01 3.68116535e-02 3.90629351e-01 3.24199110e-01 1.19231439e+00 4.63357925e-01 -6.14101529e-01 6.78782403e-01 3.88494521e-01 8.06957662e-01 7.43356705e-01 -1.82108986e+00 8.85793269e-01 9.42410886e-01 6.35445833e-01 8.93326044e-01 3.54824334e-01 -5.32523572e-01 -4.35974240e-01 -1.11963153e+00 1.69183290e+00 -5.18667161e-01 8.55175436e-01 -6.39451742e-01 -1.06126702e+00 8.22519898e-01 6.46956742e-01 -1.87272891e-01 8.37029576e-01 4.06981528e-01 -5.72972953e-01 3.03083599e-01 -1.12026083e+00 4.13279027e-01 7.34797299e-01 -2.25737408e-01 -1.08412516e+00 -2.20567472e-02 9.63773012e-01 -1.85651407e-01 -1.28115094e+00 3.81436974e-01 2.35496446e-01 -8.25952053e-01 7.29351461e-01 -7.59386957e-01 2.48454571e-01 -2.44552687e-01 -1.27984270e-01 -1.09642434e+00 -4.44745839e-01 -1.39923513e-01 2.18035549e-01 1.82157230e+00 9.50617790e-01 -4.92039949e-01 7.81632423e-01 4.89413351e-01 -1.97386116e-01 -6.63638413e-01 -8.35030317e-01 -8.15337718e-01 6.46568596e-01 -4.63760763e-01 3.77320409e-01 1.12083256e+00 -2.08088830e-01 4.19680923e-01 -2.19576620e-02 1.34416729e-01 2.83931702e-01 -2.48321414e-01 1.51386961e-01 -1.32550883e+00 1.41490865e-02 -4.45725322e-01 -3.90453041e-01 -1.20255125e+00 2.88070321e-01 -8.42949390e-01 4.33651030e-01 -1.92374182e+00 -1.19308077e-01 -6.78104937e-01 -5.56543171e-01 8.20683897e-01 -5.56151681e-02 3.33239287e-01 1.85391143e-01 -9.22893509e-02 -7.48773873e-01 2.78388947e-01 6.12562597e-01 4.53412123e-02 -2.61072636e-01 -3.21925670e-01 -5.60451210e-01 6.11262918e-01 7.51461387e-01 -6.20275557e-01 2.97712356e-01 -6.18722498e-01 -6.06419034e-02 -3.69431406e-01 -3.02864779e-02 -1.18047059e+00 2.57335752e-01 1.46445557e-01 5.36104620e-01 -5.17045557e-01 -3.07669630e-03 -7.36527801e-01 -1.79895937e-01 1.59310684e-01 -4.17316079e-01 2.90978134e-01 3.93200696e-01 3.23708117e-01 -4.31306809e-01 -6.23726666e-01 5.19995868e-01 -1.01833440e-01 -9.31850076e-01 -3.79719451e-04 -6.85032904e-01 3.52895588e-01 8.41762900e-01 -1.18300706e-01 -1.43739492e-01 1.66772380e-01 -9.71103013e-01 3.27742279e-01 2.07028180e-01 4.31474447e-01 5.84968925e-02 -1.05614090e+00 -6.84366107e-01 -9.87997726e-02 -1.29097313e-01 2.41787031e-01 5.86226806e-02 4.17831570e-01 -7.52559364e-01 5.60468078e-01 1.60841586e-03 -4.40344542e-01 -1.08458567e+00 3.84213418e-01 1.69758961e-01 -9.86557484e-01 -3.78671944e-01 7.42626309e-01 -7.32797023e-04 -1.03502619e+00 1.28697619e-01 -5.36427855e-01 -7.30979025e-01 3.45906705e-01 3.05773914e-01 2.33163610e-01 2.45487660e-01 -8.50942433e-01 -2.96491534e-01 4.47003573e-01 -2.67741412e-01 -1.73130304e-01 1.55079091e+00 1.99347168e-01 -1.03821665e-01 5.97534776e-01 1.26510954e+00 1.88974947e-01 -7.73394108e-01 -1.27387345e-01 8.02984416e-01 2.60514468e-01 -2.88152516e-01 -7.62951732e-01 -5.12763441e-01 1.02903426e+00 4.75314826e-01 6.22342527e-01 7.02723503e-01 2.59158555e-02 9.87160981e-01 5.50541937e-01 3.82664129e-02 -9.73020136e-01 -4.40992445e-01 1.09271562e+00 1.38835341e-01 -9.52569902e-01 -3.33445758e-01 -3.47542375e-01 -5.70456386e-01 1.19073331e+00 5.13029695e-01 -3.92291754e-01 4.89841342e-01 4.57684815e-01 2.18368873e-01 5.92509992e-02 -4.90060955e-01 -4.80306268e-01 3.06789249e-01 4.89640266e-01 9.01930511e-01 1.81697935e-01 -4.02283758e-01 7.31749356e-01 -1.38274997e-01 -3.88085753e-01 4.42267001e-01 1.02092505e+00 -6.10189736e-01 -1.46075463e+00 -1.32614523e-01 4.67024088e-01 -1.10878551e+00 -4.24781621e-01 -2.24349052e-01 1.14018905e+00 4.51956362e-01 1.00717819e+00 4.52036411e-01 -1.77527547e-01 6.77779555e-01 4.32035089e-01 2.99124941e-02 -9.82147872e-01 -1.18396676e+00 -6.39663413e-02 3.68332475e-01 -1.48716792e-01 -4.69891667e-01 -6.44495964e-01 -1.57975459e+00 -5.62893935e-02 -5.48850119e-01 4.36492741e-01 5.84310949e-01 1.00112557e+00 3.11415315e-01 4.49826062e-01 2.07359895e-01 -6.66571319e-01 -3.23512644e-01 -1.03475463e+00 -5.35050392e-01 5.24604261e-01 2.33084723e-01 -3.91032606e-01 -2.50942200e-01 1.16068788e-01]
[9.800224304199219, 9.664308547973633]
954f763c-a458-40a1-abae-6c37dd790437
a-corpus-for-multilingual-document
1805.09821
null
http://arxiv.org/abs/1805.09821v1
http://arxiv.org/pdf/1805.09821v1.pdf
A Corpus for Multilingual Document Classification in Eight Languages
Cross-lingual document classification aims at training a document classifier on resources in one language and transferring it to a different language without any additional resources. Several approaches have been proposed in the literature and the current best practice is to evaluate them on a subset of the Reuters Corpus Volume 2. However, this subset covers only few languages (English, German, French and Spanish) and almost all published works focus on the the transfer between English and German. In addition, we have observed that the class prior distributions differ significantly between the languages. We argue that this complicates the evaluation of the multilinguality. In this paper, we propose a new subset of the Reuters corpus with balanced class priors for eight languages. By adding Italian, Russian, Japanese and Chinese, we cover languages which are very different with respect to syntax, morphology, etc. We provide strong baselines for all language transfer directions using multilingual word and sentence embeddings respectively. Our goal is to offer a freely available framework to evaluate cross-lingual document classification, and we hope to foster by these means, research in this important area.
['Xi-An Li', 'Holger Schwenk']
2018-05-24
a-corpus-for-multilingual-document-1
https://aclanthology.org/L18-1560
https://aclanthology.org/L18-1560.pdf
lrec-2018-5
['cross-lingual-document-classification']
['natural-language-processing']
[-1.13655843e-01 -2.98157543e-01 -4.01354551e-01 -3.93245399e-01 -7.91405976e-01 -1.00601876e+00 1.06475210e+00 4.29534405e-01 -9.18019652e-01 9.75984275e-01 3.36855710e-01 -3.96629602e-01 2.41388217e-01 -6.99033499e-01 -5.22867382e-01 -4.42372799e-01 1.92893878e-01 6.68231487e-01 1.20669022e-01 -3.88880372e-01 1.35647714e-01 3.41262370e-01 -1.24178886e+00 3.93457562e-01 1.04020679e+00 4.96735543e-01 2.70129383e-01 3.52068812e-01 -3.07987690e-01 4.00652319e-01 -4.89495128e-01 -8.60518277e-01 2.92316340e-02 -3.01347762e-01 -1.11448109e+00 -1.04474723e-01 4.50288355e-01 1.12088751e-02 4.84658889e-02 1.09363461e+00 5.24454951e-01 -1.15251638e-01 1.05066252e+00 -8.50815952e-01 -8.88048530e-01 9.68634188e-01 -2.84960657e-01 1.94892794e-01 2.71477461e-01 -3.77332568e-01 1.11472785e+00 -8.85328114e-01 1.01751947e+00 1.26061010e+00 6.32621646e-01 4.63280171e-01 -9.83420253e-01 -4.23340023e-01 3.71589094e-01 2.67563254e-01 -1.52566791e+00 -3.36712509e-01 7.10948765e-01 -5.44869840e-01 9.82568026e-01 6.19500801e-02 1.62613973e-01 1.48655927e+00 7.85673261e-02 9.15204287e-01 1.37057972e+00 -1.07153940e+00 -1.46811932e-01 8.87303174e-01 3.41118097e-01 2.77165055e-01 2.41309986e-01 -1.09654821e-01 -2.19055548e-01 2.27324054e-01 9.84516218e-02 -3.49208444e-01 -5.31761765e-01 -2.57682145e-01 -1.30284202e+00 9.55010712e-01 1.17396386e-02 9.46864307e-01 1.81640625e-01 -4.26438272e-01 8.29818249e-01 5.18655717e-01 7.00228274e-01 2.56905556e-01 -7.80141354e-01 -4.38009910e-02 -7.39829719e-01 -1.46933878e-03 8.73833597e-01 1.06714022e+00 7.43855536e-01 -2.54918098e-01 -1.73083405e-04 1.24805832e+00 2.46661454e-01 5.07391453e-01 8.45707893e-01 -3.09736460e-01 7.72875369e-01 2.77572364e-01 -1.23650871e-01 -7.10249960e-01 -1.44247517e-01 -3.24185461e-01 -6.10720217e-01 -1.92925349e-01 3.81468654e-01 -2.49397889e-01 -4.65581566e-01 1.68268800e+00 1.01512522e-01 -3.90831649e-01 4.39904839e-01 4.66022283e-01 7.57806659e-01 6.37565553e-01 1.04048900e-01 -4.30046171e-02 1.56519699e+00 -1.05864239e+00 -6.07825100e-01 -7.19376579e-02 1.06647277e+00 -1.18646348e+00 1.31563389e+00 2.65551567e-01 -7.75144041e-01 -6.24583721e-01 -1.07364690e+00 -2.29633555e-01 -9.65748131e-01 4.19109374e-01 4.66990024e-01 8.86695683e-01 -8.57834876e-01 3.38849515e-01 -6.25956237e-01 -7.40226209e-01 -9.29732919e-02 -1.99066587e-02 -6.39411986e-01 -4.30633456e-01 -1.54843152e+00 1.36672401e+00 6.83661580e-01 -4.36547160e-01 -3.78460348e-01 -4.14268732e-01 -8.89949262e-01 -2.08412006e-01 -1.41487122e-01 -2.43367478e-01 1.00500309e+00 -9.58672285e-01 -1.46749461e+00 1.29637754e+00 7.37540200e-02 -2.54291862e-01 5.67714632e-01 -4.79666190e-03 -6.01348341e-01 -1.97465375e-01 1.94736227e-01 5.88705719e-01 1.89414129e-01 -1.03774905e+00 -8.04203451e-01 -4.92104083e-01 1.12022720e-01 3.75281036e-01 -8.53352547e-01 4.17633325e-01 -5.93231559e-01 -7.92676210e-01 -4.44153994e-01 -1.02630711e+00 2.15126410e-01 -5.48887849e-01 -1.45655587e-01 -5.59344530e-01 4.52407807e-01 -7.29781687e-01 1.12683022e+00 -2.10045123e+00 1.51579857e-01 -1.63547099e-01 -6.30728245e-01 1.79698452e-01 -2.91625828e-01 6.37252986e-01 1.14016354e-01 3.54922593e-01 -1.81275174e-01 -5.37118137e-01 1.80155322e-01 3.68020654e-01 -2.74539024e-01 5.00074327e-01 9.07786414e-02 6.86087251e-01 -6.72519624e-01 -5.29807985e-01 2.33625751e-02 6.61638618e-01 -5.90673685e-01 8.08348414e-03 8.24774355e-02 3.81338835e-01 -2.16764614e-01 3.01579893e-01 7.04030812e-01 4.25073236e-01 3.99498671e-01 -1.29045904e-04 -4.61790502e-01 8.05282533e-01 -1.07391906e+00 1.91694558e+00 -7.73529589e-01 5.76331019e-01 -8.04049000e-02 -1.17082822e+00 7.40047216e-01 4.31096405e-01 1.24107614e-01 -5.52018821e-01 2.72052974e-01 5.14534831e-01 2.16395244e-01 -1.68484360e-01 6.15904272e-01 -1.79445863e-01 -3.01550031e-01 5.47512054e-01 4.75840747e-01 -3.30021918e-01 6.69938684e-01 -1.64292902e-01 5.49906611e-01 1.43030763e-01 3.04794312e-01 -6.63239896e-01 8.08860660e-01 -6.53528497e-02 1.54673994e-01 3.13152939e-01 -1.81398526e-01 4.23445225e-01 1.45074636e-01 -1.58449993e-01 -8.89518082e-01 -9.46566224e-01 -7.71403134e-01 1.34165311e+00 -2.30272114e-01 -4.65928346e-01 -7.50769675e-01 -9.51845288e-01 -7.60268643e-02 8.28059554e-01 -5.51189959e-01 1.36120752e-01 -6.00964785e-01 -9.23216999e-01 6.65206730e-01 3.77626479e-01 3.25426757e-01 -1.06436872e+00 8.89951140e-02 1.12284549e-01 -2.16927662e-01 -1.32671499e+00 -5.87326586e-01 4.28925037e-01 -6.26269460e-01 -9.59838212e-01 -9.01540935e-01 -1.28967774e+00 4.88889068e-01 -4.81992587e-02 1.34814239e+00 -2.74685413e-01 5.56694297e-03 4.58562493e-01 -6.81510568e-01 -5.18892169e-01 -6.75381124e-01 7.16923535e-01 1.81920543e-01 -1.95226714e-01 6.79470718e-01 -2.07874149e-01 2.25641429e-02 2.49090254e-01 -8.82763386e-01 -3.24560940e-01 3.00337940e-01 6.98309898e-01 4.57522660e-01 -1.86187234e-02 3.91582161e-01 -1.18351901e+00 7.46804893e-01 -4.16909695e-01 -4.30321902e-01 4.38545316e-01 -3.89093846e-01 1.19390585e-01 8.13601553e-01 -3.67234468e-01 -1.09999943e+00 -2.63043225e-01 -3.79268944e-01 3.70935321e-01 -5.20901442e-01 7.37268329e-01 -3.76032323e-01 1.76860586e-01 5.53414524e-01 3.49108487e-01 -5.41781962e-01 -7.48886764e-01 4.65925843e-01 1.18494725e+00 1.98008135e-01 -8.96114230e-01 4.99778181e-01 7.72437006e-02 -6.41790628e-01 -9.54848230e-01 -7.09059000e-01 -3.64298761e-01 -1.15257072e+00 2.11343795e-01 8.20628166e-01 -9.97180760e-01 -1.13735877e-01 3.43876600e-01 -1.46598732e+00 -1.34760246e-01 -1.38955474e-01 7.46129036e-01 -1.71100974e-01 4.68742222e-01 -6.62410200e-01 -4.05985177e-01 -1.71826407e-01 -1.18474448e+00 8.52701306e-01 -2.97737837e-01 -2.87575036e-01 -1.67048240e+00 4.58948731e-01 9.08716023e-02 3.99698883e-01 -2.60835350e-01 1.09785008e+00 -9.14615691e-01 9.37482417e-02 -1.35952964e-01 -2.09647492e-02 8.42306435e-01 3.73362154e-01 7.05215186e-02 -9.20718968e-01 -6.28824174e-01 -5.08017018e-02 -4.72466081e-01 8.08988690e-01 2.90304720e-02 6.47971749e-01 2.99684778e-02 -2.41773129e-01 4.61350441e-01 1.40147269e+00 1.93885535e-01 3.30552727e-01 5.11401892e-01 5.16007900e-01 9.21724498e-01 3.29662591e-01 -3.03157009e-02 6.60162508e-01 6.80689871e-01 -1.24739692e-01 8.69133323e-03 -2.52002418e-01 -4.12569977e-02 7.30520427e-01 1.63910365e+00 -3.04514579e-02 -5.33215284e-01 -1.06353354e+00 7.71603167e-01 -1.29385746e+00 -5.64157128e-01 4.22439836e-02 2.19955754e+00 1.29575658e+00 -3.88139784e-02 -2.68107653e-01 7.74569511e-02 5.86651862e-01 -1.43431975e-02 1.87002614e-01 -4.35104072e-01 -3.83605629e-01 2.59980708e-01 4.22740966e-01 8.04435492e-01 -1.31235218e+00 1.27431250e+00 6.26179504e+00 8.76406550e-01 -1.30764616e+00 2.99368560e-01 3.51343870e-01 4.04563934e-01 -3.51315588e-01 -1.25876382e-01 -1.34229493e+00 4.88344669e-01 1.06821823e+00 -2.77103603e-01 1.93843111e-01 7.57022977e-01 -3.83573890e-01 1.69900134e-01 -1.26862586e+00 7.29642749e-01 5.21418214e-01 -7.93418825e-01 1.79815680e-01 -1.81429580e-01 8.15412581e-01 5.05308211e-01 -1.45757469e-04 5.87302864e-01 4.45156544e-01 -8.42015028e-01 5.83115280e-01 -2.81634014e-02 9.07391489e-01 -9.04907584e-01 9.82020020e-01 3.56635958e-01 -8.96353245e-01 4.82786059e-01 -6.12219095e-01 1.30841672e-01 -1.19292989e-01 3.30654591e-01 -6.84132755e-01 7.10196316e-01 6.15076244e-01 8.22889686e-01 -7.77743340e-01 6.17244065e-01 -4.04501587e-01 6.02953136e-01 -2.16735572e-01 -1.05231732e-01 1.68915495e-01 -3.31318736e-01 1.94719344e-01 1.74762833e+00 5.06306767e-01 -7.06707060e-01 4.39826667e-01 2.75370389e-01 -2.58465081e-01 8.95898461e-01 -7.62116611e-01 1.14096999e-02 1.45264462e-01 1.01251638e+00 -6.72027588e-01 -3.72414827e-01 -8.58934760e-01 1.09418213e+00 5.92860222e-01 1.35720000e-01 -5.08914948e-01 -6.61660492e-01 5.08102000e-01 -1.37353605e-02 1.85964763e-01 -3.01632136e-01 -7.75214285e-04 -1.61081791e+00 -7.23974430e-04 -8.74936819e-01 4.14944530e-01 -1.97008178e-01 -1.70477092e+00 1.05469692e+00 7.51350075e-02 -1.11220098e+00 -3.47932070e-01 -1.16748440e+00 -1.04337297e-01 1.06796587e+00 -1.95767677e+00 -1.23244488e+00 2.23741755e-01 6.69441342e-01 5.70323050e-01 -6.14478469e-01 1.24445856e+00 6.99568272e-01 -3.31582844e-01 8.56307924e-01 5.10832667e-01 4.15266782e-01 1.25958490e+00 -1.24651635e+00 2.66835541e-01 5.64108253e-01 5.60830593e-01 6.62399471e-01 4.75307435e-01 -2.84649998e-01 -8.64899337e-01 -1.12039196e+00 1.56089354e+00 -6.85930610e-01 8.99444222e-01 -5.96258700e-01 -8.50510955e-01 8.86488676e-01 8.63179684e-01 -1.15537494e-02 9.90321577e-01 4.58502769e-01 -3.86922091e-01 5.92717417e-02 -8.74698400e-01 6.69661462e-01 6.61456823e-01 -6.19803727e-01 -7.36658394e-01 5.01163661e-01 4.33863401e-01 -2.70147957e-02 -8.81238103e-01 1.40887976e-01 2.85440117e-01 -6.62977815e-01 6.41304553e-01 -7.67597795e-01 4.20435280e-01 -8.48904625e-02 -4.58430767e-01 -1.84774852e+00 -1.51749104e-01 8.88374541e-03 5.05442321e-01 1.77196372e+00 5.15999377e-01 -8.19841862e-01 2.67571330e-01 -2.78564841e-02 -2.90605605e-01 -3.27147782e-01 -9.79891360e-01 -1.06801748e+00 1.16551399e+00 -6.29503846e-01 3.40102136e-01 1.28151560e+00 9.89677906e-02 6.33318305e-01 -1.79272279e-01 -1.50634497e-01 1.76777393e-01 5.05491644e-02 6.63007259e-01 -1.32209694e+00 -1.41697571e-01 -4.62739319e-01 -2.64953315e-01 -1.09274697e+00 6.54636204e-01 -1.61011827e+00 -1.00627951e-01 -1.39629328e+00 2.38094002e-01 -4.47777003e-01 -4.03667539e-01 3.39854985e-01 1.29535869e-02 4.54400092e-01 1.24936342e-01 3.72443311e-02 -3.95603567e-01 5.41951001e-01 1.08666313e+00 -3.14947367e-01 1.86465949e-01 -2.84801185e-01 -6.77664101e-01 8.15738201e-01 7.68170655e-01 -5.04093230e-01 -8.95379335e-02 -8.43886852e-01 -1.12021551e-01 -6.04966164e-01 -3.39796513e-01 -7.91584909e-01 -1.06237784e-01 -3.22434939e-02 2.22339645e-01 -2.98973322e-01 -3.16277891e-02 -8.78956437e-01 -5.77425659e-01 2.52319634e-01 -3.16643804e-01 1.39835566e-01 2.43296266e-01 5.41744977e-02 -6.47525549e-01 -6.98110759e-01 8.96131158e-01 -1.51373938e-01 -4.63922143e-01 1.48286372e-01 -5.68740904e-01 4.45756823e-01 8.85596275e-01 2.46101826e-01 -2.09355533e-01 -4.68873838e-03 -5.14213741e-01 1.33900354e-02 6.26062870e-01 7.92823613e-01 -1.73542678e-01 -1.52861249e+00 -1.02835310e+00 2.71271586e-01 3.79895836e-01 -5.37997007e-01 -5.24211414e-02 5.49973607e-01 -5.08043230e-01 8.92015934e-01 -3.42939168e-01 -3.58608842e-01 -1.35050774e+00 5.22602975e-01 6.22532815e-02 -6.38219655e-01 -5.71600124e-02 7.02567935e-01 -7.40622217e-03 -1.11015892e+00 2.27307007e-01 -3.23357344e-01 -5.81999660e-01 4.58809853e-01 3.82899582e-01 7.29372501e-02 3.95789385e-01 -1.03940105e+00 -4.96244520e-01 7.11723149e-01 -2.92004347e-01 -1.73373312e-01 1.21661806e+00 -1.20212704e-01 -2.99303651e-01 6.91574991e-01 1.63571620e+00 6.11682117e-01 -4.51130182e-01 -2.32004493e-01 2.68398106e-01 -1.24605916e-01 -1.46047309e-01 -7.32281268e-01 -7.82602906e-01 1.19301271e+00 5.70442021e-01 1.28348261e-01 8.42540026e-01 -1.73871685e-02 4.46029454e-01 4.38064903e-01 4.62587684e-01 -1.40149057e+00 -4.48236197e-01 1.20381403e+00 8.88393164e-01 -1.43567657e+00 -2.07039773e-01 -3.21339548e-01 -5.60120523e-01 1.11849415e+00 3.71773332e-01 -1.04687419e-02 7.96768486e-01 3.20650905e-01 3.61976594e-01 3.47306103e-01 -5.42433143e-01 -2.37992182e-01 3.87962252e-01 6.79916739e-01 1.39028335e+00 -4.35466692e-02 -8.74046087e-01 7.32752025e-01 -5.50646007e-01 -1.94995627e-01 3.10639709e-01 7.77876675e-01 -1.22839324e-01 -1.89914739e+00 -3.01522702e-01 3.40303481e-02 -7.80375540e-01 -4.10541922e-01 -4.14981633e-01 1.07138002e+00 3.77954245e-01 7.30529368e-01 1.01349019e-01 8.67883638e-02 3.34277809e-01 4.14750844e-01 6.87257886e-01 -9.34136808e-01 -4.87963259e-01 -7.48221651e-02 1.23774461e-01 2.62436748e-01 -4.64729339e-01 -7.35734463e-01 -7.34161794e-01 -1.12670749e-01 -1.33071572e-01 3.95882577e-01 7.86949754e-01 8.57057154e-01 -3.63865271e-02 4.71033961e-01 3.70493233e-01 -7.20615566e-01 -3.81033212e-01 -1.32053411e+00 -6.90014124e-01 5.31500876e-01 -8.06493387e-02 -5.67745686e-01 -3.58221203e-01 2.28503555e-01]
[10.943464279174805, 9.973381042480469]
a5d1bbfb-7fee-4d17-949c-cc7756d6d292
joint-task-self-supervised-learning-for
1909.11895
null
https://arxiv.org/abs/1909.11895v1
https://arxiv.org/pdf/1909.11895v1.pdf
Joint-task Self-supervised Learning for Temporal Correspondence
This paper proposes to learn reliable dense correspondence from videos in a self-supervised manner. Our learning process integrates two highly related tasks: tracking large image regions \emph{and} establishing fine-grained pixel-level associations between consecutive video frames. We exploit the synergy between both tasks through a shared inter-frame affinity matrix, which simultaneously models transitions between video frames at both the region- and pixel-levels. While region-level localization helps reduce ambiguities in fine-grained matching by narrowing down search regions; fine-grained matching provides bottom-up features to facilitate region-level localization. Our method outperforms the state-of-the-art self-supervised methods on a variety of visual correspondence tasks, including video-object and part-segmentation propagation, keypoint tracking, and object tracking. Our self-supervised method even surpasses the fully-supervised affinity feature representation obtained from a ResNet-18 pre-trained on the ImageNet.
['Ming-Hsuan Yang', 'Xiaolong Wang', 'Sifei Liu', 'Jan Kautz', 'Shalini De Mello', 'Xueting Li']
2019-09-26
joint-task-self-supervised-learning-for-1
http://papers.nips.cc/paper/8324-joint-task-self-supervised-learning-for-temporal-correspondence
http://papers.nips.cc/paper/8324-joint-task-self-supervised-learning-for-temporal-correspondence.pdf
neurips-2019-12
['unsupervised-video-object-segmentation']
['computer-vision']
[-5.69466352e-02 -9.82402563e-02 -5.92976928e-01 -4.94501889e-01 -9.67241764e-01 -5.54212570e-01 4.87332374e-01 3.25633258e-01 -5.55374861e-01 4.68428850e-01 2.78328449e-01 4.56979543e-01 2.49655228e-02 -7.02314377e-01 -1.18300116e+00 -4.14640963e-01 -9.39282849e-02 4.33474749e-01 9.82913315e-01 3.22179385e-02 1.43962383e-01 6.32000864e-01 -1.66639972e+00 4.21653986e-01 4.64578897e-01 1.11687052e+00 2.99345851e-01 5.30430138e-01 -1.68115884e-01 7.56647170e-01 7.92893395e-02 -1.61873549e-01 3.55705291e-01 -2.38532573e-01 -9.79299307e-01 3.15790474e-01 1.17419398e+00 -4.20980245e-01 -5.80914617e-01 1.30190277e+00 -6.49100319e-02 2.23151803e-01 4.64782238e-01 -1.16560423e+00 -5.54051876e-01 3.92015755e-01 -8.43740165e-01 6.00317359e-01 4.23884124e-01 6.80116266e-02 1.10260499e+00 -9.88260031e-01 9.53493834e-01 1.20149565e+00 9.57331538e-01 4.00933176e-01 -1.27375054e+00 -5.94202042e-01 3.82838696e-01 1.94391578e-01 -1.54080307e+00 -3.79123539e-01 6.74472034e-01 -6.69244826e-01 9.57771063e-01 -1.63405091e-01 8.56826246e-01 5.38079321e-01 1.81790814e-01 7.84565866e-01 9.26830411e-01 -2.02144369e-01 3.04626785e-02 -2.28722930e-01 -7.72842159e-03 1.20804334e+00 7.20613822e-02 4.27488267e-01 -8.09866190e-01 -9.22798663e-02 1.53135073e+00 2.97857583e-01 -1.75274402e-01 -8.34289730e-01 -1.50278974e+00 5.61668694e-01 7.03231514e-01 4.67304587e-01 -3.11805516e-01 5.33844948e-01 2.36524194e-01 1.31132349e-01 4.84538555e-01 3.24440271e-01 -5.04543900e-01 8.40550959e-02 -1.42435133e+00 -1.83786675e-02 3.64680737e-01 1.21331453e+00 1.42003155e+00 -7.76654333e-02 -3.81259024e-01 4.89245623e-01 3.47945780e-01 2.52007037e-01 3.45104188e-01 -1.08061862e+00 1.09964348e-01 6.31018400e-01 1.19836599e-01 -9.93500292e-01 -2.97106981e-01 -5.28901637e-01 -6.52933657e-01 2.29129970e-01 7.37325132e-01 3.46880347e-01 -1.01064861e+00 1.77245569e+00 4.93512511e-01 5.30008554e-01 -3.67760181e-01 1.03369784e+00 8.41749132e-01 4.94766593e-01 3.25232714e-01 -8.74427427e-03 1.24853313e+00 -1.47294855e+00 -3.97000670e-01 -1.25104472e-01 2.30701610e-01 -7.09118783e-01 6.72282577e-01 -2.52003938e-01 -1.32696104e+00 -9.83742952e-01 -6.43500209e-01 -1.65674940e-01 -2.50997126e-01 7.38075748e-02 6.37742281e-01 1.48846298e-01 -1.29048836e+00 5.61157048e-01 -8.52399588e-01 -3.53627145e-01 7.37733960e-01 4.42198217e-01 -8.00056994e-01 -4.10250574e-02 -8.88113260e-01 7.26797223e-01 2.24614874e-01 -6.60979822e-02 -9.54559267e-01 -1.10823298e+00 -1.08260012e+00 2.20435485e-02 2.76848823e-01 -9.65248048e-01 8.66766334e-01 -1.06120861e+00 -1.04020929e+00 1.39694405e+00 -3.21800500e-01 -5.05223870e-01 4.81486529e-01 -2.62250990e-01 -4.66402136e-02 4.29820567e-01 5.49687088e-01 1.34750998e+00 1.00741339e+00 -1.28971517e+00 -1.07436824e+00 -2.69287109e-01 -3.00959721e-02 6.51292428e-02 -1.58358030e-02 -2.15577465e-02 -1.08343065e+00 -9.66593802e-01 1.73764080e-01 -6.78744435e-01 -3.23688805e-01 5.82135975e-01 -9.64071825e-02 -3.51454586e-01 8.94966424e-01 -4.18747455e-01 9.18747246e-01 -1.97842848e+00 2.20388040e-01 1.54768899e-01 3.81663442e-01 2.71906219e-02 -2.42533550e-01 4.77003213e-03 -8.01586956e-02 -3.22190970e-01 -2.68142615e-02 -2.22999066e-01 -2.28889465e-01 -1.27935912e-02 -7.48228328e-03 7.34819412e-01 3.28593969e-01 1.33301818e+00 -1.17641413e+00 -9.29421902e-01 5.48213422e-01 4.59904373e-01 -6.06359243e-01 3.48431408e-01 -3.46312463e-01 4.86879557e-01 -5.33776700e-01 8.45295846e-01 4.88068610e-01 -5.60917974e-01 -2.27594599e-01 -7.10103631e-01 -1.17245950e-01 -3.09667856e-01 -1.10081220e+00 2.30171156e+00 -2.66491771e-01 7.11908460e-01 5.63649833e-02 -8.47280622e-01 6.08764112e-01 -1.17752172e-01 9.34992552e-01 -5.92123389e-01 -9.79172736e-02 -2.53291763e-02 -4.51571584e-01 -2.86611170e-01 4.37865198e-01 1.97694615e-01 5.37669286e-02 2.85574764e-01 4.55258787e-01 6.92283660e-02 1.84941143e-01 3.22922379e-01 1.02902234e+00 6.46642208e-01 2.60983974e-01 -5.61793149e-01 5.62068284e-01 1.27459690e-01 4.91375834e-01 8.13952446e-01 -3.46986383e-01 7.70849943e-01 -4.44163606e-02 -5.74748933e-01 -8.76558125e-01 -1.16626716e+00 -1.82631209e-01 1.29053652e+00 6.69305861e-01 -4.35903877e-01 -7.14734733e-01 -8.27261567e-01 3.25076491e-01 -2.52676576e-01 -7.77908027e-01 4.15789336e-02 -6.77715659e-01 -6.15314879e-02 2.54545480e-01 8.52413595e-01 5.96105099e-01 -9.58603084e-01 -4.54062939e-01 3.32263678e-01 -1.24691419e-01 -1.41239405e+00 -9.33042288e-01 2.77002811e-01 -9.07858491e-01 -1.26001370e+00 -8.27611387e-01 -1.10053146e+00 8.94605041e-01 4.09892380e-01 1.45467794e+00 1.76272094e-01 -4.59890783e-01 6.66294515e-01 -9.46929753e-02 4.07072544e-01 -1.88788682e-01 -1.18584514e-01 -4.47535962e-02 -4.40454483e-02 3.69032174e-01 -4.62362766e-01 -7.37508237e-01 6.35921597e-01 -7.43866444e-01 -8.67521763e-02 4.95264113e-01 8.53023708e-01 1.13487399e+00 -1.55847132e-01 2.10114241e-01 -6.37708187e-01 -3.40818197e-01 -9.43674669e-02 -8.33804131e-01 5.67151129e-01 -2.62601554e-01 1.36126965e-01 1.26511335e-01 -2.66138226e-01 -7.94620574e-01 5.48789680e-01 -7.34325126e-02 -8.01822603e-01 -3.15347880e-01 -1.92822188e-01 2.34423229e-03 -6.30475461e-01 3.58467877e-01 1.83436126e-01 -9.79473367e-02 -2.56854624e-01 7.20333874e-01 5.49587756e-02 9.75693047e-01 -6.02526367e-01 1.08964849e+00 6.85978234e-01 -1.82790682e-01 -4.19511855e-01 -1.06950808e+00 -9.64056134e-01 -1.14448392e+00 -4.83621269e-01 1.36795783e+00 -1.27593946e+00 -4.70386863e-01 3.59142184e-01 -1.08134544e+00 -3.77651691e-01 -6.68303967e-01 4.32456762e-01 -8.21841061e-01 3.17040205e-01 -7.75833607e-01 -1.28086731e-01 -3.23322639e-02 -1.04832804e+00 1.70151579e+00 3.47768992e-01 -1.90683976e-01 -1.21323729e+00 5.30059673e-02 2.83446193e-01 1.47372842e-01 1.88851625e-01 4.46011484e-01 -1.39599994e-01 -1.12614536e+00 4.52676881e-03 -5.16745567e-01 6.68253526e-02 1.79042622e-01 -6.52797818e-02 -8.30250859e-01 -3.06928933e-01 -5.75137198e-01 -3.76523256e-01 1.10414958e+00 6.86989367e-01 1.23390257e+00 6.60943761e-02 -6.41722679e-01 9.66176450e-01 1.38336873e+00 -3.32379043e-01 4.09995019e-01 3.98653239e-01 1.10822999e+00 3.88556689e-01 8.52644622e-01 1.14573538e-01 4.21476811e-01 9.24568117e-01 3.55090261e-01 -5.92247486e-01 -6.09837711e-01 -4.31625724e-01 6.57097623e-02 2.77382731e-01 -9.16279629e-02 2.80501127e-01 -6.05213463e-01 5.47129035e-01 -2.09585309e+00 -1.07864225e+00 -5.61173148e-02 1.95226133e+00 9.83697176e-01 2.19791029e-02 3.12178612e-01 -4.99213099e-01 9.53984976e-01 2.50734478e-01 -4.20080721e-01 4.90725338e-01 -2.15262026e-01 1.03916623e-01 7.94739723e-01 6.06237292e-01 -1.59867251e+00 1.40507615e+00 6.57582331e+00 8.66014957e-01 -8.26041520e-01 2.00737715e-01 6.91586971e-01 2.51736194e-01 -1.87070280e-01 -4.95818555e-02 -1.01011527e+00 2.07877874e-01 1.94146141e-01 3.12406093e-01 2.49027591e-02 8.87636065e-01 -7.40490034e-02 -1.35043442e-01 -1.52358234e+00 1.13226306e+00 8.50522071e-02 -1.96989202e+00 -7.15059415e-02 -1.95090488e-01 1.04818225e+00 1.83084011e-01 -2.25432873e-01 -7.57383406e-02 3.04560512e-01 -7.20881224e-01 9.20516312e-01 5.57667851e-01 9.16456223e-01 -3.94391686e-01 3.78761083e-01 -1.23926409e-01 -1.75900233e+00 1.67365283e-01 -2.69354641e-01 3.83284181e-01 2.73438692e-01 5.27417541e-01 -1.85536087e-01 3.35071534e-01 1.09234333e+00 1.18865883e+00 -5.92257798e-01 1.12175596e+00 1.02245726e-01 1.44459769e-01 -2.16455787e-01 6.14652336e-01 4.38738018e-01 1.47938393e-02 3.53682220e-01 1.51642621e+00 -1.39539957e-01 3.17912921e-02 4.94071633e-01 9.62153792e-01 -6.19721971e-02 -2.30905980e-01 -3.02115947e-01 3.04652065e-01 3.69761348e-01 1.35458481e+00 -1.00812221e+00 -4.96725380e-01 -4.60470051e-01 1.11235559e+00 3.91557753e-01 3.32231730e-01 -7.99362481e-01 1.19936287e-01 7.11321712e-01 3.70106936e-01 6.10891640e-01 -2.70636857e-01 4.03916165e-02 -1.18994570e+00 -2.24082932e-01 -4.65899229e-01 4.26107705e-01 -7.69351125e-01 -1.39354336e+00 6.08165383e-01 7.95918778e-02 -1.31060982e+00 -2.66002655e-01 -4.29952055e-01 -3.58923137e-01 5.06255567e-01 -1.72580874e+00 -1.29575932e+00 -5.48213363e-01 9.69958842e-01 7.12947667e-01 2.57414617e-02 5.95409095e-01 4.53400403e-01 -2.33405769e-01 5.93865931e-01 -1.28751069e-01 4.48450148e-01 8.10117006e-01 -1.12049603e+00 3.89182806e-01 8.86033833e-01 4.91394430e-01 5.21569490e-01 2.34480515e-01 -6.90893531e-01 -1.25680923e+00 -1.42165399e+00 6.74492300e-01 -6.40254915e-01 6.74567878e-01 -3.31919789e-01 -9.10693288e-01 5.28930247e-01 5.94229698e-02 9.13712204e-01 3.23038816e-01 -1.15378216e-01 -6.55304253e-01 -2.52384841e-01 -1.00310254e+00 2.07666755e-01 1.46896029e+00 -7.37322807e-01 -4.72999185e-01 4.39031273e-01 5.96849322e-01 -5.56666017e-01 -1.03614640e+00 3.53178203e-01 5.63225865e-01 -9.92343545e-01 1.37296009e+00 -5.99452972e-01 1.94304302e-01 -5.34421980e-01 -1.12214372e-01 -8.70322406e-01 -6.86341763e-01 -6.88103437e-01 -1.53359339e-01 1.10068333e+00 7.17493668e-02 -4.53223921e-02 1.13828111e+00 3.41502279e-01 -9.73992497e-02 -6.81809247e-01 -8.16070318e-01 -6.61415637e-01 -2.76537985e-01 -1.63699016e-01 3.08718115e-01 8.42660666e-01 -3.82597029e-01 2.41285432e-02 -2.26883546e-01 1.41073987e-01 1.01454568e+00 3.59993368e-01 7.49954402e-01 -1.09491181e+00 -3.06053698e-01 -5.20298064e-01 -8.32622468e-01 -1.43583000e+00 4.76865083e-01 -6.83237433e-01 2.16705799e-01 -1.39602900e+00 4.47735280e-01 -6.25535250e-01 -3.07333201e-01 5.98902822e-01 -2.28506193e-01 8.50187182e-01 1.50690615e-01 3.85605812e-01 -1.24565339e+00 2.67507315e-01 1.30590475e+00 -1.67224720e-01 -6.72389194e-02 -2.14632913e-01 -2.28348926e-01 7.97157228e-01 2.68315166e-01 -4.87421870e-01 -1.66736379e-01 -4.07288730e-01 -2.56314903e-01 5.30988649e-02 7.24598348e-01 -1.05449760e+00 5.70920885e-01 -6.59928247e-02 6.68046534e-01 -6.67335451e-01 2.18155846e-01 -9.42752719e-01 2.22140122e-02 2.34941408e-01 -4.17492181e-01 1.90461189e-01 1.20892003e-01 9.61839795e-01 -4.39356923e-01 1.95775375e-01 9.98431206e-01 -2.97777086e-01 -1.36860526e+00 7.34271824e-01 -9.14860889e-02 1.80823058e-01 1.12257874e+00 -6.49078786e-01 -1.15363307e-01 -1.72382712e-01 -9.54902649e-01 1.73370719e-01 7.00906515e-01 6.09876990e-01 6.42120540e-01 -1.36400235e+00 -4.54843670e-01 2.46355072e-01 2.53083736e-01 6.06582388e-02 3.40053767e-01 9.26586032e-01 -4.34336543e-01 3.04754734e-01 -5.38856268e-01 -1.19091940e+00 -1.27330625e+00 4.12868947e-01 5.16711831e-01 -1.56063884e-01 -8.04864645e-01 1.27828157e+00 5.56249499e-01 -6.54640570e-02 5.26749969e-01 -3.94198626e-01 3.88317928e-02 -1.02295354e-01 3.53873104e-01 1.51760727e-01 -9.50384215e-02 -1.06490791e+00 -5.73894799e-01 1.27481186e+00 -9.97768193e-02 1.71992347e-01 1.13591409e+00 -3.77789140e-01 -1.73237100e-01 1.79538369e-01 1.38896704e+00 -2.22231716e-01 -1.91023076e+00 -6.38698757e-01 -1.08095087e-01 -7.23443091e-01 1.98278978e-01 -4.66105521e-01 -1.32529521e+00 5.17542601e-01 6.32162690e-01 -2.79329777e-01 8.99992645e-01 4.49427336e-01 8.19917917e-01 1.68235719e-01 5.50531507e-01 -9.67206001e-01 4.06526536e-01 2.46236488e-01 4.65200037e-01 -1.53397298e+00 1.09789781e-01 -6.47395909e-01 -4.27777320e-01 1.05693102e+00 8.86899650e-01 -3.85560900e-01 7.50598073e-01 3.38048130e-01 5.80297131e-03 -3.46871644e-01 -3.59100103e-01 -4.90483820e-01 8.22433770e-01 8.32907319e-01 4.49088693e-01 -3.70067626e-01 2.22845569e-01 4.42241244e-02 3.82934093e-01 -1.55190974e-01 -5.16848378e-02 7.40005672e-01 -5.11581182e-01 -1.10537803e+00 -3.04695636e-01 2.78537452e-01 -2.86284506e-01 -7.43577182e-02 -1.56720147e-01 9.88263607e-01 2.79382139e-01 5.70951164e-01 4.91031855e-01 -1.60840511e-01 5.36627732e-02 -3.81959885e-01 7.53506601e-01 -6.53247297e-01 -6.44088030e-01 2.43011966e-01 -2.64057189e-01 -1.13260341e+00 -1.02364087e+00 -6.83121026e-01 -1.39207900e+00 3.19590829e-02 -3.05272490e-01 5.83419800e-02 1.37512952e-01 8.82436454e-01 2.87425965e-01 4.19019252e-01 3.24315012e-01 -1.25777006e+00 -1.31966233e-01 -3.84541452e-01 -3.86660010e-01 5.37344575e-01 5.10899425e-01 -8.19011271e-01 6.20748177e-02 3.61846924e-01]
[8.95966911315918, -0.32564806938171387]
fba4dff7-e140-4fab-8280-192e8864f1bf
multitask-vocal-burst-modeling-with-resnets
2206.12494
null
https://arxiv.org/abs/2206.12494v1
https://arxiv.org/pdf/2206.12494v1.pdf
Multitask vocal burst modeling with ResNets and pre-trained paralinguistic Conformers
This technical report presents the modeling approaches used in our submission to the ICML Expressive Vocalizations Workshop & Competition multitask track (ExVo-MultiTask). We first applied image classification models of various sizes on mel-spectrogram representations of the vocal bursts, as is standard in sound event detection literature. Results from these models show an increase of 21.24% over the baseline system with respect to the harmonic mean of the task metrics, and comprise our team's main submission to the MultiTask track. We then sought to characterize the headroom in the MultiTask track by applying a large pre-trained Conformer model that previously achieved state-of-the-art results on paralinguistic tasks like speech emotion recognition and mask detection. We additionally investigated the relationship between the sub-tasks of emotional expression, country of origin, and age prediction, and discovered that the best performing models are trained as single-task models, questioning whether the problem truly benefits from a multitask setting.
['Brendan Jou', 'Brian Eoff', 'Krishna Somandepalli', 'Josh Belanich']
2022-06-24
null
null
null
null
['sound-event-detection']
['audio']
[ 1.01310713e-02 -2.46426184e-02 1.91138074e-01 -3.91860962e-01 -1.17033505e+00 -4.78298843e-01 6.74622774e-01 -1.03357531e-01 -6.36390328e-01 3.01057726e-01 4.28203404e-01 8.92994180e-02 -5.74123785e-02 2.08956882e-01 -4.67148781e-01 -3.46757948e-01 -9.70088243e-02 3.87558222e-01 -2.13471293e-01 1.00147925e-01 -1.24828294e-02 -2.46181041e-02 -1.68593442e+00 9.19459581e-01 3.68219197e-01 1.14321661e+00 1.87256548e-03 1.04204595e+00 3.17841768e-03 9.05943871e-01 -1.02676666e+00 -6.89758539e-01 -2.91083574e-01 -2.08349809e-01 -8.21727812e-01 -6.30990267e-02 7.74869084e-01 1.65168986e-01 5.99782206e-02 5.87182105e-01 8.87690306e-01 1.46749496e-01 7.45373547e-01 -1.35522163e+00 -4.52678442e-01 6.67128026e-01 -2.07917199e-01 5.07252634e-01 4.71219391e-01 -4.02497441e-01 1.04354501e+00 -7.48906136e-01 5.24502516e-01 1.41635740e+00 8.58551562e-01 5.58797538e-01 -1.06596816e+00 -1.07697988e+00 1.90592915e-01 9.51835364e-02 -1.15444350e+00 -9.28812444e-01 4.19895321e-01 -7.61792719e-01 1.32054543e+00 5.26675701e-01 5.50170243e-01 1.77526891e+00 4.39072475e-02 7.49595881e-01 1.49516392e+00 -4.31326210e-01 -2.31257528e-01 5.16157687e-01 1.53902313e-02 7.58761108e-01 -7.60064244e-01 -1.75424432e-03 -1.04974687e+00 -4.05733019e-01 2.90543199e-01 -8.15222263e-01 6.30178377e-02 3.54898900e-01 -1.14289391e+00 6.84009492e-01 -4.16854411e-01 5.90846896e-01 -1.57715216e-01 2.24747926e-01 9.04122472e-01 4.21000570e-01 1.17616308e+00 3.56075376e-01 -6.46158636e-01 -6.13308728e-01 -1.12764490e+00 2.22992986e-01 8.11462581e-01 6.62059188e-01 6.65851170e-03 1.51951402e-01 -4.30905193e-01 1.44316149e+00 9.33033228e-02 1.35613590e-01 6.85947597e-01 -1.05649173e+00 1.56295255e-01 -1.09319091e-01 -2.27244273e-01 -3.63900214e-01 -6.27728820e-01 -4.63330775e-01 -1.62077531e-01 -1.31274566e-01 4.46446002e-01 -4.74734038e-01 -7.62836099e-01 1.78970909e+00 -9.95329171e-02 2.18898073e-01 -2.43633538e-01 3.52701992e-01 9.46124971e-01 5.69690049e-01 6.38915122e-01 -4.56208110e-01 1.59287286e+00 -1.02405357e+00 -8.07619393e-01 -2.14559197e-01 7.38066256e-01 -1.08134961e+00 1.05215430e+00 8.14504445e-01 -1.13850892e+00 -6.99275374e-01 -7.29684293e-01 2.37938732e-01 -3.83251071e-01 3.50660026e-01 7.18289316e-01 1.20741153e+00 -9.61506486e-01 2.45041803e-01 -3.97295356e-01 -4.09374505e-01 6.73295036e-02 1.82360619e-01 -2.88782418e-01 5.13207257e-01 -9.65932488e-01 1.06027770e+00 -1.50704548e-01 -4.19787884e-01 -9.35328841e-01 -8.95212829e-01 -5.97471654e-01 -1.14982635e-01 6.04258068e-02 -2.64951438e-01 1.62819350e+00 -1.11626363e+00 -1.36886501e+00 1.45811450e+00 -1.75601348e-01 -2.20378757e-01 3.23973477e-01 -3.28179717e-01 -9.28316653e-01 -6.09940551e-02 -1.28293768e-01 5.14304817e-01 1.16503179e+00 -9.27948594e-01 -7.51220524e-01 -1.81688026e-01 -3.52023244e-01 4.49445173e-02 -6.23880744e-01 1.11466718e+00 -8.16022828e-02 -6.67238116e-01 -6.46582127e-01 -1.01453519e+00 1.53383493e-01 -6.32738769e-01 -1.63623855e-01 -5.97710252e-01 5.27183771e-01 -8.86310399e-01 1.38132226e+00 -2.36145067e+00 -3.36417817e-02 -2.30329320e-01 8.69586840e-02 -7.05207288e-02 -1.35126740e-01 1.86011836e-01 -4.69385743e-01 3.92569453e-01 4.61295217e-01 -9.39474344e-01 2.82928109e-01 -2.45216176e-01 -1.85190633e-01 2.79250234e-01 -7.66338855e-02 4.61354703e-01 -3.83652329e-01 -7.08246052e-01 -7.83016160e-03 3.03686291e-01 -3.46521676e-01 1.43251196e-01 4.90660258e-02 2.38011345e-01 -1.03981808e-01 5.79273283e-01 8.82115886e-02 2.43702874e-01 -1.42962083e-01 -7.03334212e-02 -4.17472303e-01 3.37200195e-01 -6.08347237e-01 1.72573805e+00 -8.43792617e-01 8.41514409e-01 5.26748896e-01 -7.98298597e-01 6.60474896e-01 9.73075092e-01 6.58209980e-01 -4.25820380e-01 1.97805315e-01 1.62901342e-01 2.26983324e-01 -7.03494072e-01 3.33456904e-01 -4.43662167e-01 -3.78491163e-01 1.90251738e-01 4.93870944e-01 -2.50890493e-01 -6.76539093e-02 -9.84607041e-02 9.02087808e-01 1.69583127e-01 8.15058798e-02 -3.33077163e-01 1.12370461e-01 -4.07874763e-01 3.61926317e-01 6.77406371e-01 -4.27887827e-01 6.09092474e-01 4.64705318e-01 -1.52175501e-01 -6.57772779e-01 -8.11825752e-01 -3.44104081e-01 1.98709059e+00 -9.43475544e-01 -7.61710584e-01 -8.50786269e-01 -7.65389323e-01 -3.77512455e-01 8.55791211e-01 -8.12083125e-01 8.54666233e-02 -2.87685305e-01 -9.28255081e-01 1.14419961e+00 2.86286741e-01 -2.70500153e-01 -1.14118147e+00 -6.05802834e-01 2.34809533e-01 -2.98616022e-01 -1.44041455e+00 -4.87408340e-01 6.15251720e-01 -2.66394615e-01 -4.80836332e-01 -7.81354249e-01 -7.62032688e-01 -1.76223949e-01 -6.99166834e-01 1.16924608e+00 -3.18214446e-01 -6.03387535e-01 7.27240980e-01 -2.95600384e-01 -1.09088516e+00 -6.17009103e-01 1.91090882e-01 7.90953860e-02 7.69887865e-02 3.33492965e-01 -5.36619365e-01 -8.67293403e-02 1.44906521e-01 -4.72587049e-01 -1.25221759e-01 3.18389505e-01 6.46756709e-01 1.05452724e-02 -4.33453083e-01 1.02056420e+00 -6.82581067e-01 8.23066473e-01 -4.11535323e-01 -5.80504946e-02 3.33334148e-01 -2.98700362e-01 -3.42284024e-01 7.80821443e-02 -8.96543086e-01 -9.95480061e-01 1.07651345e-01 -3.76112849e-01 -4.57206249e-01 -4.98450190e-01 1.50460958e-01 2.92528898e-01 8.38979408e-02 3.88052076e-01 -2.41103634e-01 -2.90794909e-01 -4.66049939e-01 1.22460902e-01 9.97742772e-01 4.36342627e-01 -8.28467369e-01 5.69417961e-02 -2.58303769e-02 -4.74251360e-01 -1.08640361e+00 -9.27675009e-01 -4.78964090e-01 -3.58705312e-01 -6.05311036e-01 1.24451065e+00 -1.11593056e+00 -8.64026189e-01 4.80993807e-01 -1.19232309e+00 -4.52177912e-01 -3.65509875e-02 6.48202300e-01 -6.89201593e-01 2.72264741e-02 -6.69429898e-01 -1.27107573e+00 -2.48416066e-01 -9.61162984e-01 1.26840448e+00 -2.20942333e-01 -8.37833047e-01 -1.06622469e+00 2.98426509e-01 7.14882851e-01 2.96739310e-01 1.35998592e-01 1.10404825e+00 -9.96660113e-01 4.72686291e-01 1.92453023e-02 1.54248402e-01 4.94480848e-01 -1.90833032e-01 8.92749652e-02 -1.62765157e+00 4.91341427e-02 1.48493245e-01 -8.19708645e-01 8.98964167e-01 5.45040786e-01 1.40359437e+00 2.27826700e-01 -1.95173427e-01 2.73340255e-01 6.29038274e-01 2.20249787e-01 2.28092402e-01 1.04704993e-02 3.27158451e-01 1.01771390e+00 4.44914758e-01 4.85260308e-01 4.04597431e-01 8.74954820e-01 4.10288833e-02 -9.70033780e-02 -2.39056394e-01 -1.17250100e-01 7.21740961e-01 1.05478549e+00 -1.52950332e-01 -1.58239380e-01 -7.92731345e-01 7.78445601e-01 -1.55788040e+00 -9.28850770e-01 -7.37737864e-02 1.76912892e+00 7.18950570e-01 6.43585995e-02 5.42943895e-01 -3.78789380e-02 5.36661685e-01 3.92208248e-01 4.10693288e-02 -1.15260768e+00 2.01988034e-02 5.63637257e-01 1.30187258e-01 2.10807651e-01 -1.28088284e+00 8.44106674e-01 7.93533421e+00 1.21478343e+00 -1.15506029e+00 7.02031672e-01 7.54146338e-01 -7.01528013e-01 1.36345625e-01 -4.94563639e-01 -8.63109231e-01 1.97508961e-01 1.66346633e+00 -5.20032085e-02 3.79088372e-01 7.10676789e-01 9.32623819e-02 -4.12526689e-02 -1.21548867e+00 1.01176977e+00 6.21980965e-01 -6.40286148e-01 -5.65138102e-01 1.77417338e-01 5.50600290e-01 1.36489093e-01 3.34036738e-01 7.76618600e-01 1.89828798e-02 -1.21464825e+00 1.17073679e+00 3.98535013e-01 8.26026976e-01 -6.54655814e-01 1.57973096e-01 3.16405147e-01 -8.71826470e-01 -3.02959025e-01 8.21395516e-02 -7.42368996e-02 1.78740904e-01 2.52725959e-01 -8.51422906e-01 2.68685222e-01 9.59832430e-01 2.86303163e-01 -4.64925289e-01 6.99901164e-01 3.07254791e-01 9.20147657e-01 -2.40748987e-01 -6.42874613e-02 9.41211134e-02 2.53666937e-01 5.22414684e-01 1.82353950e+00 5.34083307e-01 -3.05005819e-01 1.21177979e-01 4.22245890e-01 -1.58551604e-01 5.51174521e-01 -4.40944195e-01 -3.31398696e-01 2.36626882e-02 1.58817649e+00 -4.49435860e-01 -2.02022672e-01 -6.42490804e-01 8.12525690e-01 2.20998675e-01 1.51488364e-01 -9.53586340e-01 -8.81189108e-02 5.93823791e-01 -5.61157912e-02 8.45825523e-02 8.46553668e-02 -1.91724271e-01 -8.02011311e-01 -3.87313694e-01 -1.04209340e+00 4.65161473e-01 -8.11127603e-01 -1.24142146e+00 8.15523744e-01 1.53853357e-01 -5.81861317e-01 -4.44032729e-01 -6.06632352e-01 -5.87858498e-01 7.09377050e-01 -1.12872767e+00 -1.42560661e+00 1.23969823e-01 4.15442437e-01 8.14550281e-01 -3.30794454e-01 1.30679059e+00 5.52624166e-01 -5.80169499e-01 8.59911382e-01 -4.12980795e-01 -1.59716427e-01 1.27915573e+00 -1.29115856e+00 9.59420502e-02 2.87354350e-01 5.66800356e-01 1.77692488e-01 7.52793729e-01 -2.38893494e-01 -7.58644640e-01 -6.81533813e-01 1.10447431e+00 -7.33557224e-01 9.92005885e-01 -7.42904067e-01 -3.98651034e-01 7.80304611e-01 5.34823179e-01 -3.67346197e-01 1.11711717e+00 6.95487022e-01 -4.10173953e-01 1.61157265e-01 -8.40353489e-01 2.68890530e-01 8.30954075e-01 -8.10525596e-01 -4.92787123e-01 5.38086832e-01 4.80322808e-01 -2.11651966e-01 -1.13806570e+00 3.63904148e-01 9.48022425e-01 -8.21088910e-01 8.57969820e-01 -7.92341352e-01 5.61317921e-01 6.57955527e-01 -2.53659099e-01 -1.43448114e+00 -1.41187012e-01 -8.44425142e-01 2.45972812e-01 1.64733768e+00 6.10562265e-01 -1.56326592e-01 3.48206103e-01 3.71727079e-01 -4.85723436e-01 -6.50456667e-01 -1.18206620e+00 -6.83010578e-01 3.51671427e-01 -9.07711506e-01 4.77469973e-02 8.82615209e-01 8.36385190e-02 5.08735299e-01 -6.50580704e-01 -1.66856140e-01 1.40679553e-01 -2.96771169e-01 4.15746242e-01 -1.35017896e+00 -3.87000322e-01 -6.37687624e-01 -1.44545749e-01 -3.58308822e-01 6.89117908e-01 -9.01044846e-01 1.32778883e-01 -1.19463575e+00 3.13057214e-01 4.46636640e-02 -4.20470953e-01 5.09194195e-01 -3.96724045e-02 3.91572088e-01 2.18965262e-01 -4.04945672e-01 -5.96450567e-01 1.72251478e-01 8.00409138e-01 9.48461369e-02 2.58950293e-02 1.68991894e-01 -7.15113878e-01 9.97585356e-01 6.48994803e-01 -6.64658666e-01 -1.29993886e-01 -3.39127570e-01 1.30239710e-01 1.52564213e-01 2.29228601e-01 -1.00389218e+00 -6.31494820e-02 1.36019841e-01 3.03928047e-01 -1.94015726e-01 1.06471467e+00 -3.50987494e-01 -8.61000083e-03 -4.06968109e-02 -6.29050434e-01 1.18435234e-01 5.07992566e-01 1.51715800e-02 -2.29700446e-01 -3.54276985e-01 5.70394397e-01 -1.15948789e-01 -2.93356419e-01 -4.68607098e-02 -9.72605944e-01 2.24486604e-01 8.47710311e-01 1.13957509e-01 -1.62020922e-01 -5.15525162e-01 -1.24816215e+00 -1.66936114e-01 -1.98808610e-01 9.01949942e-01 1.15404189e-01 -1.00762892e+00 -1.02053344e+00 -1.76348880e-01 1.45119995e-01 -1.14859343e+00 3.49031240e-01 1.12145805e+00 2.59063393e-01 3.63368869e-01 -6.80747703e-02 -3.15175712e-01 -2.13948774e+00 1.85233131e-01 3.45529884e-01 -3.34488243e-01 -6.64061829e-02 1.12770438e+00 2.19693810e-01 -5.02142131e-01 5.29378712e-01 -2.35910431e-01 -2.22656086e-01 6.30708575e-01 2.98633665e-01 3.93390894e-01 1.26412973e-01 -6.62391961e-01 -3.00017476e-01 3.78235817e-01 -7.33551988e-03 -6.53372407e-01 1.29865956e+00 7.52665550e-02 -4.20322716e-02 1.03456748e+00 1.29501402e+00 3.69830132e-01 -6.17293000e-01 1.76529720e-01 1.94984928e-01 2.91151479e-02 2.56029040e-01 -1.16320384e+00 -6.07228816e-01 1.05828905e+00 7.51086116e-01 4.08712715e-01 1.01785457e+00 2.97169536e-01 4.45636779e-01 3.14154290e-02 -1.17794000e-01 -1.44358265e+00 2.76333332e-01 5.20220280e-01 1.06301296e+00 -9.19483662e-01 -2.74026930e-01 -3.41350704e-01 -9.03259099e-01 8.85525703e-01 5.18187881e-01 3.66524488e-01 6.44643664e-01 5.29870033e-01 3.01233798e-01 -2.32779399e-01 -1.11698794e+00 -1.63130656e-01 5.25834024e-01 5.47775626e-01 9.17627037e-01 2.37169847e-01 -1.78880677e-01 1.16890311e+00 -5.97835302e-01 -2.88829058e-01 1.59165144e-01 3.47572058e-01 -2.39615545e-01 -1.05686605e+00 -2.99946249e-01 4.71754611e-01 -1.30763805e+00 -1.21022262e-01 -6.71543539e-01 6.59077108e-01 4.60155219e-01 1.22843373e+00 7.69486949e-02 -5.23773730e-01 3.47669661e-01 8.73482823e-01 7.00212359e-01 -8.31673324e-01 -1.35212660e+00 4.88738149e-01 8.81037474e-01 -2.74085313e-01 -5.04837632e-01 -9.64584947e-01 -7.27644086e-01 3.95526588e-01 -2.78098881e-01 1.90420225e-01 1.12386858e+00 8.77126098e-01 -1.16226552e-02 8.58104050e-01 4.82920647e-01 -6.99855030e-01 -3.29761773e-01 -1.42144763e+00 -5.05035341e-01 1.79416582e-01 1.28002837e-01 -4.34923917e-01 -4.55202341e-01 9.77710858e-02]
[13.592513084411621, 5.7543134689331055]
fd1a42c9-0649-4c74-b04c-231a6c49b4b5
eco-amlp-a-decision-support-system-using-an
1706.07679
null
http://arxiv.org/abs/1706.07679v1
http://arxiv.org/pdf/1706.07679v1.pdf
ECO-AMLP: A Decision Support System using an Enhanced Class Outlier with Automatic Multilayer Perceptron for Diabetes Prediction
With advanced data analytical techniques, efforts for more accurate decision support systems for disease prediction are on rise. Surveys by World Health Organization (WHO) indicate a great increase in number of diabetic patients and related deaths each year. Early diagnosis of diabetes is a major concern among researchers and practitioners. The paper presents an application of \textit{Automatic Multilayer Perceptron }which\textit{ }is combined with an outlier detection method \textit{Enhanced Class Outlier Detection using distance based algorithm }to create a prediction framework named as Enhanced Class Outlier with Automatic Multi layer Perceptron (ECO-AMLP). A series of experiments are performed on publicly available Pima Indian Diabetes Dataset to compare ECO-AMLP with other individual classifiers as well as ensemble based methods. The outlier technique used in our framework gave better results as compared to other pre-processing and classification techniques. Finally, the results are compared with other state-of-the-art methods reported in literature for diabetes prediction on PIDD and achieved accuracy of 88.7\% bests all other reported studies.
['Hammad Afzal', 'Raheel Nawaz', 'Khawar Khurshid', 'Mehreen Ahmed', 'Maham Jahangir']
2017-06-23
null
null
null
null
['diabetes-prediction']
['medical']
[-5.07768616e-02 -2.59715080e-01 5.38110584e-02 -7.32258201e-01 -2.63666630e-01 2.08255515e-01 2.19359696e-01 1.39520121e+00 -4.51644570e-01 9.55728650e-01 1.02500431e-01 -1.14914812e-01 -3.93996686e-01 -5.45731843e-01 -2.42991596e-01 -5.38169444e-01 -2.27570608e-01 1.00798821e+00 -5.30721620e-02 1.19086832e-01 5.73324263e-01 6.17370307e-01 -1.62906563e+00 7.54314125e-01 1.33418083e+00 1.00068712e+00 -3.98128808e-01 8.37139845e-01 -6.97413012e-02 1.03446496e+00 -3.11623663e-01 -1.46055177e-01 4.44817215e-01 -3.97278458e-01 -2.14932308e-01 -2.37842157e-01 3.67754251e-01 -3.22667003e-01 7.47759938e-02 7.74096549e-01 7.82796025e-01 -6.71281070e-02 1.11078501e+00 -1.32079828e+00 -5.56397676e-01 3.54134947e-01 -6.07541263e-01 4.59270179e-01 1.77476779e-01 -2.06082329e-01 2.49545664e-01 -7.20399201e-01 2.34003708e-01 9.64074552e-01 1.17962444e+00 6.74433410e-02 -1.28683484e+00 -6.86421454e-01 -2.83755362e-01 4.37671214e-01 -1.17198896e+00 -2.56752253e-01 5.12266934e-01 -7.17077374e-01 1.26492918e+00 3.74073803e-01 4.00984198e-01 5.98465621e-01 7.49230325e-01 3.55778039e-01 1.43898702e+00 -4.89692569e-01 3.29513550e-01 3.81846994e-01 4.41336483e-01 6.81017399e-01 7.84630716e-01 2.68604249e-01 -3.81871343e-01 -4.73519534e-01 4.10305977e-01 4.68753904e-01 5.04091918e-01 -6.58663139e-02 -1.01522398e+00 7.59220421e-01 -4.91675623e-02 3.58195379e-02 -1.01403582e+00 -3.68635088e-01 6.74598753e-01 5.79201758e-01 5.29329658e-01 1.88147828e-01 -7.38589108e-01 -2.68926658e-02 -8.96824062e-01 3.16328287e-01 8.78704131e-01 4.40772980e-01 -4.45716940e-02 1.10471368e-01 4.56056148e-02 9.96198177e-01 7.43206918e-01 2.64790982e-01 3.93737942e-01 -6.61691189e-01 3.67256492e-01 1.34714508e+00 1.85715199e-01 -9.74764109e-01 -6.71184778e-01 -3.83755416e-01 -1.20270693e+00 6.91578150e-01 5.66423297e-01 -3.18566978e-01 -1.10551977e+00 7.97644734e-01 1.65547132e-01 3.34841520e-01 5.06226301e-01 3.55236262e-01 6.86447561e-01 3.82858455e-01 3.66608024e-01 -2.82982975e-01 1.02722228e+00 -4.36062276e-01 -4.92044419e-01 4.71870452e-01 6.46180093e-01 -9.54038262e-01 3.15747350e-01 1.13623869e+00 -5.49254894e-01 -3.07059139e-01 -9.31997180e-01 5.76357365e-01 -4.74623203e-01 2.23685727e-01 5.94355285e-01 7.74458826e-01 -4.84717578e-01 7.29289353e-01 -1.00290179e+00 -9.55045879e-01 6.86760366e-01 7.04604208e-01 -4.38568681e-01 -1.32234693e-01 -5.23774803e-01 1.08093476e+00 6.39755487e-01 -2.35114157e-01 -3.82087648e-01 -7.79542267e-01 -4.93875891e-01 -6.36939883e-01 -4.81075764e-01 -5.63712597e-01 6.93177700e-01 -9.04248774e-01 -1.06317317e+00 8.61791551e-01 9.83705446e-02 -7.02599883e-01 6.30973279e-01 -3.47505659e-01 -9.72983778e-01 -1.10790215e-01 -1.46628231e-01 3.71632636e-01 1.36152953e-01 -8.26165855e-01 -9.93295670e-01 -8.15787792e-01 -8.28394115e-01 8.00770074e-02 1.47514179e-01 3.16765010e-01 6.93083465e-01 -7.29940057e-01 5.90302169e-01 -6.47081256e-01 -4.06275183e-01 -1.63563401e-01 -5.01819134e-01 -3.12974304e-01 6.23716295e-01 -8.58809710e-01 1.26907587e+00 -1.71641695e+00 -3.02340984e-01 3.74925256e-01 7.48717180e-03 3.02711040e-01 5.23904741e-01 7.09407151e-01 -4.43512678e-01 -1.11294843e-01 -5.64317266e-03 1.90537125e-02 -1.19497411e-01 5.13907313e-01 1.35667101e-01 5.69895566e-01 9.47229415e-02 -5.39490283e-02 -4.06356215e-01 -4.06027466e-01 6.09610200e-01 3.55069399e-01 -6.77989721e-01 2.00779596e-03 -8.37718695e-02 5.12914419e-01 -3.23817402e-01 1.26649690e+00 6.64721310e-01 3.41661900e-01 -1.82112470e-01 -1.66087940e-01 -2.26337090e-01 -4.99643147e-01 -1.58238780e+00 9.75230515e-01 4.85028654e-01 1.48240700e-01 -8.53743136e-01 -1.32011664e+00 1.30850065e+00 4.83751953e-01 5.95644593e-01 -3.79965872e-01 1.98047578e-01 6.25320077e-01 3.07233334e-01 -1.15770960e+00 -1.50571600e-01 -1.44468889e-01 4.01650071e-01 -6.75795674e-02 -4.27740291e-02 9.84064698e-01 1.41789511e-01 -2.77815402e-01 1.05026996e+00 2.39745885e-01 7.68273592e-01 -3.93825024e-01 6.01332009e-01 3.61320943e-01 8.80923450e-01 7.25510240e-01 -6.50148571e-01 4.59615767e-01 5.94813406e-01 -1.01674068e+00 -1.23503900e+00 -1.01749933e+00 -6.62843347e-01 6.45501256e-01 -5.39229810e-01 1.05271272e-01 -1.64690152e-01 -4.36558604e-01 6.50455296e-01 6.32343650e-01 -4.73656386e-01 3.50866556e-01 -2.81793684e-01 -1.45769346e+00 8.39867055e-01 4.20886934e-01 8.21058810e-01 -8.85400951e-01 -3.43754530e-01 5.03402114e-01 4.38878894e-01 -4.79698986e-01 6.99814558e-01 2.55880952e-01 -1.38828635e+00 -1.22824621e+00 -4.68095779e-01 -8.14887464e-01 6.27261758e-01 -9.06072021e-01 1.03238690e+00 -2.02236190e-01 -6.84668660e-01 -2.42511347e-01 -3.14186335e-01 -1.15790796e+00 -5.76892078e-01 -4.27018136e-01 3.94776851e-01 -2.20469475e-01 1.18787861e+00 -8.63219440e-01 -8.42917800e-01 -1.52021348e-01 -5.49584210e-01 -3.77543241e-01 9.30367708e-01 3.48478794e-01 5.23484409e-01 4.10704285e-01 1.07263863e+00 -1.10369742e+00 3.10861945e-01 -9.42848265e-01 -3.78895938e-01 -9.17067975e-02 -9.51806128e-01 1.01154344e-03 3.99467617e-01 -5.67256510e-02 -5.69874406e-01 1.25204623e-01 -2.19331756e-01 1.85250521e-01 -8.63969684e-01 3.62844199e-01 1.69133693e-01 8.38137344e-02 7.22514868e-01 9.60191265e-02 1.67111412e-01 -1.00317693e+00 -5.01233935e-01 1.12492335e+00 2.29825303e-01 -2.20219001e-01 -4.75353412e-02 1.47882253e-01 2.60392100e-01 -9.31200504e-01 -4.16001856e-01 -5.66238344e-01 -7.78364837e-01 -5.93287237e-02 8.63385558e-01 -8.41494679e-01 -6.79004908e-01 1.05885375e+00 -7.06226587e-01 3.83128732e-01 2.86865413e-01 9.99491274e-01 -1.56874090e-01 1.15716346e-01 -5.67725778e-01 -1.17397225e+00 -4.99319941e-01 -8.50446105e-01 3.24615419e-01 5.32291196e-02 -4.34017032e-01 -1.06019926e+00 5.30799687e-01 4.85005200e-01 3.50650817e-01 1.15204763e+00 1.43454731e+00 -1.63287044e+00 1.92987313e-03 -6.02432966e-01 -1.36051327e-01 6.66948318e-01 1.44673213e-01 1.22593060e-01 -5.85950196e-01 -2.75058709e-02 -3.69459242e-01 2.05383390e-01 5.77613652e-01 5.32958746e-01 5.59776723e-01 -4.35813010e-01 -3.04869831e-01 1.91435277e-01 2.20514560e+00 6.38605356e-01 6.18178308e-01 6.16228700e-01 3.28063577e-01 -5.81911542e-02 3.32296610e-01 9.00562525e-01 4.46046472e-01 2.09245995e-01 2.92375863e-01 1.29816845e-01 3.70869488e-01 4.89889532e-01 1.36236593e-01 4.70588803e-01 -4.38174069e-01 9.70218554e-02 -1.29577041e+00 3.69347483e-01 -2.02538514e+00 -1.01414526e+00 -6.18409276e-01 2.40128422e+00 5.57261229e-01 9.49668735e-02 4.08959329e-01 5.59633315e-01 5.80206931e-01 -7.43750632e-01 -5.14054000e-01 -9.21864331e-01 1.94536626e-01 3.21850479e-01 5.26587963e-01 2.49023005e-01 -1.35393691e+00 5.21844774e-02 5.86800718e+00 -1.68861240e-01 -9.04349983e-01 -1.16428971e-01 6.41588569e-01 -5.57511188e-02 6.47833943e-01 -4.24892128e-01 -6.48810029e-01 7.45056033e-01 1.09409738e+00 1.59369826e-01 -1.66637704e-01 8.64344120e-01 4.93417770e-01 -4.31302756e-01 -1.20237637e+00 8.32209051e-01 2.02381030e-01 -9.04512227e-01 2.68832654e-01 1.21490443e-02 8.27897906e-01 3.25580299e-01 -4.81828064e-01 1.97319940e-01 2.86409885e-01 -1.12017906e+00 -1.06913365e-01 1.20784545e+00 4.94242273e-02 -1.15402889e+00 1.47316384e+00 1.49538770e-01 -5.90674758e-01 -3.22632551e-01 -3.85236382e-01 -3.97110432e-01 -2.70993620e-01 9.76570249e-01 -8.79657686e-01 5.41570663e-01 8.59830201e-01 6.23320401e-01 -5.81176519e-01 1.78109300e+00 6.06583238e-01 6.84249282e-01 -4.80215311e-01 1.41434073e-01 7.71603594e-03 -2.78753072e-01 5.96285641e-01 1.17245495e+00 4.29095089e-01 1.61310911e-01 2.78909773e-01 1.53642111e-02 4.34209824e-01 6.41712070e-01 -6.16994619e-01 1.43354148e-01 3.31975430e-01 5.50717831e-01 -4.89135891e-01 -5.60773015e-01 -5.42925060e-01 7.86375344e-01 -1.49355993e-01 -3.09785847e-02 -5.60887694e-01 -6.26637995e-01 6.66022301e-01 2.65757084e-01 -5.23256436e-02 1.22021854e-01 -5.86483657e-01 -7.44214654e-01 -1.54565781e-01 -1.09738708e+00 9.68884110e-01 -4.02709961e-01 -1.62337959e+00 2.59713471e-01 -1.66874751e-01 -1.29700994e+00 1.94829166e-01 -6.54109895e-01 -5.44032514e-01 6.84315264e-01 -1.15751302e+00 -8.82001400e-01 -3.00751925e-01 5.50296843e-01 4.05193031e-01 -7.54211426e-01 1.17892683e+00 7.81366944e-01 -9.44536269e-01 3.78580987e-01 5.91363788e-01 3.58233564e-02 8.86908710e-01 -1.32741916e+00 -6.33245587e-01 7.55912244e-01 -6.30805314e-01 3.89697403e-01 8.45124006e-01 -9.69482124e-01 -8.66792917e-01 -1.37001014e+00 1.15756345e+00 -5.12789547e-01 2.13590384e-01 4.42481667e-01 -7.83508420e-01 8.12479138e-01 2.63284206e-01 -1.00996383e-01 1.33350039e+00 -4.82142158e-02 1.38663650e-01 -5.96084535e-01 -1.87094164e+00 1.42279088e-01 2.38799304e-01 3.70163858e-01 -8.42202961e-01 3.44160110e-01 -2.17882052e-01 -1.95553884e-01 -1.40991223e+00 6.15520537e-01 7.49365747e-01 -1.36008072e+00 7.83970177e-01 -9.64840829e-01 2.52958179e-01 -3.41970295e-01 -5.31436682e-01 -8.72654974e-01 -9.94577780e-02 -2.15460092e-01 -1.57761678e-01 1.28179479e+00 3.73190671e-01 -8.37231815e-01 7.67596841e-01 6.47333145e-01 -6.77542314e-02 -8.15711081e-01 -7.59863436e-01 -4.03789639e-01 -5.27869463e-02 -3.83514196e-01 2.89251488e-02 1.23887229e+00 -9.72176492e-02 -1.44850343e-01 -3.57377946e-01 6.26212597e-01 1.15906274e+00 -4.67668891e-01 3.99274945e-01 -2.00166345e+00 9.88183394e-02 9.76420194e-02 -1.29353082e+00 4.02995825e-01 -7.53378987e-01 -8.09495091e-01 -6.79809690e-01 -1.68069661e+00 2.96717942e-01 -5.17407596e-01 -9.66493845e-01 5.87576210e-01 1.41603515e-01 3.56164485e-01 -4.01965678e-01 3.30101430e-01 -3.85253191e-01 -2.92998075e-01 3.47739577e-01 3.66369426e-01 -6.21209145e-01 2.83840328e-01 -5.55380940e-01 1.03241539e+00 1.04589927e+00 -6.84839725e-01 -7.01895729e-02 3.52063007e-03 7.11803958e-02 -2.78985023e-01 3.32187325e-01 -1.64396405e+00 2.17932053e-02 -1.69641972e-01 1.21420574e+00 -1.01986778e+00 -2.04154029e-01 -1.03705060e+00 5.44768214e-01 9.20245588e-01 -1.10711850e-01 3.91566724e-01 9.06922668e-03 5.68927824e-01 6.70672441e-03 -6.80638012e-03 9.80898082e-01 -1.44264519e-01 -6.31589115e-01 -4.27416805e-03 -6.27101183e-01 -4.37385201e-01 1.52878261e+00 -4.84932274e-01 -2.48917297e-01 3.19321275e-01 -1.14003909e+00 1.34671390e-01 2.30651423e-01 2.50399828e-01 4.72928137e-01 -1.28900540e+00 -1.04532754e+00 2.24302098e-01 3.54344398e-01 -4.08241123e-01 -1.32153984e-02 1.26694238e+00 -1.26240146e+00 3.81344885e-01 -7.66810477e-01 -6.51898146e-01 -1.72596383e+00 2.44982302e-01 2.49178976e-01 -1.57149490e-02 -8.43923569e-01 3.78744513e-01 -9.50834572e-01 -3.95206153e-01 6.00491643e-01 -5.81017733e-01 -5.13180912e-01 7.77974129e-02 4.62023526e-01 1.02593744e+00 1.70439959e-01 -2.35714868e-01 -6.47968948e-01 2.59232342e-01 -3.46691817e-01 4.42912608e-01 1.72717929e+00 1.77440271e-01 -4.33318228e-01 7.05382466e-01 7.37628877e-01 -4.38073397e-01 -5.66561580e-01 5.43122329e-02 5.64721882e-01 -2.45252952e-01 -1.34905428e-01 -1.44689429e+00 -4.82526690e-01 3.40850502e-01 1.56794488e+00 -9.68990773e-02 1.02433157e+00 -6.35962784e-01 5.03481925e-01 8.04962873e-01 -1.36491388e-01 -1.20668912e+00 -2.72579640e-01 1.69624314e-01 6.02982998e-01 -1.61063004e+00 3.62034917e-01 1.05249964e-01 -7.57510424e-01 1.29998696e+00 3.68922561e-01 -5.87210834e-01 9.86844242e-01 1.54189050e-01 4.12886947e-01 8.20196513e-03 -7.11596012e-01 3.39203805e-01 1.38992518e-02 7.56887972e-01 6.56709373e-01 4.58860919e-02 -5.81931114e-01 5.98519266e-01 2.25726902e-01 7.69760191e-01 2.95968264e-01 1.20249426e+00 -9.52367961e-01 -1.16672802e+00 -4.57642227e-01 1.32982385e+00 -1.01493049e+00 -1.27725258e-01 -1.40055746e-01 8.73892307e-01 6.33619726e-01 8.50802600e-01 4.51535918e-02 -9.44105908e-02 3.73872876e-01 7.63078511e-01 3.32713127e-01 -3.31402361e-01 -6.18631899e-01 -3.38742256e-01 4.17999387e-01 -2.53417790e-01 -4.87689286e-01 -8.12101483e-01 -1.24878156e+00 -3.76252383e-01 7.92533904e-02 -8.94203782e-03 7.86195457e-01 8.81357789e-01 4.43598241e-01 4.74952579e-01 3.12268853e-01 -2.62468547e-01 -2.86609948e-01 -1.09058440e+00 -8.02668929e-01 2.98802793e-01 2.35811427e-01 -4.08439100e-01 -9.64822993e-02 3.19820255e-01]
[8.389045715332031, 4.933208465576172]
daa1a7e8-b5e3-45f6-a6cd-79599fb0d56c
tunet-a-block-online-bandwidth-extension
2110.13492
null
https://arxiv.org/abs/2110.13492v5
https://arxiv.org/pdf/2110.13492v5.pdf
TUNet: A Block-online Bandwidth Extension Model based on Transformers and Self-supervised Pretraining
We introduce a block-online variant of the temporal feature-wise linear modulation (TFiLM) model to achieve bandwidth extension. The proposed architecture simplifies the UNet backbone of the TFiLM to reduce inference time and employs an efficient transformer at the bottleneck to alleviate performance degradation. We also utilize self-supervised pretraining and data augmentation to enhance the quality of bandwidth extended signals and reduce the sensitivity with respect to downsampling methods. Experiment results on the VCTK dataset show that the proposed method outperforms several recent baselines in both intrusive and non-intrusive metrics. Pretraining and filter augmentation also help stabilize and enhance the overall performance.
['Andy W. H. Khong', 'Anh H. T. Nguyen', 'Viet-Anh Nguyen']
2021-10-26
null
null
null
null
['bandwidth-extension', 'audio-super-resolution', 'audio-super-resolution', 'bandwidth-extension']
['audio', 'audio', 'music', 'speech']
[ 2.35405907e-01 -5.82987480e-02 -6.04936004e-01 -4.40414816e-01 -8.51375699e-01 -1.84924796e-01 7.61275768e-01 -3.77032250e-01 -4.41291779e-01 6.52007759e-01 2.40305841e-01 -5.22144377e-01 -2.47393981e-01 -4.52365041e-01 -5.16878843e-01 -4.84879076e-01 -2.78228998e-01 -2.07813337e-01 3.38282764e-01 -5.49114533e-02 5.42772710e-02 3.41594934e-01 -1.21048045e+00 5.55926085e-01 9.64989603e-01 1.51219976e+00 -1.85360119e-01 4.10836756e-01 7.14147389e-02 9.33473468e-01 -6.19315863e-01 -2.92775840e-01 4.76237506e-01 -2.09322780e-01 -7.02727437e-01 5.28614409e-02 4.98917222e-01 -3.58676970e-01 -7.14773178e-01 7.00357080e-01 6.26536429e-01 3.39184374e-01 5.44463813e-01 -1.31510222e+00 -1.67301208e-01 4.63540226e-01 -8.78327906e-01 7.67253697e-01 -7.34039322e-02 6.68060854e-02 7.10708261e-01 -1.06600273e+00 2.27713343e-02 1.35624552e+00 8.45082402e-01 2.60388374e-01 -1.32276452e+00 -1.07213712e+00 2.72369921e-01 6.23272419e-01 -1.59861469e+00 -8.96520138e-01 6.30212367e-01 1.12265058e-01 1.03902566e+00 2.92125940e-01 1.62274912e-01 1.21033108e+00 1.29763588e-01 7.65136003e-01 9.64176118e-01 -5.12086093e-01 3.05848103e-02 1.12012543e-01 2.45627597e-01 8.06281924e-01 -1.47172153e-01 4.44592714e-01 -1.05261552e+00 -4.00695741e-01 6.11826479e-01 -2.78097063e-01 -2.00060710e-01 -1.29719496e-01 -8.04098487e-01 8.02972674e-01 3.58732283e-01 1.65736333e-01 -1.04280621e-01 2.34362289e-01 6.05413377e-01 4.77374643e-01 6.10804141e-01 1.90307423e-01 -5.00744402e-01 -2.94562995e-01 -1.12942684e+00 -2.66707212e-01 3.97592813e-01 8.85368645e-01 4.06500995e-01 5.42936325e-01 -3.94004703e-01 1.02081239e+00 1.68617681e-01 3.42389435e-01 8.37336183e-01 -1.00844467e+00 5.72529972e-01 2.30615139e-01 -2.37800837e-01 -8.61646652e-01 -4.32442605e-01 -8.83746445e-01 -7.88718224e-01 -2.15841964e-01 -7.44397119e-02 2.17353590e-02 -8.79590452e-01 1.59522450e+00 2.76522666e-01 6.34492040e-01 -2.79039532e-01 6.65861845e-01 5.03309488e-01 6.93406641e-01 -1.01310842e-01 -4.20717657e-01 1.17098558e+00 -1.30571127e+00 -1.11842453e+00 -1.84834421e-01 5.18602788e-01 -7.24029899e-01 1.15186715e+00 5.17914057e-01 -7.75343657e-01 -6.65488124e-01 -1.27385104e+00 1.47035951e-02 -7.37789571e-02 1.58622310e-01 9.15144801e-01 1.09282267e+00 -6.15713358e-01 5.32414079e-01 -9.53509212e-01 1.22078046e-01 6.82163715e-01 3.78277630e-01 -2.45212410e-02 1.55353427e-01 -1.38743913e+00 8.11417878e-01 2.58286923e-01 -1.66675895e-02 -6.83689773e-01 -1.04571068e+00 -6.48447573e-01 -2.42746792e-05 4.58515495e-01 -3.38292211e-01 1.19703746e+00 -6.97657883e-01 -1.92271614e+00 1.30239755e-01 -8.52719247e-02 -9.94300008e-01 5.39738357e-01 -4.88362759e-01 -8.56655896e-01 2.94861019e-01 -3.28755587e-01 6.32814586e-01 1.31402099e+00 -5.17609417e-01 -7.14708745e-01 1.51417255e-01 -3.98931235e-01 8.83313417e-02 -7.66637862e-01 -3.01858813e-01 -5.97723126e-01 -1.02248883e+00 -9.79856849e-02 -7.07042515e-01 2.03224383e-02 -8.65428224e-02 -2.42350727e-01 1.46924868e-01 1.43411446e+00 -6.18671358e-01 1.50179660e+00 -2.24518800e+00 -4.30043966e-01 3.14635187e-01 1.87888946e-02 4.23625350e-01 -2.95867205e-01 1.75599381e-01 -1.48285806e-01 -1.83220327e-01 -4.26063016e-02 -3.16277266e-01 -1.96810901e-01 2.42880136e-01 -4.93046910e-01 7.31408596e-01 1.08400017e-01 7.68717766e-01 -3.50854605e-01 -4.40474123e-01 4.21244353e-01 6.25900745e-01 -8.30674410e-01 6.65906593e-02 6.61939234e-02 3.79573882e-01 -4.66924533e-02 4.29741561e-01 6.00658953e-01 -1.45488411e-01 -7.20529109e-02 -5.20782471e-01 8.77614990e-02 6.67832851e-01 -1.07464612e+00 1.64039981e+00 -8.31529498e-01 6.76160336e-01 -1.88455373e-01 -1.20422602e+00 5.51922977e-01 3.10142457e-01 4.32904810e-01 -9.07011151e-01 2.82530844e-01 -2.01433510e-01 -1.21224098e-01 -1.98172256e-01 3.65368515e-01 5.49133420e-02 2.11195685e-02 2.64557242e-01 3.44136804e-01 5.18563032e-01 -1.59131974e-01 1.50109142e-01 9.33160007e-01 -5.08089252e-02 4.35951591e-01 -3.78014088e-01 5.92282236e-01 -5.37264347e-01 4.89881486e-01 8.55629206e-01 -1.34198055e-01 -1.54586136e-02 1.65458918e-02 -3.55511904e-02 -5.15243053e-01 -1.05283272e+00 -4.48433965e-01 1.38741589e+00 1.93835590e-02 -5.24781048e-01 -4.35021609e-01 -5.94831467e-01 7.83136487e-03 7.79243827e-01 -2.78448910e-01 -5.77183843e-01 -4.06255931e-01 -1.01276338e+00 7.74735391e-01 4.18635696e-01 8.18661034e-01 -1.71384335e-01 -3.54879051e-01 2.65492439e-01 -5.37233531e-01 -1.41356599e+00 -8.12586367e-01 3.73535931e-01 -1.00798500e+00 -5.40428698e-01 -4.17113602e-02 -2.62768537e-01 1.20843641e-01 3.79911035e-01 5.09709120e-01 -3.75196338e-01 -3.21895719e-01 -2.39132773e-02 -3.06555152e-01 -3.10240239e-01 -1.28868192e-01 2.08321407e-01 2.55204201e-01 3.33728611e-01 9.10675153e-02 -7.58941352e-01 -5.57282150e-01 4.97177541e-01 -5.82106948e-01 -1.30787328e-01 6.21789217e-01 1.05676162e+00 3.25409949e-01 2.34788135e-01 6.30769610e-01 -7.44705677e-01 4.75795209e-01 -3.19749743e-01 -6.15255594e-01 -8.51032212e-02 -9.94293571e-01 3.23349237e-01 6.26619160e-01 -8.88056934e-01 -1.26589572e+00 -3.54603112e-01 5.89795932e-02 -6.53442502e-01 4.69990879e-01 3.15822631e-01 -1.35386914e-01 -5.27858317e-01 6.97073460e-01 2.58287251e-01 1.17320754e-02 -4.68730509e-01 5.30407369e-01 8.20108533e-01 6.75655067e-01 -2.78319627e-01 9.44474578e-01 5.03309190e-01 3.28363031e-02 -8.85540128e-01 -9.20936644e-01 -4.08215582e-01 -2.69129366e-01 -1.57951519e-01 6.65677562e-02 -1.19846857e+00 -6.68918788e-01 2.57208794e-01 -6.87727332e-01 -2.04621494e-01 -3.76076102e-01 5.92519701e-01 -4.17240232e-01 4.04611528e-01 -7.68837214e-01 -9.13914323e-01 -6.07577085e-01 -7.27035284e-01 8.75390887e-01 -1.62468374e-01 -1.45099983e-01 -7.27972984e-01 -3.58453631e-01 3.37736905e-01 7.82682776e-01 -2.12374851e-01 6.14699066e-01 -4.99355376e-01 -3.65023583e-01 -1.52889982e-01 -1.33395359e-01 4.58595693e-01 1.28941059e-01 -5.22409320e-01 -1.24562216e+00 -4.75851417e-01 1.50038600e-01 -2.52825409e-01 1.00802636e+00 3.76336843e-01 1.42845035e+00 -5.82793832e-01 -3.82960320e-01 9.37486947e-01 1.10809720e+00 7.74277374e-02 5.79604030e-01 2.04862535e-01 3.59033287e-01 1.39130682e-01 6.44217491e-01 5.94999313e-01 9.78124607e-03 9.07647431e-01 8.44705030e-02 -2.27075204e-01 -1.83926955e-01 -3.39814037e-01 6.72020853e-01 9.45929527e-01 2.63150752e-01 -1.88774324e-03 -3.51172328e-01 2.52399385e-01 -1.61633027e+00 -1.04743457e+00 1.65069640e-01 1.97921002e+00 9.88413155e-01 3.45197678e-01 2.70460099e-01 5.77559948e-01 4.44410086e-01 2.11743563e-01 -4.85442758e-01 -3.26863557e-01 4.45259139e-02 5.64690113e-01 8.22176993e-01 6.36876285e-01 -1.31606162e+00 9.57436800e-01 7.43199825e+00 1.36090624e+00 -1.17654324e+00 5.53593218e-01 5.82796633e-01 -5.76122105e-01 3.01424056e-01 -3.90742540e-01 -8.17578852e-01 4.42056865e-01 1.53766024e+00 -2.31514260e-01 6.48989916e-01 8.14919829e-01 3.57321531e-01 -2.60269139e-02 -9.56807435e-01 1.09409976e+00 7.08765760e-02 -1.21536517e+00 -3.73437232e-03 -6.81814551e-02 2.82801896e-01 8.39221254e-02 3.41764838e-01 5.14760792e-01 -9.36824176e-03 -8.15896749e-01 6.04313791e-01 2.25760058e-01 9.61070836e-01 -1.06242251e+00 3.91623855e-01 2.83899635e-01 -1.38935411e+00 -4.00858700e-01 -2.49209538e-01 -8.67940485e-02 -3.75859737e-02 8.04638088e-01 -7.99532235e-01 4.75622147e-01 5.73272765e-01 3.67737055e-01 -4.69531775e-01 7.60985553e-01 -8.34605694e-02 8.74610543e-01 -5.36625683e-01 2.64982671e-01 3.73766311e-02 1.12543792e-01 5.28039098e-01 1.33814931e+00 -4.25985642e-02 -1.21917456e-01 2.40653232e-01 3.09960067e-01 -5.53309582e-02 -8.38928223e-02 -2.02072546e-01 2.12787703e-01 8.77605736e-01 1.06495333e+00 -1.72063842e-01 -1.36305898e-01 -5.00628293e-01 1.03710556e+00 1.63428634e-01 3.09711576e-01 -1.33723533e+00 -4.36516881e-01 7.06089020e-01 3.72341052e-02 3.79069924e-01 -7.79726356e-02 -8.74823630e-02 -1.00067282e+00 -1.64849445e-01 -8.84296715e-01 6.78996623e-01 -2.80499607e-01 -9.56407547e-01 3.92726898e-01 -7.15034008e-02 -1.36850512e+00 -1.31615773e-01 -9.91713628e-02 -1.49136946e-01 5.81165195e-01 -1.63183403e+00 -1.07400537e+00 -6.55529499e-02 7.77249575e-01 5.83257616e-01 -2.41965607e-01 7.57916749e-01 6.31215751e-01 -6.64746225e-01 1.38104105e+00 7.92095251e-03 -2.52802223e-01 7.50606418e-01 -5.48285246e-01 1.71958968e-01 1.08752441e+00 1.51286453e-01 6.10296190e-01 6.35684788e-01 -1.65539280e-01 -1.26732719e+00 -1.58156061e+00 1.97713584e-01 2.02802505e-04 6.19343221e-01 -5.72685599e-01 -6.63391173e-01 5.38171768e-01 -3.34362276e-02 1.49744153e-01 7.94679582e-01 1.98640496e-01 -6.41445577e-01 -4.82817173e-01 -1.13616526e+00 4.73514795e-01 1.09690475e+00 -6.87131047e-01 -4.51952398e-01 1.82742044e-01 8.16601515e-01 -4.03425604e-01 -7.73920000e-01 4.04720515e-01 5.90910733e-01 -6.25078261e-01 1.05909801e+00 -4.38754082e-01 -1.65118083e-01 -1.13641948e-01 -2.08685890e-01 -1.15179861e+00 -4.39973414e-01 -1.13831782e+00 -7.14386940e-01 1.08130181e+00 3.53859544e-01 -6.63452446e-01 8.10169756e-01 -3.76774726e-04 -5.36550432e-02 -6.13903224e-01 -1.34596443e+00 -1.13477409e+00 -4.02974755e-01 -7.02016711e-01 2.74402410e-01 8.21861923e-01 2.84530997e-01 6.64390802e-01 -8.91165495e-01 3.34368199e-01 6.66662693e-01 -1.07015990e-01 5.71360826e-01 -9.12515640e-01 -5.42120993e-01 -1.35939345e-01 -3.78292769e-01 -1.31994128e+00 -5.13078682e-02 -9.93086338e-01 -3.03537160e-01 -6.09734297e-01 -6.33770484e-04 -7.10745901e-02 -5.71179152e-01 6.32058740e-01 -7.47566521e-02 3.55704367e-01 -7.49819130e-02 3.48274894e-02 -6.10300243e-01 7.16386497e-01 7.19887972e-01 -6.50864169e-02 -4.11968112e-01 -1.59306720e-01 -6.90636158e-01 5.62005818e-01 7.41665065e-01 -4.55164135e-01 -6.37770534e-01 -1.91446409e-01 -6.51475787e-01 1.38352690e-02 6.11329898e-02 -1.22241700e+00 1.63650379e-01 -9.26983450e-03 4.32258725e-01 -5.70845723e-01 5.32280564e-01 -7.81795442e-01 -4.98182178e-02 6.52546346e-01 -3.56053561e-01 -3.17777246e-01 5.63720942e-01 8.27638924e-01 2.79935566e-03 4.29047704e-01 1.29902530e+00 6.00928605e-01 -4.89355534e-01 3.63523453e-01 -5.81191659e-01 -1.18045472e-01 8.23334038e-01 -4.47582044e-02 -3.56474370e-01 -4.98444825e-01 -4.62086380e-01 1.90247998e-01 -3.37311774e-01 4.96677339e-01 4.72234160e-01 -1.46419442e+00 -2.55249500e-01 4.44429159e-01 7.84421265e-02 -6.14912331e-01 2.34752938e-01 1.15627277e+00 1.30000412e-01 7.82437980e-01 3.83640602e-02 -5.99771976e-01 -1.23672485e+00 4.47611421e-01 4.68802184e-01 -4.85934645e-01 -6.50320292e-01 1.01764452e+00 3.94219644e-02 -1.78839564e-02 7.13223636e-01 -1.60343915e-01 1.99840069e-02 -8.62518623e-02 7.16761470e-01 5.70621133e-01 4.04553831e-01 -2.81635582e-01 -4.52061117e-01 1.97940134e-02 -3.90065581e-01 -1.82963088e-01 1.17989850e+00 -2.12278068e-01 3.71942848e-01 3.62964779e-01 1.31448758e+00 -6.76245913e-02 -1.03888595e+00 -5.89991987e-01 -1.12814754e-02 -5.43059409e-01 7.73530900e-01 -7.52710640e-01 -1.12205172e+00 6.09321594e-01 1.01450729e+00 -4.17711824e-01 1.48057806e+00 -5.19172966e-01 9.77085292e-01 4.51967537e-01 4.43143576e-01 -1.24954617e+00 1.93100914e-01 4.36433345e-01 6.19945824e-01 -6.96324646e-01 3.38585138e-01 -7.13987172e-01 -3.51318598e-01 8.64018738e-01 4.50121194e-01 1.11446559e-01 9.40100312e-01 7.80863941e-01 -1.96502045e-01 2.57258415e-01 -1.11713612e+00 -1.61566306e-02 3.91398758e-01 5.17982781e-01 2.37785995e-01 -2.23542616e-01 -4.97700691e-01 6.46523654e-01 -1.34972766e-01 -2.71256524e-03 5.07995300e-02 8.07187676e-01 -3.72459233e-01 -1.02775192e+00 -3.69214088e-01 7.70710111e-01 -6.15914941e-01 -1.90521166e-01 2.97803455e-03 8.11679542e-01 -1.48090780e-01 1.20270884e+00 1.49781421e-01 -7.85429716e-01 3.82614732e-01 1.79042101e-01 5.61100483e-01 -2.53615946e-01 -4.94725615e-01 5.07785797e-01 3.57904702e-01 -9.37913299e-01 -4.06047821e-01 -3.54052335e-01 -9.54215169e-01 -4.98448581e-01 -8.78005624e-01 4.39974703e-02 3.20302367e-01 9.47681367e-01 6.02219880e-01 8.38764787e-01 1.00801432e+00 -5.72117925e-01 -8.43866289e-01 -1.21923316e+00 -3.63396078e-01 2.81464569e-02 3.94232541e-01 -9.42371964e-01 -4.41324621e-01 -8.07664841e-02]
[14.556925773620605, 6.064140796661377]
95c9cbfc-5a1a-4866-be0a-80d35e9f35e0
continual-active-learning-for-efficient
2106.03351
null
https://arxiv.org/abs/2106.03351v1
https://arxiv.org/pdf/2106.03351v1.pdf
Continual Active Learning for Efficient Adaptation of Machine Learning Models to Changing Image Acquisition
Imaging in clinical routine is subject to changing scanner protocols, hardware, or policies in a typically heterogeneous set of acquisition hardware. Accuracy and reliability of deep learning models suffer from those changes as data and targets become inconsistent with their initial static training set. Continual learning can adapt to a continuous data stream of a changing imaging environment. Here, we propose a method for continual active learning on a data stream of medical images. It recognizes shifts or additions of new imaging sources - domains -, adapts training accordingly, and selects optimal examples for labelling. Model training has to cope with a limited labelling budget, resembling typical real world scenarios. We demonstrate our method on T1-weighted magnetic resonance images from three different scanners with the task of brain age estimation. Results demonstrate that the proposed method outperforms naive active learning while requiring less manual labelling.
['Georg Langs', 'Johannes Hofmanninger', 'Matthias Perkonigg']
2021-06-07
null
null
null
null
['age-estimation', 'age-estimation']
['computer-vision', 'miscellaneous']
[ 6.83704674e-01 2.62564570e-01 -4.94122952e-01 -8.16134155e-01 -9.57762122e-01 -2.77731687e-01 3.72436672e-01 4.38472748e-01 -1.04234755e+00 7.14639604e-01 -8.28229189e-02 -1.13144696e-01 -3.21326554e-01 -2.36317724e-01 -6.06039703e-01 -7.22187519e-01 -7.15276599e-01 1.16177404e+00 5.77135980e-01 4.21850473e-01 1.59624219e-01 5.77287376e-01 -1.01336539e+00 2.01428488e-01 4.82716560e-01 8.73029411e-01 4.75944161e-01 7.41387248e-01 -7.23888800e-02 8.74695480e-01 -5.59300721e-01 5.92134967e-02 3.77522141e-01 4.98542972e-02 -8.88362169e-01 5.71392715e-01 3.11632156e-01 -3.87996823e-01 -9.12919864e-02 8.50210905e-01 1.02693653e+00 4.57690880e-02 3.78311634e-01 -9.85125303e-01 -1.42573178e-01 7.41571307e-01 -5.08137703e-01 9.97275472e-01 -1.67960376e-01 5.18418074e-01 -7.45563302e-03 -6.77114785e-01 6.64144039e-01 5.71471691e-01 8.02820027e-01 8.18098545e-01 -1.31438816e+00 -6.76797211e-01 2.58074880e-01 4.98938233e-01 -9.81558561e-01 -1.01949561e+00 4.96735513e-01 -7.08782852e-01 8.38532388e-01 -1.05105955e-02 7.22638547e-01 1.08165383e+00 2.70628691e-01 6.28171384e-01 9.36684549e-01 -5.67805767e-01 6.92928493e-01 3.24632456e-05 2.52754629e-01 6.23651445e-01 7.56008103e-02 1.45209078e-02 -8.11798930e-01 -4.14663166e-01 6.65151358e-01 -2.06590131e-01 -2.48943806e-01 -8.60682249e-01 -1.39129698e+00 5.20900488e-01 4.93279397e-02 4.06392604e-01 -7.85307109e-01 -1.19616792e-01 6.29528403e-01 3.92745882e-01 4.84005868e-01 5.19892991e-01 -6.42281055e-01 8.56921077e-02 -1.15185213e+00 -3.06868464e-01 2.95886576e-01 8.87026191e-01 4.08901930e-01 5.44784479e-02 4.31846678e-02 8.19248915e-01 1.63753584e-01 1.35517687e-01 1.07518733e+00 -6.66503727e-01 5.95701300e-02 1.69737682e-01 -1.07617036e-01 -1.53697625e-01 -7.88348079e-01 -5.81590772e-01 -5.45072019e-01 3.26184124e-01 2.34742165e-01 -2.05628708e-01 -1.31055832e+00 1.22799051e+00 3.75189155e-01 1.58058360e-01 -6.50567889e-01 4.70464081e-01 4.26990956e-01 -1.54567778e-01 5.37251711e-01 -9.16982532e-01 7.93310165e-01 -6.41912162e-01 -8.60197186e-01 -5.62671900e-01 6.78092420e-01 -3.55738074e-01 7.01818585e-01 4.98382390e-01 -1.27186418e+00 -5.51591694e-01 -1.03016114e+00 6.96179807e-01 -7.85241276e-02 -4.88215297e-01 6.86421692e-01 6.98260784e-01 -1.21050990e+00 3.32875669e-01 -1.31673777e+00 -1.17172442e-01 6.44285202e-01 9.19573665e-01 -2.50931025e-01 1.06738113e-01 -1.02828884e+00 1.02451682e+00 7.53712237e-01 6.08721972e-02 -9.79527950e-01 -1.13386631e+00 -6.55164897e-01 -2.86220551e-01 6.68912351e-01 -4.26035643e-01 1.66999400e+00 -1.27681935e+00 -1.28539300e+00 1.02596104e+00 4.31046993e-01 -7.20614552e-01 6.92859590e-01 1.67156577e-01 -6.56163156e-01 1.82721540e-01 -7.00272322e-02 7.29738355e-01 1.14920449e+00 -9.45191801e-01 -4.32514012e-01 -5.26259661e-01 -3.91011298e-01 3.23449403e-01 -1.52229518e-01 1.04498953e-01 -1.08595610e-01 -1.91612422e-01 1.89625561e-01 -8.50342333e-01 -5.96615136e-01 8.38092808e-03 8.84464830e-02 2.52978325e-01 6.24532402e-01 -2.70248562e-01 9.88332212e-01 -2.02754617e+00 -2.31185600e-01 1.70464069e-01 4.50728744e-01 1.08703636e-01 2.42532089e-01 -4.00334090e-01 -6.15955234e-01 -3.31654996e-01 -2.59926140e-01 -6.16895258e-02 -4.25546557e-01 2.25725353e-01 1.18105926e-01 6.03113830e-01 -1.45571366e-01 6.01309419e-01 -1.03476250e+00 -8.99475634e-01 1.02704048e-01 -2.18888056e-02 -8.12105089e-02 2.11333230e-01 3.30704800e-03 9.20287192e-01 -5.99063188e-02 5.95144629e-01 3.41876119e-01 -4.96444702e-01 5.60470641e-01 2.71946304e-02 1.16573751e-01 -1.58295915e-01 -8.09243143e-01 1.88384771e+00 -2.38996983e-01 3.76045734e-01 -1.20585263e-01 -1.39274752e+00 8.41879904e-01 6.04854703e-01 1.15328860e+00 -1.03618407e+00 9.41207781e-02 3.22763354e-01 5.10185361e-01 -8.08762848e-01 3.98698309e-03 -1.36228442e-01 2.65858799e-01 5.26401043e-01 3.54979873e-01 2.31769770e-01 6.42727986e-02 5.03778905e-02 1.44634211e+00 -2.63711065e-01 4.97209996e-01 -2.17306823e-01 1.38242409e-01 4.00414988e-02 6.10411346e-01 1.14964640e+00 -8.27811480e-01 3.53019029e-01 -6.74065351e-02 -7.86334097e-01 -1.10894251e+00 -9.61075187e-01 -4.59023297e-01 1.20566463e+00 -3.89424384e-01 3.62452179e-01 -4.68172699e-01 -1.07364416e+00 -4.69197690e-01 4.27671432e-01 -7.09666908e-01 -1.67985365e-01 -1.01140046e+00 -1.38582098e+00 7.04130977e-02 2.70065248e-01 2.19614208e-01 -1.30824208e+00 -1.45014691e+00 5.17272234e-01 2.23408058e-01 -8.19237947e-01 -1.99657708e-01 7.30640113e-01 -1.31231749e+00 -1.11777055e+00 -8.22093666e-01 -7.90079415e-01 8.07663143e-01 -2.92243809e-01 1.35325134e+00 -1.69709008e-02 -5.31988204e-01 6.93695843e-01 -2.96487987e-01 -6.51457906e-01 -5.34285605e-01 3.99864525e-01 -3.01645063e-02 -1.06308937e-01 1.81975499e-01 -3.82082999e-01 -7.99088776e-01 1.07181810e-01 -8.33257794e-01 1.85037255e-02 6.44164562e-01 7.98157632e-01 7.26193190e-01 -2.96170805e-02 6.67360306e-01 -1.45692980e+00 4.34803367e-01 -6.81316495e-01 -4.69713211e-01 5.33187449e-01 -9.94145989e-01 -1.69734553e-01 5.16556390e-02 -9.28019762e-01 -1.04217970e+00 5.20041406e-01 2.07780272e-01 -1.37676403e-01 -1.32147938e-01 3.07646275e-01 2.36000106e-01 -1.33981019e-01 1.09723520e+00 -1.33862495e-02 -7.24085467e-03 -1.86031550e-01 -8.31257701e-02 4.04118121e-01 7.64480591e-01 -3.06620359e-01 2.72714019e-01 3.28199863e-01 -3.61569166e-01 -3.81593347e-01 -5.26733696e-01 -3.95705491e-01 -1.22515619e+00 -7.02697575e-01 2.34318376e-01 -4.88691598e-01 -5.80171049e-02 6.39950454e-01 -6.15783036e-01 -9.18021619e-01 -5.81094623e-01 7.63812125e-01 -6.76205039e-01 1.79457963e-01 -2.73716718e-01 -4.93283212e-01 -5.21262348e-01 -1.42830503e+00 6.12383723e-01 -3.38889360e-02 -3.94176126e-01 -1.30110371e+00 3.51954281e-01 1.01891451e-01 6.44637585e-01 9.58821550e-02 1.03021622e+00 -8.61288190e-01 -3.34536314e-01 -2.12791357e-02 4.52630371e-01 -1.31406352e-01 3.30721945e-01 -4.84974444e-01 -8.50925922e-01 -5.98912716e-01 4.20313656e-01 -2.73671299e-01 4.54229176e-01 7.40521371e-01 1.48513317e+00 1.56150116e-02 -2.91703373e-01 4.01432842e-01 1.09476233e+00 7.99605429e-01 4.67656374e-01 6.91844940e-01 1.34187758e-01 4.85335797e-01 4.84872967e-01 3.63127798e-01 -1.49395064e-01 4.26354349e-01 2.36272305e-01 -4.53288615e-01 -1.03705600e-01 7.04369485e-01 -1.31715257e-02 9.24532771e-01 2.00642154e-01 2.89681703e-01 -1.66527390e+00 5.38406730e-01 -1.78205287e+00 -7.66414821e-01 2.97775328e-01 2.41596746e+00 1.08269954e+00 6.85630918e-01 1.71293154e-01 6.09258339e-02 9.10847783e-01 -1.58802867e-01 -1.35127795e+00 6.69048950e-02 3.12039912e-01 3.37467909e-01 1.00297940e+00 1.69426933e-01 -1.05855668e+00 2.89756000e-01 7.73224878e+00 2.45015919e-01 -1.57488275e+00 7.60251641e-01 8.22506487e-01 -3.38673443e-01 2.78431326e-01 -2.52040386e-01 -1.50448397e-01 5.22089124e-01 1.15390027e+00 -2.54847463e-02 -2.36852188e-02 7.20638812e-01 4.10514399e-02 -3.22338045e-01 -1.30123639e+00 1.09441829e+00 7.00234026e-02 -1.35263920e+00 -5.12317240e-01 -9.74296629e-02 3.83644342e-01 3.60629231e-01 2.04983100e-01 1.49986133e-01 2.38034606e-01 -6.86250746e-01 7.64727652e-01 6.30229414e-01 1.03201604e+00 -2.52549797e-01 4.82925057e-01 4.21817690e-01 -3.79851192e-01 -2.30424672e-01 -3.03863287e-02 5.93930960e-01 2.52810508e-01 4.12817895e-01 -1.40238154e+00 8.60739592e-03 7.43934333e-01 2.54526287e-01 -8.17728162e-01 1.40283895e+00 3.70945603e-01 6.81695521e-01 7.42861405e-02 6.70392156e-01 -1.01523206e-01 5.08640587e-01 3.56268466e-01 1.13134515e+00 -6.80299383e-03 5.45532703e-02 1.90045327e-01 2.05731895e-02 2.18460560e-01 1.56965986e-01 -2.76257604e-01 3.14407051e-01 5.13087690e-01 9.60976303e-01 -1.31607509e+00 -5.96459329e-01 -2.96908826e-01 7.00925291e-01 2.56678045e-01 1.25178218e-01 -6.75362706e-01 2.91936249e-01 -3.80577892e-01 4.28516209e-01 -1.28936648e-01 -1.94573358e-01 -1.96512774e-01 -8.47917736e-01 -4.59346771e-01 -9.63303089e-01 8.42866600e-01 -6.39647901e-01 -1.24149632e+00 5.92818618e-01 3.86058569e-01 -1.09411788e+00 -2.22603753e-01 -3.14482421e-01 -2.77206093e-01 4.27662492e-01 -1.10907090e+00 -5.15064240e-01 -1.96691960e-01 7.65989423e-01 9.43573117e-01 -7.07355618e-01 7.35773802e-01 4.71621275e-01 -3.53214443e-01 4.73692447e-01 1.33562624e-01 1.16962738e-01 9.70623016e-01 -1.18779194e+00 3.14268827e-01 6.12962008e-01 -1.22920424e-01 4.55249369e-01 6.31780148e-01 -7.87798941e-01 -9.44204807e-01 -7.75626063e-01 3.57213765e-01 -1.53628424e-01 6.80144668e-01 -2.37712905e-01 -1.22457087e+00 9.16354477e-01 1.79498509e-01 2.06191316e-01 9.18759108e-01 3.35775912e-02 9.08687413e-02 -9.55490172e-02 -1.44591200e+00 8.15559253e-02 1.05033398e+00 -1.67648584e-01 -5.50062239e-01 7.20271587e-01 2.30542883e-01 -7.31295109e-01 -8.33399117e-01 5.37116349e-01 4.71916914e-01 -6.66853428e-01 8.06880772e-01 -6.70362830e-01 -4.34471488e-01 3.36072057e-01 4.52965051e-01 -1.26305747e+00 -5.38512647e-01 -5.05830646e-01 -2.47198984e-01 4.22360271e-01 3.53059053e-01 -4.99106079e-01 9.27789569e-01 7.93396354e-01 -3.01437646e-01 -5.96552730e-01 -1.04213870e+00 -4.77955580e-01 -9.99292657e-02 -3.35418731e-01 5.19411385e-01 1.34537530e+00 -1.75882518e-01 2.12698966e-01 -8.44493061e-02 -1.59948214e-03 9.63112772e-01 -5.70762813e-01 1.73245102e-01 -1.33213043e+00 -4.24017817e-01 -7.96440691e-02 -4.56081748e-01 -3.21670115e-01 -2.39585005e-02 -6.80493236e-01 1.92378223e-01 -1.09714794e+00 3.01610917e-01 -9.35429871e-01 -6.11987889e-01 5.91440320e-01 7.85401165e-02 2.09591836e-01 -2.40534872e-01 6.80814266e-01 -8.85653377e-01 -9.68832299e-02 8.75208199e-01 -3.93339008e-01 -2.44170740e-01 8.42137486e-02 -1.92138940e-01 6.63263023e-01 8.78168762e-01 -7.43269742e-01 -8.89263153e-01 -5.73275566e-01 2.75854915e-01 2.10080177e-01 -9.95332301e-02 -1.14873564e+00 5.42958558e-01 -4.02564220e-02 6.52963638e-01 -2.77247190e-01 -1.55199647e-01 -8.90194833e-01 4.91794765e-01 7.31300533e-01 -7.88085938e-01 2.51214802e-01 9.39919129e-02 3.67687404e-01 2.22257391e-01 -5.78092277e-01 1.16588306e+00 -6.67595088e-01 -1.10110712e+00 7.46333897e-01 -6.55548394e-01 1.25142798e-01 8.97957861e-01 -5.00001371e-01 -6.16644807e-02 1.35047570e-01 -1.61377919e+00 1.30866513e-01 9.06702951e-02 3.97050142e-01 4.72603112e-01 -1.06437969e+00 -5.75143337e-01 6.81033880e-02 1.36611342e-01 1.68503016e-01 5.40178418e-01 1.13570380e+00 -4.33499277e-01 -6.11665472e-02 -6.04638398e-01 -9.23388183e-01 -1.37758005e+00 5.25293827e-01 7.44751453e-01 -3.54269445e-01 -7.62572169e-01 5.96841276e-01 -4.99227270e-02 -3.06706518e-01 4.85938400e-01 2.19055682e-01 -3.31818819e-01 2.24346802e-01 5.19255161e-01 6.11324608e-02 7.74689078e-01 -3.80299926e-01 -2.03118205e-01 -1.69894621e-01 -6.37825668e-01 -1.89300239e-01 1.46552467e+00 -3.12698394e-01 3.72242689e-01 8.18360090e-01 7.20013678e-01 -6.40862048e-01 -1.29418218e+00 -7.73778617e-01 4.71016556e-01 -3.95712495e-01 2.62931198e-01 -9.93253410e-01 -1.06742918e+00 5.52759469e-01 1.59938920e+00 -2.51052547e-02 1.03799510e+00 -1.33739989e-02 3.20912778e-01 2.91492313e-01 6.38867974e-01 -1.32790995e+00 3.75971377e-01 1.79339424e-01 5.23256063e-01 -1.27171457e+00 2.93993235e-01 2.29001731e-01 -7.05980599e-01 1.19264948e+00 6.01903379e-01 5.04713356e-01 6.67453468e-01 5.65582752e-01 4.19227749e-01 -4.89451081e-01 -7.03148365e-01 1.80822030e-01 -1.10604718e-01 9.13568020e-01 2.37671033e-01 1.08339209e-02 9.39297751e-02 -1.25320658e-01 1.99379012e-01 1.92950934e-01 5.20909965e-01 1.31792116e+00 -4.98105645e-01 -9.89662170e-01 -5.15084386e-01 7.13189602e-01 -2.46299192e-01 1.86344415e-01 1.34365052e-01 6.57300889e-01 5.39650857e-01 2.96628535e-01 4.77653831e-01 2.22309113e-01 1.76419467e-01 4.39402997e-01 9.85232294e-01 -1.02240396e+00 -5.81034422e-01 7.48963207e-02 -1.78481936e-01 -2.09683761e-01 -8.02777946e-01 -1.09326398e+00 -1.25886905e+00 2.42083162e-01 -2.59962112e-01 -1.66708112e-01 7.05156028e-01 9.03434217e-01 1.06618628e-01 6.36393785e-01 5.70324659e-01 -5.91502070e-01 -4.87597972e-01 -9.42366719e-01 -4.72081780e-01 3.97091717e-01 2.84390748e-01 -8.55668008e-01 -2.42368486e-02 4.66361910e-01]
[14.698555946350098, -2.1386606693267822]
fda8c548-bcbd-43bc-966e-9d6697ce037f
evaluating-active-learning-heuristics-for
1807.03083
null
https://arxiv.org/abs/1807.03083v2
https://arxiv.org/pdf/1807.03083v2.pdf
Evaluating Active Learning Heuristics for Sequential Diagnosis
Given a malfunctioning system, sequential diagnosis aims at identifying the root cause of the failure in terms of abnormally behaving system components. As initial system observations usually do not suffice to deterministically pin down just one explanation of the system's misbehavior, additional system measurements can help to differentiate between possible explanations. The goal is to restrict the space of explanations until there is only one (highly probable) explanation left. To achieve this with a minimal-cost set of measurements, various (active learning) heuristics for selecting the best next measurement have been proposed. We report preliminary results of extensive ongoing experiments with a set of selection heuristics on real-world diagnosis cases. In particular, we try to answer questions such as "Is some heuristic always superior to all others?", "On which factors does the (relative) performance of the particular heuristics depend?" or "Under which circumstances should I use which heuristic?"
['Wolfgang Schmid', 'Patrick Rodler']
2018-07-09
null
null
null
null
['sequential-diagnosis']
['medical']
[ 1.89751893e-01 2.05341622e-01 -2.09643140e-01 -4.02784497e-01 -5.85977912e-01 -4.11927074e-01 4.68368292e-01 5.15898049e-01 8.60157385e-02 7.20997274e-01 -2.01721117e-01 -8.40529263e-01 -5.53530931e-01 -3.51742476e-01 -5.49250059e-02 -7.81591117e-01 -3.12869877e-01 8.02043974e-01 5.05132079e-01 4.82898168e-02 4.21823859e-01 6.33882821e-01 -1.48168600e+00 6.40245378e-02 5.87666929e-01 5.22825122e-01 -3.52459401e-02 8.82254601e-01 1.01572789e-01 9.79244530e-01 -9.23752785e-01 3.45584869e-01 7.96364322e-02 -7.80075908e-01 -9.62061465e-01 8.78102899e-01 -2.93838620e-01 -2.38911703e-01 1.37242883e-01 1.00006568e+00 -1.43244173e-02 2.04793233e-02 3.53885412e-01 -1.56854832e+00 1.74983412e-01 5.14521480e-01 -1.12889536e-01 5.48097968e-01 6.75246894e-01 3.91278327e-01 8.34832311e-01 -3.23538870e-01 1.00648142e-01 9.38153267e-01 1.47150367e-01 4.56629217e-01 -1.51882207e+00 -3.56324792e-01 4.47265506e-01 2.33423516e-01 -1.50281179e+00 -7.55482674e-01 6.97307408e-01 -2.27379814e-01 1.23366141e+00 5.91949105e-01 2.60939449e-01 7.36733317e-01 4.27051783e-01 3.27635556e-01 8.64843309e-01 -5.08926749e-01 5.57530165e-01 2.42651284e-01 2.35769778e-01 7.82324612e-01 5.96471429e-01 1.76900804e-01 -2.05466881e-01 -8.84961545e-01 4.56819862e-01 -5.40968105e-02 -2.23988533e-01 -3.63275856e-01 -9.42204416e-01 4.03598696e-01 -2.33212769e-01 4.11528289e-01 -5.09299517e-01 6.52615800e-02 6.19300716e-02 6.10993028e-01 2.64562387e-02 1.08938074e+00 -7.03522682e-01 -1.31113276e-01 -6.39748037e-01 1.12591624e-01 9.84061241e-01 4.05721664e-01 6.92538977e-01 3.17684919e-01 2.88352102e-01 2.47049257e-01 2.85898685e-01 2.23813117e-01 9.71878767e-02 -9.03927743e-01 1.03096142e-01 6.88836753e-01 5.04349053e-01 -5.04761338e-01 -4.44958657e-01 -1.39956713e-01 -1.35175660e-01 5.06750524e-01 4.38200861e-01 -3.13546091e-01 -6.30743623e-01 1.33545542e+00 1.06258951e-01 2.49389812e-01 -1.03097878e-01 7.55076170e-01 -1.20763116e-01 3.64397019e-01 -1.99133411e-01 -8.35668147e-01 6.63453519e-01 -2.78520584e-01 -4.39827770e-01 -2.68488824e-01 6.70446038e-01 -8.58299255e-01 6.72867239e-01 6.44311547e-01 -9.35181320e-01 -2.76947230e-01 -1.14738381e+00 1.03762817e+00 1.20756291e-01 -2.88527757e-01 4.37243104e-01 5.84508777e-01 -8.23858678e-01 5.08987129e-01 -1.18899941e+00 -2.69697875e-01 -5.83056509e-01 6.58146441e-01 -8.99481848e-02 8.61528665e-02 -8.25332940e-01 9.87628341e-01 3.37223470e-01 2.74689142e-02 -1.11492383e+00 1.12203903e-01 -3.82454008e-01 2.04165250e-01 9.64950144e-01 -4.54587579e-01 1.37112200e+00 -8.72657299e-01 -1.11647916e+00 8.48670006e-02 -2.42984653e-01 -3.34551066e-01 2.03301638e-01 2.17908472e-01 -7.65620053e-01 2.37069398e-01 -6.48729131e-02 -1.81044638e-01 7.84613669e-01 -1.29974496e+00 -7.77376473e-01 -2.77274400e-01 3.92889231e-01 1.69602297e-02 -5.37523581e-03 1.51968688e-01 -5.68893105e-02 9.18532014e-02 5.10952175e-01 -1.13052869e+00 -3.04936498e-01 -4.78204042e-01 -6.52843535e-01 -3.21478307e-01 8.47904921e-01 -1.84339851e-01 1.31750190e+00 -2.00429010e+00 -1.56725228e-01 4.50123131e-01 2.09828481e-01 1.11061642e-02 3.16146404e-01 7.56833434e-01 -3.10708731e-01 7.94114098e-02 5.91753349e-02 5.77367730e-02 -2.41270170e-01 3.96735400e-01 -2.74135619e-01 6.69489086e-01 1.32183224e-01 -1.73507468e-03 -8.29601049e-01 -4.80230123e-01 4.31525856e-01 -2.66665041e-01 -1.71349794e-01 4.59567249e-01 -2.07366824e-01 4.75479603e-01 -7.72906125e-01 7.55589366e-01 -1.49923991e-02 -5.19161761e-01 3.99655372e-01 4.21772242e-01 5.21494262e-02 4.13788259e-01 -1.27371407e+00 3.03891391e-01 -2.65356421e-01 3.38371992e-01 -1.92175210e-02 -1.03272355e+00 7.80460119e-01 7.58453190e-01 6.35712087e-01 -2.81025678e-01 1.66829959e-01 2.60237426e-01 5.42553306e-01 -6.34031236e-01 8.62870459e-03 -9.13404003e-02 8.79466534e-03 7.12968230e-01 -3.29102874e-01 3.40460062e-01 1.49031579e-01 -7.23971128e-02 1.84524071e+00 -5.23813784e-01 5.66542089e-01 -3.18543613e-01 5.35393000e-01 1.17164887e-01 7.06074238e-01 8.57050002e-01 -4.20605451e-01 3.07726473e-01 8.46151292e-01 -3.55782598e-01 -7.40985930e-01 -9.09549117e-01 1.40144274e-01 5.00881374e-01 4.19253081e-01 -3.62629175e-01 -4.27469611e-01 -9.89636362e-01 -8.65111426e-02 1.15188682e+00 -3.27348053e-01 -2.40630805e-01 -3.43396991e-01 -7.00322270e-01 -9.91515256e-03 3.22130859e-01 1.29713714e-01 -1.02412486e+00 -7.98360944e-01 3.77220631e-01 8.36114362e-02 -7.72847176e-01 1.22447968e-01 6.63272917e-01 -1.00045609e+00 -1.60159981e+00 2.42193744e-01 -5.09737395e-02 1.22470200e+00 3.76378864e-01 9.17911470e-01 7.11434066e-01 9.28783673e-04 3.62167031e-01 -3.35438579e-01 -1.01668395e-01 -7.67975330e-01 -1.32332250e-01 2.89276272e-01 -2.98299760e-01 3.01044375e-01 -2.42382616e-01 -2.24001050e-01 6.12730443e-01 -6.47824645e-01 -5.27077556e-01 6.22800291e-01 6.56385303e-01 2.04133660e-01 9.47193086e-01 4.18841332e-01 -1.16292799e+00 9.41043198e-01 -6.52321637e-01 -7.13717222e-01 4.94375437e-01 -1.22414219e+00 3.11373115e-01 9.45684135e-01 -4.32870418e-01 -8.49799633e-01 6.58954084e-02 -4.00445573e-02 -3.72155547e-01 -6.86882317e-01 3.71320486e-01 -2.39320606e-01 3.59245062e-01 6.49761498e-01 2.08879918e-01 -2.04049945e-01 -2.72744536e-01 -3.96071911e-01 3.22978735e-01 3.07309747e-01 -4.59570378e-01 5.93493879e-01 6.30994141e-02 -8.68001357e-02 -5.42457044e-01 -5.02812743e-01 -5.61731577e-01 -2.32290670e-01 -1.90368608e-01 2.41067559e-01 -1.45470470e-01 -8.66556346e-01 -1.50586978e-01 -8.43986571e-01 -2.46820971e-01 6.28126487e-02 2.99654633e-01 -4.78881866e-01 3.02839428e-01 -1.64306015e-01 -1.37890041e+00 1.51543155e-01 -1.36891496e+00 6.82269633e-01 1.34111553e-01 -9.68641102e-01 -1.01745689e+00 6.71151057e-02 1.76865652e-01 9.67022032e-02 -7.11679682e-02 1.11693716e+00 -1.16425538e+00 -4.22421157e-01 -5.25858760e-01 3.28029215e-01 1.17885932e-01 4.59351748e-01 2.41994634e-01 -6.25487447e-01 -6.54837430e-01 2.52222717e-01 2.68109500e-01 1.27957314e-01 1.16857633e-01 9.36485767e-01 -3.64828855e-01 -5.89240491e-01 -8.04340169e-02 1.31507313e+00 9.97379363e-01 3.49336952e-01 2.82880872e-01 2.27588549e-01 3.91705215e-01 7.56112516e-01 3.93095791e-01 -3.79234292e-02 5.82311094e-01 5.16462266e-01 3.58437374e-02 2.86787838e-01 2.58094579e-01 4.41594720e-01 3.46423894e-01 2.53671288e-01 -5.74079454e-01 -1.12590337e+00 4.94386137e-01 -1.92028415e+00 -7.96944499e-01 6.70763105e-02 2.62424564e+00 1.09503657e-01 9.70542729e-01 1.70112312e-01 5.56363344e-01 6.39827251e-01 -1.08548932e-01 -8.84284317e-01 -4.43049729e-01 4.25457865e-01 -4.28877622e-01 4.46055055e-01 6.76369786e-01 -7.13981688e-01 2.03777894e-01 6.90299749e+00 2.24292740e-01 -1.14346004e+00 -2.67022669e-01 3.06640416e-01 6.30496070e-02 -1.32490084e-01 7.51755714e-01 -7.83234596e-01 5.42467237e-01 1.38126254e+00 -2.32893392e-01 1.29151613e-01 6.78029835e-01 4.60746616e-01 -5.26433408e-01 -1.35739672e+00 3.66943151e-01 -2.55530894e-01 -8.15321743e-01 -4.70811427e-01 2.57805914e-01 4.00999635e-01 -3.43617827e-01 -4.00776893e-01 9.13942903e-02 4.20853198e-01 -5.54631591e-01 3.91509354e-01 4.19348121e-01 4.64631990e-02 -8.49382043e-01 8.49179268e-01 7.76204705e-01 -6.86085939e-01 -4.59590912e-01 8.84006545e-02 -2.66211212e-01 8.78186449e-02 4.78076011e-01 -1.38592696e+00 3.40324879e-01 3.08258772e-01 -1.73140913e-01 -4.79660183e-01 9.76465344e-01 -5.39208531e-01 9.61838484e-01 -3.72086614e-01 -5.14805205e-02 1.48987863e-02 3.62971365e-01 7.86171913e-01 7.11820900e-01 7.61588588e-02 2.29558796e-01 5.68277061e-01 4.55161303e-01 8.15040767e-01 -3.69106054e-01 -5.57138741e-01 -1.98289573e-01 8.71579945e-01 8.99511218e-01 -1.31144929e+00 -5.19171834e-01 -3.73382241e-01 5.72149932e-01 -1.70741528e-01 3.37167531e-01 -5.26379645e-01 -1.78274542e-01 4.08976287e-01 4.86221462e-01 2.02092268e-02 -2.69043505e-01 -2.28729248e-01 -9.55580175e-01 -1.59235328e-01 -9.57462728e-01 6.00797296e-01 -5.93970358e-01 -9.32722211e-01 7.19123185e-01 3.20412576e-01 -1.24885356e+00 -1.02871609e+00 -7.19993562e-02 -7.27530122e-01 6.49737656e-01 -6.73004866e-01 -1.11949898e-01 1.73503533e-01 3.43493193e-01 6.72051370e-01 -2.87196845e-01 9.77567732e-01 -1.64870217e-01 -6.80739462e-01 1.10181212e-01 -1.63906544e-01 -3.52658689e-01 6.35458946e-01 -1.32387769e+00 2.10384414e-01 1.21080363e+00 1.34488046e-01 6.42947733e-01 1.61541069e+00 -8.16441178e-01 -1.29806638e+00 -3.84715945e-01 9.45108950e-01 -4.50815111e-01 7.17556596e-01 1.05018541e-01 -1.09460509e+00 7.24746227e-01 -8.76043737e-02 -3.25734317e-01 5.34152925e-01 4.28640336e-01 4.67400439e-02 -2.27889605e-02 -1.17996073e+00 4.35186505e-01 4.20559019e-01 -2.99303263e-01 -5.68802178e-01 2.62176991e-01 1.73860326e-01 5.24403192e-02 -5.28774023e-01 4.11961585e-01 8.18092078e-02 -1.22647715e+00 5.02808928e-01 -5.81203580e-01 -2.10514124e-02 -5.60889840e-01 -3.13185640e-02 -1.20211756e+00 -5.07209778e-01 -5.12558818e-01 -2.02196524e-01 9.28533018e-01 7.14325130e-01 -8.44120860e-01 8.82020354e-01 1.02714789e+00 -1.17891923e-01 -8.72193873e-01 -6.44373417e-01 -9.54322815e-01 -5.87735415e-01 -4.86516505e-01 5.82351625e-01 9.25722539e-01 3.76605093e-01 3.35672170e-01 -2.41648555e-01 6.43961430e-01 5.38293481e-01 2.93918252e-01 5.63795865e-01 -1.19078863e+00 -7.19900906e-01 -2.10370734e-01 -3.81735772e-01 -5.50156534e-01 1.20710460e-02 -1.15515895e-01 3.04114401e-01 -1.07046247e+00 4.55402620e-02 -4.44261074e-01 -4.85646188e-01 7.59373605e-01 -2.70289093e-01 -4.36971992e-01 -5.42887412e-02 4.22913909e-01 -7.17033148e-01 -1.42087191e-01 3.71179670e-01 1.71519935e-01 -2.90275693e-01 5.74957192e-01 -6.46504939e-01 6.69554532e-01 8.22362185e-01 -5.19471049e-01 -7.75051355e-01 -7.82962590e-02 8.65240693e-02 1.04120159e+00 9.24448073e-02 -8.72079909e-01 4.04563785e-01 -6.32074118e-01 1.28270388e-01 -3.31821471e-01 1.90742344e-01 -1.10208750e+00 4.72263992e-01 7.35075355e-01 -4.33948427e-01 3.35881889e-01 1.02249250e-01 4.71722424e-01 -9.23174173e-02 -5.91049254e-01 7.33489037e-01 -1.29522622e-01 -8.31134558e-01 -1.04209401e-01 -1.01049352e+00 -5.76250255e-01 1.17272305e+00 -2.93624908e-01 -6.93713082e-03 -5.12327313e-01 -1.02084267e+00 2.88894355e-01 5.55803597e-01 3.17305267e-01 3.62769514e-01 -7.41410673e-01 -1.68375671e-01 4.95844893e-02 2.08393723e-01 -5.05932033e-01 -2.05762863e-01 7.76882589e-01 -1.78561494e-01 6.08238339e-01 1.08569451e-01 -4.39450473e-01 -1.63657427e+00 7.67238200e-01 3.65372688e-01 -1.17717266e-01 -6.53352976e-01 4.87993598e-01 -2.13686511e-01 2.41582662e-01 6.65276870e-02 8.35900828e-02 2.33432427e-01 -4.19625014e-01 4.04511720e-01 6.15053117e-01 3.02490562e-01 -2.33459920e-01 -5.36291420e-01 -1.91304430e-01 -3.10443461e-01 -1.20224535e-01 9.77315068e-01 -4.25063759e-01 1.17238738e-01 7.35935628e-01 7.04605758e-01 -1.94210932e-01 -1.15826380e+00 -7.77623728e-02 3.61203223e-01 -6.31367743e-01 8.81680697e-02 -8.03405583e-01 -7.99139798e-01 4.92415011e-01 4.36338872e-01 1.10784543e+00 1.38906109e+00 4.77394648e-02 2.74265081e-01 5.10283947e-01 8.12650621e-01 -8.43183696e-01 -1.52861759e-01 6.84543001e-03 4.87746537e-01 -1.12982988e+00 2.67800003e-01 -2.36045957e-01 -5.53208828e-01 1.28917158e+00 7.57454753e-01 -2.56293654e-01 4.26586688e-01 3.85435313e-01 1.99457984e-02 -5.05680025e-01 -1.45404041e+00 -9.87830535e-02 -9.69470739e-02 1.43311113e-01 1.74932852e-01 2.08627015e-01 -2.44945034e-01 -2.02984869e-01 8.86602700e-02 -2.36673474e-01 8.10351133e-01 1.13520896e+00 -7.75868297e-01 -1.18923581e+00 -7.80508399e-01 6.63994133e-01 -2.53327161e-01 6.57028496e-01 -8.17095339e-01 8.85599852e-01 1.02533765e-01 1.51000524e+00 -4.75487113e-02 -5.53312242e-01 4.27657306e-01 3.95286046e-02 2.48374939e-01 -1.00253153e+00 -3.01490873e-01 3.74057353e-01 3.99431020e-01 -5.42076409e-01 1.05404526e-01 -9.49671507e-01 -1.30781627e+00 -3.28230232e-01 -5.77921510e-01 4.25737917e-01 3.41461062e-01 1.51147318e+00 9.61306412e-03 5.83981156e-01 9.48116064e-01 -1.28561452e-01 -7.53838718e-01 -5.79470098e-01 -6.09620333e-01 1.24448918e-01 3.99404407e-01 -8.62971187e-01 -6.32679224e-01 -1.95768535e-01]
[5.423994541168213, 2.699327230453491]
1bf3d4c9-faf2-446b-9c37-cb8061d81899
episodic-policy-gradient-training
2112.01853
null
https://arxiv.org/abs/2112.01853v1
https://arxiv.org/pdf/2112.01853v1.pdf
Episodic Policy Gradient Training
We introduce a novel training procedure for policy gradient methods wherein episodic memory is used to optimize the hyperparameters of reinforcement learning algorithms on-the-fly. Unlike other hyperparameter searches, we formulate hyperparameter scheduling as a standard Markov Decision Process and use episodic memory to store the outcome of used hyperparameters and their training contexts. At any policy update step, the policy learner refers to the stored experiences, and adaptively reconfigures its learning algorithm with the new hyperparameters determined by the memory. This mechanism, dubbed as Episodic Policy Gradient Training (EPGT), enables an episodic learning process, and jointly learns the policy and the learning algorithm's hyperparameters within a single run. Experimental results on both continuous and discrete environments demonstrate the advantage of using the proposed method in boosting the performance of various policy gradient algorithms.
['Svetha Venkatesh', 'Dung Nguyen', 'Kien Do', 'Thommen K. George', 'Majid Abdolshah', 'Hung Le']
2021-12-03
null
null
null
null
['policy-gradient-methods']
['methodology']
[-2.00001210e-01 -1.40805349e-01 -8.48758042e-01 -7.98831284e-02 -7.52152056e-02 -4.99410838e-01 7.72234976e-01 -1.20695621e-01 -8.39336157e-01 1.40653801e+00 4.00331020e-02 -4.44743186e-01 -2.74768442e-01 -9.15052891e-01 -4.48570758e-01 -1.08104050e+00 -1.93097502e-01 5.31535864e-01 1.83352038e-01 -1.24988556e-01 6.99584305e-01 4.54913408e-01 -1.70414019e+00 -2.34610662e-01 8.42874348e-01 9.42280471e-01 4.26434278e-01 6.13609433e-01 -3.99756789e-01 4.72928435e-01 -6.84474647e-01 2.08972394e-01 1.13828212e-01 -3.63620460e-01 -6.54001832e-01 -1.53971151e-01 -3.96903336e-01 -4.38564569e-01 -4.15075749e-01 7.60937035e-01 5.78350544e-01 7.89549768e-01 3.05820137e-01 -8.96241069e-01 -9.68840048e-02 4.31945801e-01 -7.57149756e-02 4.84995186e-01 4.06538010e-01 4.91147459e-01 4.20838207e-01 -6.26723111e-01 5.57021081e-01 1.17529047e+00 2.71991163e-01 4.94845152e-01 -7.21749187e-01 -6.97735310e-01 5.70588589e-01 5.85425138e-01 -9.11295772e-01 -2.47514918e-01 5.76816380e-01 7.65671581e-02 1.29478669e+00 -1.84965536e-01 1.34666526e+00 1.08210528e+00 4.44264978e-01 6.83666587e-01 1.23801994e+00 -5.59831440e-01 9.82748568e-01 1.92900434e-01 -2.70898789e-01 7.95187712e-01 9.94550064e-02 8.21165323e-01 -6.56877935e-01 -4.78779197e-01 8.00001621e-01 -9.27043706e-02 -1.97347164e-01 -4.63817447e-01 -7.64973998e-01 7.60023594e-01 -5.88996857e-02 -7.31420964e-02 -8.63923788e-01 3.56644720e-01 3.67667675e-01 6.40356243e-01 6.34033456e-02 4.94937360e-01 -4.40406621e-01 -6.76739156e-01 -4.28515464e-01 4.59310532e-01 1.03380191e+00 4.95888263e-01 8.36731791e-01 4.41384524e-01 -3.49117011e-01 9.53063011e-01 2.79319942e-01 7.39782870e-01 1.03089094e+00 -8.61500919e-01 1.26236573e-01 2.80982405e-01 6.54160738e-01 -3.16765815e-01 -1.97455585e-01 -3.54494005e-01 -1.31162927e-01 5.23863792e-01 -1.15212880e-01 -6.17309988e-01 -1.07878828e+00 1.41384339e+00 9.78133857e-01 5.30180693e-01 2.41210610e-01 6.42857492e-01 1.61021397e-01 9.34226930e-01 4.09539521e-01 -7.61031926e-01 6.67695284e-01 -7.97843516e-01 -8.37918878e-01 -2.19462395e-01 2.84820080e-01 -4.80721444e-01 9.52749431e-01 4.03140396e-01 -1.14073718e+00 -3.06360394e-01 -1.26714957e+00 9.65086699e-01 -3.91528785e-01 -2.32477039e-01 6.55988753e-01 4.51230556e-01 -9.22765434e-01 9.88845944e-01 -9.70580935e-01 -2.47818455e-01 2.11215261e-02 5.17675698e-01 6.03525996e-01 4.14130896e-01 -1.38788462e+00 1.06603849e+00 9.50774789e-01 -1.34399712e-01 -1.37134600e+00 -5.52612841e-01 -4.35243338e-01 3.75859998e-02 6.59012675e-01 -5.90375125e-01 1.72608280e+00 -7.19618857e-01 -2.67050600e+00 1.24568053e-01 -1.00475639e-01 -5.03024578e-01 3.20837677e-01 -3.38183045e-01 -6.57015741e-01 -1.49441659e-02 -5.30719995e-01 3.40080857e-01 1.46109474e+00 -1.03755438e+00 -1.20242906e+00 -9.08833519e-02 -1.57405868e-01 9.13537621e-01 -2.59333134e-01 -3.78627628e-01 -3.09951842e-01 -2.03358054e-01 -1.27730742e-01 -1.02575994e+00 -2.79416561e-01 -7.07671046e-01 2.38279939e-01 -2.97182947e-01 1.13167489e+00 -4.79189306e-02 1.59358561e+00 -1.90915465e+00 9.33357701e-02 7.94096410e-01 -5.10260344e-01 6.40165806e-01 7.92947114e-02 5.11179805e-01 3.63156468e-01 -2.80609965e-01 1.34845644e-01 2.57355064e-01 1.06252261e-01 6.42332196e-01 -6.35876894e-01 1.95772305e-01 -4.38353658e-01 6.52908444e-01 -1.12286019e+00 -3.92237365e-01 3.16667557e-01 1.50385678e-01 -3.74879181e-01 5.53779125e-01 -5.71758270e-01 3.47120970e-01 -9.78468239e-01 3.81406456e-01 2.04089627e-01 -4.40234914e-02 6.95717633e-01 4.03530657e-01 -4.04202968e-01 4.23878968e-01 -1.43032539e+00 1.30663896e+00 -5.91337800e-01 -1.99258979e-02 -1.80926040e-01 -7.62520850e-01 9.13353264e-01 4.28403229e-01 4.28644836e-01 -1.06555796e+00 -9.91497189e-02 1.69043317e-01 -3.02952737e-01 -5.62019765e-01 6.49493515e-01 2.88867801e-01 3.93443257e-01 8.10002506e-01 7.20949620e-02 -5.19176796e-02 3.32362205e-01 -1.87108323e-01 8.28913867e-01 6.20681465e-01 5.54379642e-01 -1.08452542e-02 3.81992728e-01 1.05433494e-01 7.18073905e-01 1.26281440e+00 -1.15037821e-01 -6.73776269e-01 1.36623204e-01 -6.20883942e-01 -9.65587139e-01 -9.90219235e-01 -1.08387999e-01 1.59870923e+00 1.37763783e-01 -3.15966249e-01 -4.52568144e-01 -6.76581800e-01 2.33621418e-01 8.34059834e-01 -4.63918895e-01 -3.06966037e-01 -8.31702650e-01 -1.00565922e+00 4.42505851e-02 1.58314854e-01 9.88834500e-01 -1.56976712e+00 -1.11509717e+00 7.56860733e-01 4.19998735e-01 -3.05720687e-01 -1.20015480e-01 5.11848211e-01 -1.15216088e+00 -8.29758704e-01 -2.77151436e-01 -5.82670033e-01 3.70161444e-01 -1.76779032e-01 8.95646751e-01 1.96399271e-01 4.34284694e-02 6.02422297e-01 -2.88328260e-01 -2.19021663e-01 -4.36185598e-01 6.88398704e-02 2.64431417e-01 -2.66364008e-01 1.02249067e-02 -6.84657753e-01 -8.68303716e-01 3.74877781e-01 -5.96313179e-01 -3.44259739e-01 6.49511993e-01 1.09842813e+00 9.60003316e-01 -2.26109400e-02 8.92586350e-01 -9.35094357e-01 1.12017369e+00 -5.75869858e-01 -1.00648308e+00 4.89105940e-01 -1.44124937e+00 4.41514641e-01 5.19441187e-01 -7.49564826e-01 -1.62618601e+00 -2.03158423e-01 2.35018045e-01 -2.79825062e-01 -5.08584874e-03 4.05002534e-01 2.37889722e-01 -1.39714360e-01 5.51636219e-01 5.85300207e-01 -5.66343367e-02 -3.74306202e-01 4.78148937e-01 4.50865865e-01 3.86514604e-01 -1.06569934e+00 3.04207772e-01 1.76498182e-02 -1.55499727e-01 -3.25251490e-01 -5.16718388e-01 -2.26989701e-01 -6.73226640e-02 -3.77076298e-01 3.66342545e-01 -4.83440012e-01 -8.45748127e-01 4.47801799e-01 -4.06752884e-01 -8.21026683e-01 -4.56256926e-01 6.12912416e-01 -7.58703411e-01 -1.68458432e-01 -3.09825301e-01 -8.80774021e-01 -4.07918483e-01 -9.13468719e-01 3.01127076e-01 8.66746306e-01 2.81228521e-03 -1.17895603e+00 6.52255416e-01 -5.75080574e-01 5.49071729e-01 -2.04879150e-01 1.07472086e+00 -5.13599694e-01 -4.31692451e-01 1.57440916e-01 5.46976149e-01 -1.24456041e-01 -1.34657711e-01 -2.66991287e-01 -5.83408356e-01 -6.11427307e-01 3.96403810e-03 -3.95018607e-01 4.33241874e-01 3.43831152e-01 1.07337236e+00 -6.26044154e-01 -5.71751952e-01 8.01644325e-01 1.43447709e+00 7.96891451e-01 5.68603516e-01 8.67412746e-01 -1.72229335e-02 -2.03529984e-01 9.88027275e-01 1.12720156e+00 -1.85144410e-01 4.13277149e-01 1.69818088e-01 3.03429902e-01 3.23961884e-01 -5.26816010e-01 4.16689366e-01 3.79309446e-01 2.79750209e-02 8.10433403e-02 -5.95683634e-01 2.22625643e-01 -1.99322641e+00 -1.16048110e+00 9.62116182e-01 2.41356301e+00 1.20147765e+00 2.30092824e-01 1.09650537e-01 -2.92171061e-01 6.70821071e-01 1.86633736e-01 -1.29311287e+00 -8.65777135e-01 1.60665333e-01 4.70458657e-01 5.61903536e-01 5.83490968e-01 -7.88752913e-01 1.30888736e+00 7.32665396e+00 7.43012488e-01 -1.34600997e+00 -1.35253996e-01 2.19652846e-01 -5.27283490e-01 -6.47717416e-02 1.99580029e-01 -1.04564798e+00 5.51775098e-01 1.14172542e+00 -5.09907782e-01 1.06777072e+00 1.09944916e+00 5.29472589e-01 -5.20792663e-01 -6.71084464e-01 8.07146132e-01 -4.79853988e-01 -1.34316146e+00 1.00232847e-01 -1.86324880e-01 9.62436974e-01 -9.75787044e-02 3.71300668e-01 7.81295717e-01 7.32263386e-01 -6.99223459e-01 2.95968801e-01 6.19144380e-01 4.94068831e-01 -1.19955087e+00 2.02062145e-01 1.59578085e-01 -8.69967222e-01 -7.64034688e-01 -4.56955522e-01 4.54981625e-02 -8.37386996e-02 1.93409361e-02 -1.25253105e+00 2.32612025e-02 7.27922201e-01 3.09045523e-01 -1.97920412e-01 1.43601179e+00 -6.09505832e-01 1.00219309e+00 -3.08439761e-01 -5.59092522e-01 5.84246159e-01 -4.72497851e-01 6.93065822e-01 7.36478209e-01 2.54269242e-01 2.16302902e-01 3.58939976e-01 3.29844177e-01 2.58309782e-01 2.27855444e-01 -3.90276223e-01 6.03953004e-02 1.17017150e+00 9.67641473e-01 -4.45088118e-01 -6.00929558e-01 1.65322676e-01 7.53996611e-01 3.87656659e-01 9.63550687e-01 -6.93474531e-01 -3.78028065e-01 6.52622342e-01 -3.93781871e-01 7.25966632e-01 -1.21113785e-01 1.94385558e-01 -5.64670563e-01 -2.88524568e-01 -9.27034616e-01 5.78536391e-01 -4.07150865e-01 -6.11308634e-01 2.45869055e-01 2.85186708e-01 -1.07881832e+00 -9.62247491e-01 -1.34007454e-01 -9.10285413e-01 6.02514327e-01 -1.64202523e+00 -1.30657122e-01 1.46223798e-01 7.25235641e-01 4.32896048e-01 -5.26347399e-01 9.17151213e-01 -4.86425430e-01 -6.27337873e-01 4.44742322e-01 9.79878247e-01 -6.89513147e-01 4.50883925e-01 -1.24550211e+00 -3.15788746e-01 2.60201633e-01 -3.22415143e-01 5.46026707e-01 8.65751326e-01 -8.71260583e-01 -1.52487707e+00 -8.40359151e-01 1.64218947e-01 5.28017879e-01 5.13615429e-01 2.97533780e-01 -8.73854399e-01 4.11325246e-01 2.30411887e-01 -4.64673460e-01 5.27659714e-01 2.70063400e-01 2.94411212e-01 -1.46563098e-01 -1.06694233e+00 7.32367933e-01 8.41303170e-01 -1.47467986e-01 -4.94492501e-01 3.04333061e-01 4.31808978e-01 -6.91947579e-01 -8.06649148e-01 1.98621199e-01 5.29719651e-01 -5.60092747e-01 7.31534064e-01 -9.41192150e-01 -4.49796557e-01 -1.41098842e-01 1.73555642e-01 -1.93706548e+00 -3.06113034e-01 -1.09456694e+00 -7.59119630e-01 6.68545961e-01 2.20094979e-01 -9.99585569e-01 7.64046669e-01 4.84353185e-01 -6.64190575e-03 -1.11042166e+00 -9.66438890e-01 -6.52775884e-01 -1.46894380e-01 7.79578239e-02 1.07684517e+00 5.63082457e-01 1.94984615e-01 -3.29805575e-02 -3.56698304e-01 1.17918834e-01 3.80390167e-01 3.61249149e-01 4.15042430e-01 -5.78121781e-01 -6.72931910e-01 -4.24673826e-01 2.65617967e-01 -1.03141117e+00 3.99914145e-01 -5.56792915e-01 3.10975816e-02 -1.04611588e+00 -2.41117284e-01 -6.77237213e-01 -6.14184320e-01 4.33127224e-01 -3.47105294e-01 -7.97108829e-01 -9.06847045e-02 3.73931646e-01 -8.11581671e-01 8.72363687e-01 1.21797991e+00 1.60833657e-01 -1.00787580e+00 3.36341590e-01 -6.52157441e-02 5.16376615e-01 1.10561800e+00 -5.59580028e-01 -8.05778086e-01 -8.26299563e-02 1.17609009e-01 3.79998595e-01 -5.97248077e-02 -9.56903756e-01 2.92562336e-01 -7.21375048e-01 6.91820383e-01 -2.87039936e-01 1.14496291e-01 -4.07422483e-01 -7.73625728e-03 6.89842820e-01 -6.04052007e-01 4.57931429e-01 1.86225787e-01 9.23693538e-01 1.63737640e-01 -4.27996457e-01 7.31345475e-01 -1.52723774e-01 -1.14247286e+00 3.38124037e-01 -8.32130849e-01 1.11656502e-01 9.65546429e-01 -1.33890867e-01 -1.02879621e-01 -2.92846948e-01 -9.65778768e-01 5.96059024e-01 2.28335142e-01 2.43808120e-01 8.00552905e-01 -1.14444447e+00 -4.18525785e-02 2.52897263e-01 -3.46116573e-01 -5.13646364e-01 -9.18574333e-02 1.54731289e-01 -4.23552506e-02 3.43201607e-01 -3.64920318e-01 -3.07819605e-01 -8.50147486e-01 3.47675472e-01 7.86291122e-01 -5.62798321e-01 -7.14337468e-01 3.96251202e-01 -6.80710673e-01 -1.69658586e-01 7.05122352e-01 1.47152841e-01 -4.68997806e-01 -2.39956155e-01 5.84619701e-01 5.35944343e-01 -7.87546020e-03 3.50366741e-01 3.20900492e-02 9.21186060e-02 -1.82066977e-01 -6.41542137e-01 1.10857666e+00 -8.41408968e-02 2.23822430e-01 4.10059690e-01 6.54906809e-01 -5.53887486e-01 -1.60750747e+00 -4.85089451e-01 3.65617312e-02 -4.76057261e-01 3.62767756e-01 -1.18935835e+00 -6.50329173e-01 5.13764992e-02 1.16638613e+00 -3.18583190e-01 9.78229523e-01 -6.49773836e-01 6.40099525e-01 1.04781222e+00 4.99465138e-01 -2.12074733e+00 1.08605832e-01 7.68560767e-01 5.41288495e-01 -8.01627040e-01 1.73137471e-01 5.25563061e-01 -6.92836642e-01 1.11627185e+00 7.34979570e-01 -5.35497069e-02 6.63102984e-01 -1.47487111e-02 5.74049167e-02 2.17558984e-02 -1.23322415e+00 -9.66074914e-02 -2.82740116e-01 5.88440359e-01 -1.91326112e-01 8.70783105e-02 -5.57753205e-01 -1.57036617e-01 -2.26271283e-02 2.08661065e-01 1.49428546e-01 1.54307139e+00 -1.04762912e+00 -1.44678426e+00 -4.27661747e-01 4.84246939e-01 -7.20937178e-02 1.42437875e-01 7.80542120e-02 6.27825677e-01 -2.59680480e-01 6.56948626e-01 6.03159443e-02 -1.29495054e-01 1.69074282e-01 4.25089926e-01 5.19619823e-01 -2.87087798e-01 -7.63825119e-01 -9.53276306e-02 1.10980555e-01 -8.25406313e-01 9.56194028e-02 -6.89245582e-01 -1.75109899e+00 -1.13311030e-01 5.11223860e-02 7.58504391e-01 9.82131600e-01 9.49550688e-01 5.39057851e-01 3.94447088e-01 1.02120245e+00 -6.21633589e-01 -1.34954214e+00 -6.93449080e-01 -4.60774004e-01 1.52121156e-01 5.02523407e-02 -1.11822569e+00 -1.48385569e-01 -4.56025630e-01]
[4.112760066986084, 2.160810947418213]
8e0170ed-dde1-4130-8edf-b418d804649e
attention-based-depth-distillation-with-3d
2211.16779
null
https://arxiv.org/abs/2211.16779v2
https://arxiv.org/pdf/2211.16779v2.pdf
Attention-Based Depth Distillation with 3D-Aware Positional Encoding for Monocular 3D Object Detection
Monocular 3D object detection is a low-cost but challenging task, as it requires generating accurate 3D localization solely from a single image input. Recent developed depth-assisted methods show promising results by using explicit depth maps as intermediate features, which are either precomputed by monocular depth estimation networks or jointly evaluated with 3D object detection. However, inevitable errors from estimated depth priors may lead to misaligned semantic information and 3D localization, hence resulting in feature smearing and suboptimal predictions. To mitigate this issue, we propose ADD, an Attention-based Depth knowledge Distillation framework with 3D-aware positional encoding. Unlike previous knowledge distillation frameworks that adopt stereo- or LiDAR-based teachers, we build up our teacher with identical architecture as the student but with extra ground-truth depth as input. Credit to our teacher design, our framework is seamless, domain-gap free, easily implementable, and is compatible with object-wise ground-truth depth. Specifically, we leverage intermediate features and responses for knowledge distillation. Considering long-range 3D dependencies, we propose \emph{3D-aware self-attention} and \emph{target-aware cross-attention} modules for student adaptation. Extensive experiments are performed to verify the effectiveness of our framework on the challenging KITTI 3D object detection benchmark. We implement our framework on three representative monocular detectors, and we achieve state-of-the-art performance with no additional inference computational cost relative to baseline models. Our code is available at https://github.com/rockywind/ADD.
['Xiaoquan Wang', 'Xianzhi Li', 'Jian Pu', 'Yunzhe Wu', 'Zizhang Wu']
2022-11-30
null
null
null
null
['monocular-3d-object-detection']
['computer-vision']
[-0.08611285 0.1858308 -0.17693138 -0.62372804 -1.0700856 -0.76883 0.5620788 -0.14460602 -0.56656843 0.31674874 0.07132462 -0.35935012 0.1780152 -0.66670436 -1.195433 -0.57482684 0.5677547 0.533717 0.62771195 0.24729475 0.10578456 0.49435365 -1.7781434 0.33029404 0.94804263 1.0793831 0.5558504 0.8017531 -0.02722498 0.6449125 -0.4839418 -0.4331493 0.48017287 0.12267559 -0.69822323 0.07915352 1.118384 -0.97433215 -0.44927529 0.96596265 0.94416505 -0.0057459 0.50964326 -0.96701264 -0.96487194 0.36842307 -0.66559315 0.0899396 0.43427387 0.47837383 0.70813495 -1.3472767 0.41541368 1.440055 0.6112375 0.6256903 -1.1709403 -0.7340855 0.57375324 0.35621247 -1.4129417 -0.29976434 0.65869355 -0.47962755 1.3066766 -0.14152098 0.7011342 1.2462947 -0.3278714 1.2934666 1.0750923 -0.40340567 -0.11496741 0.24945132 -0.0931282 0.7839008 0.22105864 0.28605 -0.6648237 0.4819531 0.9466096 -0.02402948 -0.15926468 -0.6740148 -0.9373993 0.5923449 0.8933399 -0.22395372 -0.1545307 0.28693318 -0.14324388 -0.01892463 0.65121824 -0.00960045 -0.6886263 -0.03515656 -0.73266333 0.36423847 0.27114874 1.2944195 0.98797685 -0.1071234 -0.2657132 0.5669057 0.38121817 0.7225438 0.28975707 -1.0239108 0.59753734 0.75782084 0.12815735 -0.42302993 -0.45119047 -0.63238883 0.00909542 0.41478482 0.6021364 0.07711728 -1.329494 1.7298881 0.9146975 0.34978887 -0.17779809 1.2449038 1.2336698 0.16604228 0.02399127 0.5596328 1.2644763 -1.0912962 -0.05880116 -0.48594928 0.6399827 -0.66765964 1.241078 0.30717152 -1.1812627 -0.68789864 -0.8417107 -0.7137366 -0.36107403 0.09173325 0.5586474 0.69281155 -1.1496524 0.22427417 -0.9960084 -0.05744885 0.6594281 0.35918102 -0.3144258 -0.44892636 -0.87442833 0.93658125 0.31564787 -0.06779751 -1.1779491 -1.2180965 -1.0541947 -0.10913581 0.3893044 -1.0999495 1.6423734 -0.4844416 -1.518184 1.2780374 0.02358051 -0.3261646 0.7504818 -0.668688 0.43250695 0.1618586 0.23568276 1.3585713 0.756775 -1.0270368 -0.8191979 -0.5522951 0.36193195 0.6367212 -0.15957765 -0.41896316 -0.7261932 -0.23714909 0.4148492 -0.6938203 0.0445075 0.33960173 -0.4371145 -0.33454293 0.49111572 -0.4096156 0.52330196 -2.0718563 -0.07078391 -0.36173883 0.13577002 0.14436476 -0.16747166 -0.12670337 0.17895195 -0.13788079 -0.0424968 -0.68468535 0.16370755 0.07262508 -0.21598403 0.45363277 0.46923706 1.1819084 -1.082397 -0.4268376 0.63799965 0.87368417 -1.0165595 0.27651203 -0.43631497 0.5618115 -0.35675743 0.9159082 0.9324767 -0.2363988 -0.32951078 -0.16567415 -0.25296566 0.7416531 -1.0957162 2.1569588 -0.60346985 0.52447796 0.01391664 -0.6924694 0.68362945 0.01003974 0.01341877 -0.8134468 0.19019888 0.21068493 -0.32263625 -0.25460076 0.40334684 0.15658212 0.2042475 0.2809165 0.385311 -0.7041273 -0.34813827 0.10242619 1.0657121 0.79422116 0.02060098 0.11713248 0.14402929 -0.04714655 0.460521 0.8431404 -0.25545424 0.93509626 0.18290794 -0.26359636 -0.74665695 -1.3284744 -0.33426905 1.0231055 0.38663125 0.04523839 -0.67697585 -0.8259899 0.34685552 0.6282522 -0.5654842 0.09590241 -0.29834 -0.2906667 0.39200547 0.871721 0.49970722 -0.7734789 -0.70435154 0.18039568 0.12882258 -1.4263251 -0.41180325 0.25388268 -0.7605845 -1.0674597 -0.84154505 -0.6361327 0.7040435 0.687898 1.1821169 -0.13659656 -0.37545836 0.6426408 -0.15728591 -0.6137374 0.18565397 0.28319424 -0.03012894 -0.44185743 0.58486265 -0.62453526 -1.044977 0.20644306 -0.5046502 0.28834623 0.6205984 0.66952235 0.762231 -0.5972665 0.42232376 -0.6929163 -0.06240647 -0.24119017 -1.035785 -0.17929669 -0.3916515 -0.10730166 0.08516645 -0.402395 -1.2445668 0.42270142 -0.24590641 -0.8443235 -0.5077189 -0.15260373 -0.3479861 -0.30175382 0.81922233 0.24872646 -0.36566013 -0.6321096 0.68369114 0.55723774 0.7559322 -0.6808788 0.92723453 0.55212337 -0.38303986 -0.5041702 -1.3992552 -0.41040012 -0.90139574 0.00888832 0.8095386 -1.6990918 -0.88677585 0.6821835 -1.3230506 -0.65470076 -0.3048184 0.5788533 -0.5354934 -0.0183044 -0.4517473 -0.7275748 -0.13399637 -1.1605449 1.4672567 0.31401098 0.19970243 -0.7849978 -0.12629044 0.7297017 0.06612158 0.03105666 0.53843766 -0.3960607 -1.2154077 -0.03413275 -0.6429634 0.29657772 -0.1616071 -0.39527047 -1.669319 -0.04022253 -0.28381234 -0.6500402 1.0718547 0.47718108 1.2598417 0.03269405 -0.28068915 0.9878404 1.1722901 -0.19921443 0.13957201 0.25121766 1.1039947 0.5706752 0.595717 0.4034002 1.0770925 0.6203747 0.80392605 -0.05476216 -0.5715151 -0.64045745 0.22575729 0.30510718 0.30556434 -0.11649071 -0.95315963 0.92976487 -1.5626788 -0.73116434 0.02710247 2.0251644 1.1073289 0.18967682 0.077206 -0.01942607 0.4352293 -0.10171562 -0.8871342 -0.03223153 0.03689081 0.31061095 0.5186604 0.77980185 -1.0823396 1.3540719 5.0637736 0.63958687 -1.0586174 0.38739783 0.5549105 -0.59250915 -0.4590189 -0.25051025 -1.2894479 0.15921941 0.39039758 0.33024216 0.05783802 1.0611544 0.04281129 -0.13317923 -1.4022405 1.0810143 -0.05696459 -1.1777034 -0.23074643 -0.10588302 0.98127073 0.460138 0.33669573 0.50447273 0.5917568 -0.8889821 1.064854 0.02206354 0.8840426 -0.5628727 0.33819035 0.50379336 -1.0887947 -0.02079582 -0.36112112 -0.2757378 -0.02113959 0.72908604 -1.2023047 0.22277027 0.92823476 0.67721933 -0.54494464 1.1398933 -0.9477779 0.28969136 -0.5744326 0.10929894 0.2450183 0.4341348 0.35610458 0.9945927 0.3146458 0.13749678 0.08008486 1.162068 -0.18988094 -0.27889556 -0.5094064 0.6256851 0.85772854 1.0599661 -0.42510593 -0.16634327 -0.6957216 1.0341063 0.65409225 0.32563233 -0.81465894 -0.13323137 0.8985977 0.23662832 0.7353767 -0.0912823 -0.5078898 -1.1247159 0.07279815 -0.41328755 0.20720603 -1.0348042 -1.2112193 0.294708 0.05672334 -0.9205758 -0.19865645 -0.8733285 -0.41133046 0.8467536 -2.02028 -1.3757552 -0.5914197 0.57830775 0.53052104 0.32485297 0.4269201 0.43657917 -0.33293954 0.8606479 -0.55723464 -0.0426718 0.781741 -1.5247401 0.7388051 0.66727954 0.30507576 0.11139752 0.2949676 -0.46543893 -1.3264282 -1.1551144 0.7835338 -1.1636248 0.43071327 -0.6009812 -0.8053473 0.8137869 -0.18549505 0.40640578 0.23838905 0.07890598 -0.66413045 -0.02456402 -1.1525815 0.4212945 1.5976627 -0.67243224 -0.573064 0.32653823 1.2317971 -1.1163908 -0.6289149 0.458527 0.507481 -1.1773951 1.3445252 -0.13009353 0.16663054 -0.26777768 -0.15262817 -1.0819938 -0.09438318 -0.25265387 -0.3299115 1.1144482 0.3872773 -0.43963817 1.1507753 0.449956 -0.48347318 -0.7939775 -1.0669312 -0.63233364 0.21845302 -0.83902156 0.6896878 0.714674 -0.46823955 0.19557695 0.07774649 0.61832094 0.7305092 0.26302943 1.0156605 -1.0318955 -0.43864784 -0.39201722 -0.364132 -1.8342756 0.20288923 -0.8527145 0.12902613 -1.5560766 0.08629944 -0.4987109 -0.0675064 0.8060471 -0.19652414 0.44437346 0.10835962 -0.19072568 -0.51106584 0.71584773 1.437276 -0.02749079 -0.2530896 -0.07098074 -0.63711226 0.80259484 0.46725643 -0.42908278 -0.57498413 -1.1000984 0.05993097 -0.1006829 0.8691919 -0.82405895 0.29535425 -0.01961212 0.69139326 -1.0331322 0.6409336 -0.67712104 -0.57867855 0.21554562 0.01563638 -0.4320813 0.42654598 0.3072337 0.04358907 -0.07822397 0.7121491 -0.1657524 -0.9394655 0.5765213 0.26843718 0.20717812 0.8509957 -0.33799452 -0.5218929 -0.22914593 -0.536505 0.32353425 0.42442533 0.43988314 0.65805906 -1.2566199 -0.5942475 0.31131905 0.24159363 0.9263258 0.33416715 0.6041561 -0.4141719 0.50942016 0.1208913 -1.0347495 -0.8934736 0.35545594 0.4749596 0.24266768 -0.59390026 1.4866663 0.70900685 -0.8682871 0.5622844 -0.61188316 0.33809713 -0.13911203 0.48881286 0.10742152 0.11986119 -0.17316848 -0.5167248 0.92157996 -0.06615461 -0.0841054 1.2107122 -0.21981168 0.6085825 0.09487818 0.9186458 -0.25574997 -2.0930014 -0.47583252 -0.35488388 -0.6724534 0.32060844 -0.96788347 -0.9155991 1.2417295 0.5892375 -0.5209739 0.9346796 0.2814407 0.48755804 0.29720384 0.3162383 -0.7130182 0.35104734 0.57609326 0.6415707 -1.6965325 -0.12248974 -0.55501664 -0.24241017 0.5400712 1.2969187 0.00571914 0.40799627 0.20049532 0.14630157 0.04263688 -0.8170972 -0.6588951 0.48652366 0.89018977 0.37503323 -0.1402949 0.45053458 0.63374627 -0.36753246 -0.209178 0.1223015 0.7251834 -0.59499127 -0.851695 -0.29616377 -0.07134912 -0.20589802 -0.21131822 -0.298168 0.94453907 0.39238894 0.6387944 0.24234493 -0.25443858 0.48689187 -0.00748873 0.97039926 -1.0390116 -0.4075874 -0.09149434 -0.30374652 -0.63482416 -0.08673137 -0.60010195 -1.3716788 -0.2546731 -0.361503 -0.43782225 0.74493355 0.7973964 0.5835405 0.42039222 0.29661566 -1.1678333 -0.58583295 -0.8844353 -0.03886727 0.13433363 0.47258985 -0.8633028 -0.23089351 -0.23510869]
[7.900338172912598, -2.642502784729004]
f40a486a-9f10-4098-acd3-f098d8a5160b
depression-diagnosis-and-drug-response
2303.06033
null
https://arxiv.org/abs/2303.06033v1
https://arxiv.org/pdf/2303.06033v1.pdf
Depression Diagnosis and Drug Response Prediction via Recurrent Neural Networks and Transformers Utilizing EEG Signals
The Early diagnosis and treatment of depression is essential for effective treatment. Depression, while being one of the most common mental illnesses, is still poorly understood in both research and clinical practice. Among different treatments, drug prescription is widely used, however the drug treatment is not effective for many patients. In this work, we propose a method for major depressive disorder (MDD) diagnosis as well as a method for predicting the drug response in patient with MDD using EEG signals. Method: We employ transformers, which are modified recursive neural networks with novel architecture to evaluate the time dependency of time series effectively. We also compare the model to the well-known deep learning schemes such as CNN, LSTM and CNN-LSTM. Results: The transformer achieves an average recall of 99.41% and accuracy of 97.14% for classifying normal and MDD subjects. Furthermore, the transformer also performed well in classifying responders and non-responders to the drug, resulting in 97.01% accuracy and 97.76% Recall. Conclusion: Outperforming other methods on a similar number of parameters, the suggested technique, as a screening tool, seems to have the potential to assist health care professionals in assessing MDD patients for early diagnosis and treatment. Significance: Analyzing EEG signal analysis using transformers, which have replaced the recursive models as a new structure to examine the time dependence of time series, is the main novelty of this research.
['Fereidoun Nowshiravan Rahatabad', 'Arash Maghsoudi', 'Abdolkarim Saeedi']
2023-03-09
null
null
null
null
['drug-response-prediction', 'eeg', 'eeg']
['medical', 'methodology', 'time-series']
[-7.74851292e-02 -3.27548444e-01 -8.33620727e-02 -4.92707729e-01 -3.66502315e-01 -1.38685197e-01 2.33943939e-01 2.08749264e-01 -2.81051099e-01 8.38272095e-01 -1.61971390e-01 -3.20445031e-01 -4.22163010e-01 -7.89712548e-01 9.30423141e-02 -6.77500904e-01 -1.67572632e-01 5.47838688e-01 -1.83068201e-01 -1.31495193e-01 3.20464492e-01 8.50614369e-01 -1.30360067e+00 6.01533353e-01 9.70628560e-01 1.22494960e+00 5.43834381e-02 -6.36348352e-02 -1.33540947e-02 6.70987427e-01 -7.54860044e-01 -2.42534369e-01 -5.00965357e-01 -3.64061177e-01 -7.53249168e-01 -3.29623520e-01 -3.57500464e-01 -4.71415699e-01 -1.27343282e-01 7.74546325e-01 9.76835608e-01 -1.30982071e-01 6.23551846e-01 -9.55546141e-01 -4.08642024e-01 5.16155899e-01 -5.44809580e-01 5.57356834e-01 4.13618058e-01 -3.57225478e-01 1.86035171e-01 -8.75950933e-01 4.86684851e-02 7.99738646e-01 9.17492151e-01 5.63975513e-01 -1.03014970e+00 -1.07056367e+00 -2.64597982e-01 7.80544043e-01 -1.37188601e+00 -5.66277027e-01 7.63025820e-01 -4.35137868e-01 1.42022026e+00 -7.23677650e-02 8.67358863e-01 1.16111445e+00 7.38735080e-01 4.00708199e-01 1.03567719e+00 -1.75445661e-01 9.57518518e-02 1.36222646e-01 4.20170724e-01 2.08836630e-01 5.34675568e-02 -5.04117198e-02 -2.58299947e-01 -3.85653079e-01 6.16688967e-01 3.73631567e-01 -2.69477516e-01 2.83104092e-01 -7.31180549e-01 7.96368599e-01 2.13205889e-01 1.11543715e+00 -9.03059661e-01 -3.07559490e-01 7.17150331e-01 2.10557386e-01 6.97086990e-01 3.03948402e-01 -5.41274726e-01 -3.47332537e-01 -9.44953680e-01 2.13814620e-02 4.47525173e-01 1.10601626e-01 9.21379775e-02 4.27385122e-01 -1.87645033e-01 1.19598806e+00 7.17350245e-02 5.42077661e-01 1.14384520e+00 -4.29816574e-01 1.36998117e-01 9.16748285e-01 -1.63178504e-01 -1.23796213e+00 -8.67067039e-01 -7.13164091e-01 -1.26953495e+00 -2.65322059e-01 -7.90625960e-02 -9.40930471e-02 -6.68066502e-01 1.78837824e+00 -1.99060023e-01 3.02708685e-01 -4.56746630e-02 4.24120545e-01 7.50986874e-01 4.74445343e-01 9.00747776e-02 -6.59258485e-01 1.36371171e+00 -1.88779756e-01 -1.11196065e+00 -6.14369148e-03 7.64035165e-01 -4.24653113e-01 5.40452778e-01 7.60123491e-01 -1.17902792e+00 -3.84192020e-01 -9.53261793e-01 3.84438366e-01 -3.45251203e-01 5.21941304e-01 7.38995731e-01 7.66651928e-01 -1.14146554e+00 7.87405312e-01 -9.88034844e-01 -5.34756005e-01 3.94160062e-01 8.40350032e-01 -3.81583840e-01 3.08280475e-02 -1.50918806e+00 1.27248633e+00 7.60001615e-02 4.52183574e-01 -5.40851235e-01 -6.03928506e-01 -5.48018456e-01 2.07010195e-01 -5.43788671e-01 -3.96674842e-01 1.23964953e+00 -8.33307862e-01 -1.59982169e+00 8.48935843e-01 -2.90081292e-01 -7.80483723e-01 -1.19132586e-01 -6.67375252e-02 -8.60294461e-01 3.64073157e-01 -5.95391355e-02 1.66735455e-01 4.87199992e-01 -2.47196689e-01 -4.42938089e-01 -8.80657494e-01 -3.71267170e-01 -4.87525538e-02 -4.51490879e-01 1.43073112e-01 4.12981838e-01 -3.03588867e-01 3.26610088e-01 -5.54682493e-01 2.37640142e-02 -6.08322442e-01 4.37397808e-02 -4.46939796e-01 7.05978096e-01 -8.15380514e-01 1.53395283e+00 -1.85953343e+00 -3.90735686e-01 1.98493659e-01 4.12906796e-01 5.59946120e-01 1.97219476e-01 5.08156121e-01 -6.18658662e-01 -4.43981253e-02 -2.03299031e-01 3.34991924e-02 -3.19669873e-01 -2.00480953e-01 -2.09585905e-01 6.60733521e-01 2.43837297e-01 8.83894503e-01 -6.05996847e-01 1.25626370e-01 3.89798433e-01 8.72837543e-01 -3.03886756e-02 9.91770700e-02 5.87210178e-01 1.45902768e-01 -3.52158159e-01 4.74693030e-01 6.79849148e-01 -1.34407535e-01 2.56533623e-01 -1.86370060e-01 -7.74465799e-02 4.43147570e-01 -6.67616785e-01 1.14190757e+00 -2.06536755e-01 7.79203773e-01 -4.73735273e-01 -1.69352043e+00 1.08476090e+00 9.37286139e-01 5.74503362e-01 -1.31961644e+00 4.98240441e-01 4.98813242e-01 3.97109121e-01 -7.51007199e-01 -2.41529301e-01 -2.24247172e-01 3.81153703e-01 2.15593055e-01 -9.63133425e-02 3.20970863e-01 -5.08968048e-02 -2.76413232e-01 1.23445106e+00 -4.55426365e-01 4.67482924e-01 -1.39683768e-01 5.28253973e-01 -4.88454342e-01 4.68486935e-01 3.47719252e-01 -1.16984814e-01 2.20415518e-01 5.43966591e-01 -5.39125681e-01 -4.92481828e-01 -6.55308306e-01 -5.87375224e-01 3.22629064e-01 -5.18464684e-01 5.24751507e-02 -7.33347178e-01 -4.26893383e-02 -2.79366910e-01 7.10595369e-01 -7.71124005e-01 -7.84221947e-01 -3.18414301e-01 -1.15571439e+00 7.28686988e-01 5.36561310e-01 6.76717818e-01 -1.09612715e+00 -6.74125791e-01 6.26382351e-01 -3.74531865e-01 -8.08519900e-01 2.16474116e-01 5.15614688e-01 -1.44891429e+00 -1.03698885e+00 -8.17679346e-01 -8.38016808e-01 4.04625177e-01 1.69790730e-01 9.03976858e-01 -8.83010030e-02 -1.87904119e-01 1.26119629e-01 -2.31019050e-01 -2.01006010e-01 9.78460610e-02 -5.70247360e-02 1.60759181e-01 1.39992952e-01 7.48913586e-01 -1.05981004e+00 -7.96000540e-01 5.09210164e-03 -5.18762469e-01 -3.23694706e-01 5.55596173e-01 6.76401854e-01 1.84756860e-01 1.56925663e-01 1.17038286e+00 -7.83332169e-01 1.04280412e+00 -6.65213346e-01 -6.64121807e-02 -2.69591231e-02 -8.48303795e-01 -3.25022280e-01 5.41821837e-01 -5.30303001e-01 -9.15708184e-01 -3.00654173e-01 -5.31947136e-01 -1.44694000e-01 -2.31155336e-01 8.23576570e-01 -2.53494233e-02 3.76737788e-02 5.28362036e-01 3.14895600e-01 -1.58182636e-01 -5.08105934e-01 -6.00314975e-01 7.81742990e-01 1.79594696e-01 1.08205117e-02 -2.72450268e-01 1.12327784e-01 -1.42349124e-01 -8.64853024e-01 -3.53190154e-01 -4.15619373e-01 -3.30347955e-01 -5.56274503e-03 6.42264545e-01 -7.71068215e-01 -9.32591379e-01 8.43988180e-01 -1.33962870e+00 -3.05482652e-02 3.56169611e-01 8.99253309e-01 -2.39681318e-01 7.72482082e-02 -7.98933089e-01 -9.05404747e-01 -9.47418988e-01 -9.27081108e-01 6.69998944e-01 -2.51638633e-03 -3.99437875e-01 -1.03605640e+00 2.52939463e-01 -2.20250323e-01 5.62037468e-01 1.94355145e-01 1.22540975e+00 -9.82375979e-01 3.69524658e-01 -6.22446477e-01 -9.58910137e-02 3.11284035e-01 3.62520754e-01 -5.92181087e-02 -1.12784529e+00 -1.05893366e-01 9.24939811e-01 -9.12083089e-02 5.55822313e-01 9.83997881e-01 1.00081992e+00 8.06699768e-02 -4.29453969e-01 2.99230367e-01 1.20964468e+00 1.27827728e+00 1.16140652e+00 1.14235625e-01 6.89444020e-02 3.61454040e-01 2.06784159e-01 6.12917185e-01 2.86951393e-01 4.91563082e-01 2.18099952e-01 -9.78215113e-02 3.67362797e-01 3.22478443e-01 3.18978399e-01 9.52469110e-01 -1.06645869e-02 -1.27487838e-01 -1.06795812e+00 4.73302066e-01 -1.64388549e+00 -1.16741371e+00 -4.65798080e-01 2.25336576e+00 6.94347203e-01 -7.92730693e-03 3.77235077e-02 8.04907680e-01 7.02668667e-01 -5.06249368e-01 -4.07253742e-01 -5.95016778e-01 -3.01948693e-02 7.31383502e-01 -1.45494431e-01 1.75438710e-02 -7.25980520e-01 3.24914604e-01 6.36360836e+00 6.65779293e-01 -1.71134245e+00 1.86374471e-01 6.86423361e-01 1.64255604e-01 2.99238473e-01 -5.14709473e-01 -4.72757131e-01 6.39848828e-01 1.55101562e+00 -2.12212846e-01 2.71996558e-01 4.34891909e-01 6.91751957e-01 -1.16641484e-01 -1.09068644e+00 1.22743511e+00 4.58778106e-02 -9.18815374e-01 -1.73501834e-01 -1.05529584e-01 2.56944060e-01 -2.84915715e-01 2.53510058e-01 3.66281986e-01 -5.16692460e-01 -1.18295825e+00 1.53249145e-01 6.63198113e-01 6.16375268e-01 -9.46195662e-01 1.22086859e+00 3.58456492e-01 -1.00151527e+00 -1.62678748e-01 -2.62768835e-01 -4.41439241e-01 -1.90694496e-01 8.61798763e-01 -7.56893039e-01 2.26700678e-01 6.78697109e-01 1.00174022e+00 -4.44691837e-01 1.19392371e+00 2.39539314e-02 6.52210176e-01 -1.54506430e-01 -5.70244417e-02 6.59999847e-02 -1.18636213e-01 -1.36358216e-01 9.62724090e-01 7.73193598e-01 4.12965983e-01 -4.42951679e-01 5.08993506e-01 2.38969803e-01 2.03754500e-01 -7.08994448e-01 -1.93180457e-01 4.66182500e-01 1.04995382e+00 -7.31231511e-01 -3.95642191e-01 -1.92202955e-01 7.16380119e-01 1.49615616e-01 3.43560457e-01 -7.99072325e-01 -5.38217247e-01 2.21000388e-01 1.00811921e-01 -2.06434861e-01 4.17513549e-01 -3.97194624e-01 -8.71924281e-01 4.13880087e-02 -9.46476281e-01 2.09861308e-01 -8.89832497e-01 -1.07487917e+00 9.20412362e-01 -1.10537596e-01 -1.27104473e+00 -4.51989949e-01 -4.48101103e-01 -6.74983263e-01 8.85876060e-01 -1.13902676e+00 -6.95988595e-01 -3.98967490e-02 7.44515777e-01 3.10104936e-01 -1.04592659e-01 1.29717946e+00 7.95172870e-01 -8.94450605e-01 3.94595295e-01 2.87593454e-01 -1.65176332e-01 5.51500142e-01 -8.54206562e-01 -2.13534757e-01 2.56693751e-01 -5.26538193e-01 8.25882912e-01 3.62990052e-01 -4.93089020e-01 -9.11972880e-01 -7.38748729e-01 1.26317251e+00 1.31875634e-01 4.09391075e-01 -2.54361704e-02 -1.04931664e+00 3.11883509e-01 3.30439746e-01 -6.13768876e-01 8.84684324e-01 7.89339188e-04 9.70545113e-02 -4.02816057e-01 -1.40590346e+00 3.28087002e-01 4.80812669e-01 -5.85106850e-01 -6.71892226e-01 2.98087150e-01 1.51597247e-01 -1.88663658e-02 -1.05292714e+00 6.22624934e-01 7.49598384e-01 -1.13028717e+00 8.19441795e-01 -3.04822266e-01 3.68647277e-01 4.63091940e-01 1.02908954e-01 -1.34019625e+00 -5.48537791e-01 -1.24957331e-01 -1.49493113e-01 1.19863176e+00 3.08151841e-01 -1.01839161e+00 4.78216946e-01 6.60751760e-01 -5.62334694e-02 -1.00426710e+00 -7.94277489e-01 -4.83708203e-01 -1.19404353e-01 -5.48843801e-01 5.24800599e-01 1.01529694e+00 3.22356194e-01 4.46478993e-01 -1.43021390e-01 -9.31730792e-02 9.91955176e-02 -2.28575021e-01 -1.66747615e-01 -1.55014193e+00 2.89101243e-01 -5.76749027e-01 -4.68658984e-01 -2.04805985e-01 4.24220078e-02 -7.02759683e-01 -4.40717727e-01 -1.82503068e+00 2.15447187e-01 2.99196364e-03 -5.98194540e-01 6.04488492e-01 3.07741374e-01 1.55968636e-01 -4.99801517e-01 -6.64093904e-03 1.07273541e-01 4.90555853e-01 7.42668033e-01 -1.02966994e-01 -4.77956176e-01 3.67727399e-01 -6.77286088e-01 7.41596341e-01 1.16932976e+00 -6.27386272e-01 -7.62183189e-01 -1.60555467e-01 8.54488760e-02 4.42042142e-01 1.77017316e-01 -1.04593205e+00 1.35379866e-01 2.40840584e-01 8.15847576e-01 -7.80961633e-01 4.93391752e-01 -7.43380547e-01 3.97136986e-01 8.64667296e-01 -9.02778655e-02 5.03181458e-01 4.70683903e-01 -8.91880020e-02 -3.28156620e-01 -2.51558363e-01 7.61091650e-01 3.13275903e-02 -3.28109920e-01 2.65813231e-01 -8.92073929e-01 -2.68837899e-01 7.17972934e-01 -3.92197639e-01 -1.98601559e-01 -3.29294741e-01 -8.05942655e-01 -2.42213607e-01 -3.62556666e-01 1.69992253e-01 1.05166888e+00 -1.39763188e+00 -3.67203236e-01 2.26084828e-01 -2.60386080e-01 -8.22210014e-01 6.39644146e-01 1.41665125e+00 -2.39764839e-01 8.39732826e-01 -5.03120840e-01 -4.29535836e-01 -1.10383785e+00 4.08876240e-01 5.40159822e-01 -4.31019306e-01 -4.60362017e-01 5.09296119e-01 1.44658878e-01 8.34654942e-02 3.77749890e-01 -4.24103171e-01 -9.08915818e-01 4.15232718e-01 7.67798066e-01 4.23455238e-01 7.08054125e-01 -3.48977238e-01 -6.57966137e-01 4.94707912e-01 -1.93503618e-01 6.09501079e-02 1.70269573e+00 3.20826918e-02 -4.90509689e-01 6.31360412e-01 1.21672022e+00 -4.95301872e-01 -2.36994270e-02 1.70223847e-01 3.22279632e-02 8.39674398e-02 2.59897083e-01 -1.17824578e+00 -1.22934711e+00 1.33502901e+00 1.36118710e+00 4.01550591e-01 1.46653354e+00 -5.57402849e-01 6.58397794e-01 4.98286545e-01 3.46402526e-01 -8.77349496e-01 -5.02748266e-02 5.28578281e-01 8.57959449e-01 -8.29699159e-01 -3.41562659e-01 2.71601826e-01 -3.04762989e-01 1.25774443e+00 3.83717269e-01 -2.07675457e-01 8.49148691e-01 1.42833188e-01 -1.95240423e-01 -4.13731545e-01 -8.51044476e-01 5.23085035e-02 2.00395137e-01 6.74843609e-01 8.88803184e-01 1.16741294e-02 -6.40088558e-01 1.22862613e+00 3.08064520e-02 4.46620405e-01 2.32292727e-01 5.61180532e-01 -3.63304704e-01 -9.57518458e-01 -2.17040151e-01 8.93593788e-01 -8.24173748e-01 -2.97403932e-01 -3.86741847e-01 5.93075693e-01 1.40986755e-01 1.09636056e+00 -1.47795036e-01 -5.10585785e-01 5.26766539e-01 3.19988400e-01 4.94790375e-01 -5.61915040e-01 -8.86983573e-01 5.59582710e-02 -6.49123192e-02 -3.62393200e-01 -4.56458777e-01 -4.99206215e-01 -1.22892237e+00 -3.79480094e-01 -3.99087310e-01 2.29641452e-01 6.39445961e-01 1.12496972e+00 4.70570475e-01 7.77218282e-01 7.21275091e-01 -5.43954372e-01 -2.46938050e-01 -1.25073874e+00 -9.12259579e-01 -1.30747572e-01 2.27535814e-01 -8.24902892e-01 -2.48932511e-01 -2.88760483e-01]
[13.267805099487305, 3.4744272232055664]
68f4c701-c79b-487a-9216-179633404bfd
learning-the-sequential-temporal-information
1807.02857
null
http://arxiv.org/abs/1807.02857v1
http://arxiv.org/pdf/1807.02857v1.pdf
Learning The Sequential Temporal Information with Recurrent Neural Networks
Recurrent Networks are one of the most powerful and promising artificial neural network algorithms to processing the sequential data such as natural languages, sound, time series data. Unlike traditional feed-forward network, Recurrent Network has a inherent feed back loop that allows to store the temporal context information and pass the state of information to the entire sequences of the events. This helps to achieve the state of art performance in many important tasks such as language modeling, stock market prediction, image captioning, speech recognition, machine translation and object tracking etc., However, training the fully connected RNN and managing the gradient flow are the complicated process. Many studies are carried out to address the mentioned limitation. This article is intent to provide the brief details about recurrent neurons, its variances and trips & tricks to train the fully recurrent neural network. This review work is carried out as a part of our IPO studio software module 'Multiple Object Tracking'.
['Pushparaja Murugan']
2018-07-08
null
null
null
null
['stock-market-prediction']
['time-series']
[ 4.00204770e-02 -4.67577785e-01 -2.58120000e-01 -1.14510886e-01 1.68071240e-01 -2.26333901e-01 7.89515018e-01 -2.49935269e-01 -3.80418956e-01 5.91200829e-01 1.95048138e-01 -5.45296848e-01 1.34481534e-01 -5.40192306e-01 -4.38138574e-01 -4.72131222e-01 -1.98219776e-01 1.33912608e-01 4.05972719e-01 -1.94800347e-01 5.65451741e-01 7.46059775e-01 -1.66421461e+00 2.92795926e-01 7.32059777e-02 1.14844501e+00 4.54324573e-01 8.89573932e-01 -7.72425771e-01 1.65074635e+00 -6.06169403e-01 9.01255384e-02 -6.93184212e-02 -4.62268323e-01 -4.50563073e-01 -3.98496389e-01 -3.12028259e-01 2.15544235e-02 -3.91300201e-01 9.45638835e-01 6.16403818e-01 2.03156054e-01 3.28147411e-01 -9.56754446e-01 -8.63892734e-01 7.47677863e-01 -3.92711014e-01 8.77858400e-01 1.96570814e-01 -9.84484479e-02 3.62202585e-01 -1.03827846e+00 5.78575909e-01 1.17130339e+00 6.68063819e-01 4.42674816e-01 -5.25385439e-01 -8.47173631e-01 4.28161979e-01 4.59527224e-01 -9.26987529e-01 -4.19025958e-01 8.65547538e-01 -3.31261486e-01 1.36796403e+00 2.68922389e-01 6.86205328e-01 8.99740994e-01 4.06359017e-01 8.28363180e-01 9.86000896e-01 -4.76875216e-01 -5.07693999e-02 2.35340223e-01 4.11787122e-01 5.03700733e-01 -2.75219202e-01 1.89247563e-01 -6.96705341e-01 2.13583142e-01 9.50856388e-01 3.71371806e-01 -3.69164249e-04 3.73405695e-01 -1.13884401e+00 6.10237181e-01 3.95190328e-01 7.60186315e-01 -7.20014274e-01 2.50326246e-01 5.50290048e-01 5.82125604e-01 3.17247897e-01 2.60272846e-02 -4.84647036e-01 -2.67943978e-01 -7.86938429e-01 1.05549589e-01 7.83754885e-01 7.62555301e-01 2.71041632e-01 7.60854185e-01 2.32708268e-02 6.26102805e-01 6.26536906e-01 4.61473972e-01 1.21187961e+00 -4.93313819e-01 3.61356884e-01 6.68690741e-01 6.88277511e-03 -9.72082496e-01 -4.99686837e-01 -5.54479599e-01 -9.28486645e-01 3.28400880e-01 -9.91939828e-02 -4.06328768e-01 -1.12561238e+00 1.24234998e+00 9.74627435e-02 4.43456829e-01 1.46034583e-01 7.43436337e-01 1.16637349e+00 1.36512637e+00 1.07139312e-02 -6.76271737e-01 1.26726723e+00 -1.00853848e+00 -1.16953945e+00 -3.03277940e-01 1.86844364e-01 -1.04420936e+00 6.72587514e-01 4.72482182e-02 -1.09466851e+00 -5.48060060e-01 -8.55142236e-01 1.48350090e-01 -9.03030157e-01 8.78010616e-02 8.72733831e-01 2.43858844e-01 -8.20358336e-01 5.66676497e-01 -8.07248533e-01 -1.21015415e-01 -5.52731603e-02 3.92874807e-01 5.67708053e-02 6.95858061e-01 -1.41782868e+00 1.15459073e+00 5.76059520e-01 6.74414754e-01 -6.12916529e-01 -4.45314229e-01 -6.17531717e-01 -3.36511359e-02 4.50936221e-02 -2.27666870e-01 1.57142127e+00 -1.07787573e+00 -1.84718835e+00 5.71621299e-01 -3.03651690e-01 -8.70702088e-01 3.93293470e-01 -2.40003973e-01 -7.36715734e-01 -3.90436798e-01 -2.99976856e-01 4.20234680e-01 6.04536057e-01 -3.40006053e-01 -5.42759657e-01 -3.67665179e-02 -5.61616242e-01 2.80162066e-01 1.21021137e-01 7.17755735e-01 -2.29663312e-01 -8.02786767e-01 8.06747377e-02 -7.67684340e-01 -2.70210832e-01 -4.55424070e-01 -4.22785878e-02 -3.27429950e-01 1.23843086e+00 -6.39364779e-01 1.49993026e+00 -2.22522998e+00 -1.39758721e-01 -6.68969899e-02 -4.48827177e-01 5.24344921e-01 2.79803425e-01 4.47021574e-01 -4.58282113e-01 -1.50114045e-01 8.04357156e-02 -1.89257309e-01 -1.83067262e-01 3.17111239e-02 -7.07949877e-01 3.63311589e-01 7.79483095e-02 1.06074476e+00 -4.90111738e-01 -2.46611848e-01 4.16626185e-01 5.39716959e-01 2.00956821e-01 1.15684994e-01 -3.24363410e-01 3.30092043e-01 -2.93953747e-01 5.40553868e-01 2.05375299e-01 -2.66233921e-01 -2.26750419e-01 1.43239275e-01 -7.35540986e-01 4.74619478e-01 -1.49645078e+00 1.35461128e+00 -2.73552239e-01 1.12413621e+00 -3.17230105e-01 -9.06456590e-01 1.19425166e+00 6.43939793e-01 2.07056314e-01 -1.03600502e+00 2.02816948e-01 3.42575103e-01 -1.50444821e-01 -6.86691642e-01 6.24641061e-01 -2.21644804e-01 2.09627986e-01 4.53775704e-01 -9.01413187e-02 5.18739760e-01 -5.67090064e-02 -1.28510013e-01 4.55985337e-01 1.56200245e-01 1.69809297e-01 2.08683178e-01 7.39231050e-01 -1.04790777e-01 5.78492582e-01 5.26260197e-01 -8.63571689e-02 3.12704206e-01 1.85299471e-01 -7.58689225e-01 -1.02364075e+00 -5.06361187e-01 2.06778988e-01 1.09107184e+00 -3.68804067e-01 2.92247325e-01 -1.66881725e-01 4.01882380e-02 -4.39784795e-01 6.33993030e-01 -4.08613622e-01 1.28314272e-01 -6.88478291e-01 -7.23234951e-01 4.08069730e-01 5.69363654e-01 7.10218191e-01 -1.98734653e+00 -8.46780360e-01 6.29935205e-01 2.60787338e-01 -7.12404728e-01 -3.13271910e-01 3.78270954e-01 -9.07125890e-01 -5.16352117e-01 -9.18360233e-01 -1.13209283e+00 1.59204304e-01 -4.97053787e-02 8.90229583e-01 1.15106441e-01 -1.83860824e-01 -1.47379369e-01 -2.00076610e-01 -8.97262394e-01 -1.77453071e-01 1.43299893e-01 -3.55782360e-03 -1.14681020e-01 4.28789109e-01 -5.88515580e-01 -4.27660465e-01 6.57549948e-02 -8.09333980e-01 1.06304929e-01 6.65822148e-01 6.57743216e-01 4.91750747e-01 -6.64461851e-02 8.17169189e-01 -6.87062502e-01 9.48501587e-01 -4.07038212e-01 -1.02672422e+00 2.54521191e-01 -5.32317996e-01 -1.98451318e-02 4.28898543e-01 -7.60002077e-01 -1.11148691e+00 6.93063661e-02 -2.20900387e-01 -4.18848187e-01 3.13887075e-02 7.98173785e-01 5.20438910e-01 1.25803411e-01 3.42513829e-01 7.69103587e-01 1.21212177e-01 -5.55414736e-01 1.60701752e-01 7.35967994e-01 5.74997783e-01 1.48884237e-01 2.35840261e-01 1.23995997e-01 -2.56546676e-01 -7.09488750e-01 -4.71449852e-01 -6.11093104e-01 -3.31790626e-01 -5.08800328e-01 6.13557160e-01 -7.16390491e-01 -8.48591626e-01 8.28272820e-01 -1.41653740e+00 -1.46919906e-01 -2.06105426e-01 8.12078178e-01 -2.99026906e-01 -2.28485882e-01 -9.85397577e-01 -1.22012448e+00 -8.38593841e-01 -9.49830770e-01 4.16836083e-01 7.83445239e-01 -4.13692519e-02 -1.08063316e+00 2.80828178e-01 -4.72043157e-01 8.02354455e-01 1.25496015e-02 4.34761822e-01 -7.98185527e-01 -6.60653532e-01 -2.88079292e-01 -4.70143072e-02 2.49570370e-01 -3.97607731e-03 1.67197540e-01 -8.84162545e-01 1.39957666e-01 2.67722458e-01 6.69893026e-02 7.54499793e-01 6.73711360e-01 6.87355042e-01 -4.53973025e-01 -2.45629027e-01 3.48808289e-01 1.39127839e+00 9.54073727e-01 8.21413934e-01 4.41675216e-01 3.14905554e-01 5.55920422e-01 4.30682719e-01 1.40103623e-01 -3.03238388e-02 1.24903888e-01 2.03856066e-01 5.25096469e-02 2.78000087e-02 -1.54003918e-01 6.30487323e-01 1.34178329e+00 2.36756653e-02 -1.71717182e-02 -8.57005477e-01 3.93214226e-01 -1.94591951e+00 -1.50165009e+00 -2.10408449e-01 1.82464170e+00 4.68793780e-01 3.28798145e-01 -2.20196024e-02 1.28029972e-01 8.36638749e-01 2.46038541e-01 -5.61320841e-01 -7.24168360e-01 -1.71706021e-01 -2.89205573e-02 4.91989166e-01 3.97939831e-01 -9.21310902e-01 1.08060110e+00 7.20209599e+00 4.54629809e-01 -1.88759184e+00 -7.30569810e-02 4.63375121e-01 -1.68330818e-01 -2.50740200e-02 -1.99460611e-01 -9.77284610e-01 6.41552091e-01 1.48089433e+00 -9.22058746e-02 4.11408216e-01 7.70142913e-01 5.41870356e-01 -9.12114456e-02 -5.06307483e-01 1.14149785e+00 -7.84932449e-02 -1.90278137e+00 3.78325060e-02 -3.20981473e-01 4.25779045e-01 3.55280429e-01 2.32590929e-01 4.72000748e-01 7.71761760e-02 -1.02229440e+00 7.13767588e-01 9.28906977e-01 2.38045767e-01 -8.86065900e-01 6.89029813e-01 3.88206244e-01 -1.27091825e+00 -2.15550929e-01 -2.09965587e-01 -4.37643677e-01 4.59277868e-01 4.22757864e-01 -7.34600902e-01 2.36568555e-01 8.27402353e-01 7.17892885e-01 -2.41677374e-01 1.10761619e+00 7.68260881e-02 6.32461607e-01 -3.39935452e-01 -5.65945029e-01 4.73041981e-01 -2.80075282e-01 5.68606555e-01 1.21604848e+00 2.49066800e-01 1.15962677e-01 -3.33252311e-01 6.90197587e-01 1.14983238e-01 2.64797568e-01 -7.77422488e-01 -4.40533876e-01 3.12794030e-01 9.58332002e-01 -9.71967399e-01 -5.74502051e-01 -3.98526847e-01 6.81077838e-01 -1.14798211e-01 3.16343069e-01 -8.43164444e-01 -5.36454499e-01 2.70114928e-01 -2.80485094e-01 3.32542092e-01 -2.65128344e-01 -2.69319296e-01 -9.38812613e-01 -3.27739008e-02 -6.84533596e-01 3.02830577e-01 -1.18408108e+00 -7.53504038e-01 8.04618239e-01 -1.90937698e-01 -1.20556974e+00 -7.01301157e-01 -7.16247141e-01 -1.01290512e+00 1.10149503e+00 -1.43510914e+00 -7.58775294e-01 1.88509420e-01 6.75991595e-01 8.37555349e-01 -6.12344325e-01 6.59266293e-01 5.19624531e-01 -8.42537820e-01 -6.04202524e-02 2.82324776e-02 2.57221997e-01 1.37737885e-01 -8.46212745e-01 4.10775781e-01 1.00194657e+00 1.39937431e-01 7.85104215e-01 8.26882720e-01 -7.63858259e-01 -1.23548937e+00 -7.44454145e-01 1.17147744e+00 1.76919833e-01 9.05927122e-01 -7.34547228e-02 -1.01279223e+00 1.05000985e+00 5.80793142e-01 -1.40571550e-01 3.32152367e-01 -2.44423747e-01 1.70041859e-01 -1.34919703e-01 -6.97385907e-01 6.27360582e-01 4.78219032e-01 -5.32982051e-01 -8.44012260e-01 2.02430382e-01 1.01020348e+00 -5.99587560e-01 -5.63619077e-01 1.40198663e-01 6.54110551e-01 -7.56689250e-01 7.78030992e-01 -6.16005003e-01 1.20673910e-01 -4.24388200e-01 1.48787394e-01 -8.60463977e-01 -2.03015238e-01 -8.72174561e-01 -3.35200429e-01 1.34611666e+00 7.70514786e-01 -9.47682023e-01 7.59501934e-01 3.75987232e-01 -1.75316393e-01 -6.76934361e-01 -7.71144688e-01 -4.21621263e-01 -4.32383180e-01 -5.92194021e-01 5.04341960e-01 7.89208710e-01 -3.42858642e-01 4.87700015e-01 -5.56053698e-01 -9.67281312e-03 1.73151001e-01 1.07523598e-01 1.66605502e-01 -1.06180084e+00 9.29588303e-02 -6.60342991e-01 -3.22659791e-01 -1.00512969e+00 -8.99554566e-02 -4.80503440e-01 -1.94351196e-01 -1.46925700e+00 -3.20077211e-01 -2.40327507e-01 -4.58723813e-01 3.26478273e-01 2.90536046e-01 -9.97077972e-02 2.82614350e-01 3.53882402e-01 -3.76291990e-01 2.01302648e-01 1.03213704e+00 -5.09356894e-03 -4.91131842e-01 4.11471635e-01 -3.11500460e-01 5.68169951e-01 1.04696727e+00 -4.76349652e-01 -4.21408445e-01 -3.33111435e-01 4.82727796e-01 2.29256332e-01 2.61507392e-01 -7.91813016e-01 8.85143161e-01 -2.35818252e-01 5.28741002e-01 -1.27474535e+00 3.14783812e-01 -9.14387226e-01 5.78941464e-01 7.44364977e-01 -3.86965811e-01 7.45190680e-01 3.95497382e-01 2.89421111e-01 -5.50207019e-01 -4.56793755e-01 5.73205292e-01 -5.80392718e-01 -1.11618292e+00 1.04356691e-01 -7.33185649e-01 -4.91927534e-01 1.01988721e+00 -4.13107544e-01 -1.14661977e-01 -3.89044225e-01 -7.11969018e-01 1.31862924e-01 -2.94939578e-01 6.78468227e-01 6.78145111e-01 -1.37481844e+00 -3.44192177e-01 2.46035874e-01 -5.98359287e-01 -9.17246416e-02 3.07391554e-01 8.46273124e-01 -6.48732841e-01 8.95945549e-01 -2.13918507e-01 -3.79741102e-01 -1.28529251e+00 7.04393148e-01 6.64343119e-01 -3.22903335e-01 -7.16255248e-01 6.58396602e-01 -4.22026277e-01 -1.75293997e-01 5.41519463e-01 -5.05349994e-01 -8.59846234e-01 2.34797299e-01 9.50607538e-01 3.18723172e-01 3.35859135e-02 -4.69334841e-01 -3.03119540e-01 4.83060867e-01 -1.99212879e-01 -3.00068647e-01 1.47668326e+00 -1.10039279e-01 -3.27538699e-01 1.11424267e+00 1.10727942e+00 -3.86796713e-01 -8.07192624e-01 -1.59111962e-01 3.73678952e-01 2.30426744e-01 1.40192702e-01 -8.07266891e-01 -1.04282844e+00 7.88195729e-01 8.83144319e-01 4.07710910e-01 9.85440969e-01 -4.47323084e-01 6.81423366e-01 2.79779613e-01 1.99665017e-02 -1.21285439e+00 -1.89270765e-01 9.89273906e-01 9.21608567e-01 -8.90125573e-01 -3.19980294e-01 2.73028970e-01 -6.15828097e-01 1.38494360e+00 2.40566999e-01 -2.51993656e-01 9.30339217e-01 5.08548737e-01 4.29136544e-01 -2.06975594e-01 -1.07014954e+00 4.66523208e-02 1.69849575e-01 1.56770632e-01 6.70988321e-01 -2.86527932e-01 -2.25273073e-01 1.75671771e-01 -1.76764861e-01 4.49899703e-01 3.10750932e-01 1.06069481e+00 -5.32412827e-01 -7.42665589e-01 -6.04244173e-01 4.83820707e-01 -8.16605449e-01 -1.64285555e-01 -1.81522802e-01 4.98407841e-01 -2.52609432e-01 5.50243497e-01 3.98126096e-01 -9.01289284e-02 3.33597213e-01 2.99328774e-01 -1.10556528e-01 -2.91969329e-01 -9.13773239e-01 -3.74455750e-02 -7.37668425e-02 -3.40684742e-01 -6.38110757e-01 -6.19896591e-01 -1.41575444e+00 -1.97280556e-01 -2.27675483e-01 2.75978953e-01 1.16618681e+00 9.92347300e-01 2.61390060e-01 8.87092948e-01 5.06911218e-01 -8.49299192e-01 -3.05893183e-01 -1.11552858e+00 -3.26794565e-01 -1.30295008e-01 5.57866931e-01 -3.80804330e-01 -1.97136149e-01 1.54553456e-02]
[10.83125114440918, 6.249029159545898]
8b690070-e285-445a-8e1c-28cabb11e506
text-perceptron-towards-end-to-end-arbitrary
2002.06820
null
https://arxiv.org/abs/2002.06820v2
https://arxiv.org/pdf/2002.06820v2.pdf
Text Perceptron: Towards End-to-End Arbitrary-Shaped Text Spotting
Many approaches have recently been proposed to detect irregular scene text and achieved promising results. However, their localization results may not well satisfy the following text recognition part mainly because of two reasons: 1) recognizing arbitrary shaped text is still a challenging task, and 2) prevalent non-trainable pipeline strategies between text detection and text recognition will lead to suboptimal performances. To handle this incompatibility problem, in this paper we propose an end-to-end trainable text spotting approach named Text Perceptron. Concretely, Text Perceptron first employs an efficient segmentation-based text detector that learns the latent text reading order and boundary information. Then a novel Shape Transform Module (abbr. STM) is designed to transform the detected feature regions into regular morphologies without extra parameters. It unites text detection and the following recognition part into a whole framework, and helps the whole network achieve global optimization. Experiments show that our method achieves competitive performance on two standard text benchmarks, i.e., ICDAR 2013 and ICDAR 2015, and also obviously outperforms existing methods on irregular text benchmarks SCUT-CTW1500 and Total-Text.
['ShiLiang Pu', 'Zhanzhan Cheng', 'Yunlu Xu', 'Liang Qiao', 'Yi Niu', 'Sanli Tang', 'Fei Wu']
2020-02-17
null
null
null
null
['text-spotting']
['computer-vision']
[ 5.26006401e-01 -2.78434515e-01 5.98547570e-02 -2.99675226e-01 -7.91829526e-01 -3.57884109e-01 6.17322326e-01 8.84181932e-02 -3.88764709e-01 5.68065755e-02 -9.17101838e-03 -2.74689674e-01 2.08048016e-01 -7.20949054e-01 -6.23052120e-01 -7.20620453e-01 8.16837311e-01 9.78357017e-01 6.93226576e-01 9.53025445e-02 5.93640745e-01 2.70605713e-01 -1.23684716e+00 2.96058327e-01 1.02756405e+00 9.00320768e-01 5.49304128e-01 6.51442647e-01 -6.02236986e-01 5.47326565e-01 -3.85239065e-01 -3.37328136e-01 1.56585407e-02 -2.51449406e-01 -5.87944984e-01 5.97892940e-01 3.39867264e-01 -3.31717014e-01 -4.08445120e-01 9.61062551e-01 6.12503529e-01 -9.92956683e-02 9.48627710e-01 -6.31098032e-01 -6.08281314e-01 7.47960687e-01 -1.01554966e+00 -1.51869521e-01 2.09611818e-01 1.52719403e-02 1.19391358e+00 -1.15196466e+00 3.82756233e-01 1.16257560e+00 7.22753942e-01 3.86263281e-01 -8.32056284e-01 -2.33077779e-01 1.71158910e-01 -9.19350460e-02 -1.32915568e+00 -3.43614966e-01 7.57867217e-01 -3.78269136e-01 8.02209258e-01 4.25369412e-01 3.02505821e-01 9.40815389e-01 2.66367763e-01 1.44645727e+00 5.43637753e-01 -3.50232571e-01 -7.09154904e-02 -8.80558137e-03 1.51214123e-01 8.52044642e-01 2.52424300e-01 -6.32456362e-01 -4.32036936e-01 2.34869123e-01 5.92349291e-01 2.22154364e-01 -2.46263564e-01 -2.59236172e-02 -1.38604844e+00 5.87750554e-01 1.04072772e-01 3.52950007e-01 -9.75170210e-02 -2.00997308e-01 5.44361651e-01 7.52794072e-02 4.37352866e-01 1.75615828e-02 -3.51177871e-01 4.97151352e-03 -1.04535675e+00 -1.29330074e-02 6.34332180e-01 8.92004728e-01 5.55107594e-01 6.49418458e-02 -2.21349344e-01 1.21099365e+00 5.82277715e-01 8.52414727e-01 6.24855995e-01 2.91238397e-01 9.43184137e-01 1.06574380e+00 -2.52452195e-01 -9.92733300e-01 -5.27542770e-01 -2.37443686e-01 -1.01922429e+00 -3.94221693e-01 3.20746750e-01 5.03651053e-03 -1.26047695e+00 4.91098940e-01 2.45167345e-01 -2.07403138e-01 -2.73978978e-01 9.35405016e-01 1.05525577e+00 8.75583470e-01 -2.13364035e-01 1.95570007e-01 1.44792950e+00 -1.28823948e+00 -6.67117059e-01 -5.84773481e-01 6.76543832e-01 -1.26969945e+00 1.27971959e+00 4.34254289e-01 -9.42268789e-01 -3.67861569e-01 -9.88718569e-01 -3.44009638e-01 -3.01094502e-01 9.26384568e-01 7.48936832e-02 4.17694688e-01 -7.43252635e-01 7.82055482e-02 -9.74521160e-01 -5.87876320e-01 5.18785357e-01 3.84463340e-01 1.51819468e-01 1.73711553e-01 -5.43878257e-01 3.64777684e-01 5.38608193e-01 4.24017876e-01 -3.76990855e-01 4.83513772e-02 -6.78550124e-01 1.30248114e-01 6.15854621e-01 -5.75300992e-01 1.04088628e+00 -7.67274082e-01 -1.71802211e+00 1.04529262e+00 -1.64480746e-01 -2.48488307e-01 8.45998824e-01 -3.60416353e-01 -1.55995235e-01 4.92142374e-03 1.17600232e-01 4.70087200e-01 1.13146949e+00 -9.96512771e-01 -8.46885324e-01 -4.16286021e-01 -8.35083604e-01 5.11482060e-01 -5.49432456e-01 5.99055663e-02 -1.03077817e+00 -1.01431906e+00 6.35142505e-01 -5.53559601e-01 1.28778908e-02 5.41866347e-02 -9.62515831e-01 -5.49074590e-01 1.25797462e+00 -6.52116120e-01 1.08773041e+00 -1.86198723e+00 4.87125255e-02 6.10853508e-02 2.16599837e-01 2.27357179e-01 6.78974539e-02 4.01796907e-01 3.98321748e-01 -1.51890917e-02 -3.38120878e-01 -8.50019276e-01 2.34610692e-01 -3.47036086e-02 -6.97426796e-01 6.17124200e-01 1.35032892e-01 1.06460452e+00 -1.84687704e-01 -9.68823969e-01 6.49657905e-01 3.11414480e-01 -1.90791011e-01 6.27052262e-02 -6.97690248e-01 1.84276536e-01 -9.36684310e-01 9.56640303e-01 8.50280464e-01 -4.00439471e-01 -1.44019783e-01 7.33822510e-02 -1.54780179e-01 -1.76488385e-02 -1.12138999e+00 1.52031147e+00 5.70838936e-02 7.88345933e-01 -2.80488506e-02 -1.05034828e+00 1.22253180e+00 4.47786646e-03 2.66799748e-01 -6.28907323e-01 4.20058668e-01 2.51051247e-01 -4.94839728e-01 -7.53203213e-01 6.99386418e-01 1.72437876e-01 2.62663867e-02 3.65241438e-01 -3.55630279e-01 -1.76820293e-01 -2.21757181e-02 -1.23362131e-01 9.03971553e-01 3.17514837e-01 -9.42957029e-02 -1.24240570e-01 8.36302221e-01 1.33108363e-01 3.27272683e-01 7.42292702e-01 -5.80806509e-02 9.30667400e-01 4.16057289e-01 -5.22353292e-01 -9.42891240e-01 -8.60109329e-01 -2.33412951e-01 1.34929597e+00 3.81754220e-01 -3.31528395e-01 -7.56633222e-01 -7.78500855e-01 -1.83404028e-01 2.96698093e-01 -5.28105915e-01 3.36384714e-01 -8.84457588e-01 -9.92254555e-01 8.06621134e-01 5.10877311e-01 9.03467715e-01 -1.25943220e+00 -2.82557070e-01 6.50150180e-02 -2.56200522e-01 -1.38482368e+00 -7.82351613e-01 1.82376340e-01 -8.63651335e-01 -7.31961489e-01 -9.82966542e-01 -1.42519510e+00 7.91963935e-01 3.00137669e-01 7.80025601e-01 2.14640036e-01 -3.58862579e-01 -3.28123868e-02 -4.97650474e-01 -1.50867745e-01 -2.67536312e-01 3.63796860e-01 -5.62797427e-01 2.69820869e-01 3.29797238e-01 7.20729604e-02 -5.04041612e-01 5.79770386e-01 -1.02373314e+00 4.00323123e-01 1.00191092e+00 8.28351080e-01 7.79203713e-01 1.88063338e-01 2.04033822e-01 -8.07255924e-01 4.35032845e-01 1.01947337e-01 -6.60814047e-01 5.15768766e-01 -4.67507243e-01 -5.38507886e-02 8.99188817e-01 -2.48740986e-01 -1.11853564e+00 4.83029544e-01 -4.61223125e-01 -1.67538792e-01 -3.32959950e-01 3.35787058e-01 -3.71083587e-01 1.51698515e-01 4.06625628e-01 1.15118551e+00 -4.35383976e-01 -5.37057698e-01 -8.71076733e-02 1.12659597e+00 4.78629023e-01 -3.49833012e-01 9.67603505e-01 6.52149081e-01 -2.79152274e-01 -1.23191559e+00 -6.96742058e-01 -6.77852035e-01 -7.27858126e-01 1.43025070e-01 1.12461138e+00 -6.16189182e-01 -6.67751670e-01 1.05605853e+00 -1.13694322e+00 -4.30616915e-01 1.94100574e-01 7.45528191e-02 -5.13138711e-01 8.32854569e-01 -7.93919206e-01 -6.85957849e-01 -8.00789475e-01 -1.23564589e+00 1.92783880e+00 2.18838230e-01 3.59439284e-01 -9.54810083e-01 -1.45603731e-01 5.64157367e-01 1.10194437e-01 -1.05693221e-01 7.71399736e-01 -6.46143436e-01 -6.85440540e-01 -3.41657788e-01 -5.66036344e-01 -7.32510909e-02 -3.38461027e-02 1.68605223e-01 -9.54192698e-01 -2.29065076e-01 -7.07924441e-02 -3.00707608e-01 1.19254422e+00 3.42390090e-01 1.28506303e+00 -1.33391470e-02 -6.22506976e-01 6.51920557e-01 1.17390835e+00 7.68518355e-03 5.64971328e-01 4.19046670e-01 1.04823244e+00 3.29767048e-01 5.57719648e-01 4.20954496e-01 4.40167159e-01 5.07703125e-01 1.98523894e-01 -3.02408218e-01 -3.36088724e-02 -1.40914544e-01 3.49205643e-01 9.86812890e-01 4.59076494e-01 -7.42751896e-01 -1.13031971e+00 2.40573660e-01 -1.90804482e+00 -5.28192520e-01 -5.16961813e-01 1.71932995e+00 6.16326272e-01 5.30217767e-01 -2.22234745e-02 3.00161421e-01 9.39444244e-01 2.22444907e-01 -7.31256902e-01 5.00909016e-02 -3.89177084e-01 -2.08224669e-01 4.83260363e-01 1.22710712e-01 -1.38997674e+00 1.38176000e+00 4.91581011e+00 1.20582700e+00 -1.23511147e+00 -1.92762002e-01 6.57253921e-01 4.57872510e-01 1.62589028e-01 -3.22616547e-01 -9.07967746e-01 4.06935245e-01 3.58825922e-02 2.07217082e-01 1.59974381e-01 6.28722072e-01 2.07839057e-01 1.97350397e-03 -8.34720969e-01 1.27862346e+00 3.57990205e-01 -1.09429491e+00 2.57032186e-01 -2.05079958e-01 5.01023829e-01 2.39513576e-01 1.07519738e-01 2.72082478e-01 -1.21621571e-01 -1.05087113e+00 8.32875371e-01 3.72434109e-01 7.05866992e-01 -4.42036837e-01 5.81059277e-01 7.32134819e-01 -1.52890539e+00 2.79569089e-01 -7.17238784e-01 5.01799822e-01 -1.10605039e-01 5.81281126e-01 -9.72787738e-01 4.80675787e-01 5.18540025e-01 1.02296007e+00 -7.64273465e-01 1.18666947e+00 -2.04358339e-01 8.73432577e-01 -3.83788824e-01 -4.68939960e-01 3.44833285e-01 -2.45840535e-01 5.40548205e-01 1.41754389e+00 1.49792537e-01 -3.96332413e-01 3.98203582e-01 8.39065254e-01 -2.15690508e-01 5.48216283e-01 -2.17290670e-01 -5.21644093e-02 9.36875567e-02 1.45789278e+00 -1.52959156e+00 -3.04406255e-01 -3.02117437e-01 1.41711748e+00 2.17394102e-02 1.82251796e-01 -7.99394429e-01 -6.67362452e-01 -4.27504152e-01 -2.47531291e-02 5.06695807e-01 -1.39184937e-01 -8.51390004e-01 -1.48279738e+00 4.16197509e-01 -9.19333458e-01 3.89053345e-01 -8.30003083e-01 -1.13248444e+00 5.41661263e-01 -8.45582426e-01 -1.24533415e+00 3.56195122e-01 -7.36410737e-01 -8.34657490e-01 6.07332349e-01 -1.33479333e+00 -1.48675776e+00 -4.44548517e-01 4.90787655e-01 1.33462167e+00 -1.49649218e-01 4.22286361e-01 2.15621784e-01 -1.17269862e+00 7.45999753e-01 4.69788820e-01 5.14831245e-01 8.01449597e-01 -1.40815043e+00 8.05310309e-01 8.80176723e-01 1.90205067e-01 1.55259058e-01 4.75185037e-01 -9.27487791e-01 -1.78449762e+00 -1.23492491e+00 5.54537296e-01 -5.30498207e-01 6.20698988e-01 -7.50621974e-01 -1.00906301e+00 7.19837546e-01 9.04281214e-02 -3.30631465e-01 3.43346060e-03 -2.46565461e-01 4.08506095e-02 1.83141500e-01 -7.38796711e-01 8.03928792e-01 7.93100119e-01 -2.38240615e-01 -6.12295449e-01 6.53895974e-01 4.98511970e-01 -7.39847422e-01 -3.50570142e-01 1.92495540e-01 3.35329354e-01 -7.27484882e-01 5.39238513e-01 3.08895875e-02 3.38541150e-01 -2.65746504e-01 5.63186891e-02 -6.88782156e-01 1.82970628e-01 -6.74242318e-01 1.67634651e-01 1.22342932e+00 4.41165745e-01 -7.17079043e-01 1.20636046e+00 -2.77704019e-02 -3.74826521e-01 -8.29905927e-01 -7.53977716e-01 -4.11404401e-01 9.92898792e-02 -3.96438926e-01 3.92499954e-01 8.75536025e-01 -2.38604695e-01 5.68380058e-01 -3.15412611e-01 1.26022905e-01 5.18945277e-01 4.11038518e-01 7.94713736e-01 -1.21710289e+00 -1.66962549e-01 -8.18211377e-01 -1.92206666e-01 -1.98440731e+00 -8.58406723e-02 -8.94720197e-01 5.18093109e-01 -1.75885773e+00 2.81210810e-01 -4.43931252e-01 2.68784285e-01 4.38769668e-01 -4.13556635e-01 3.02310549e-02 -1.23370789e-01 5.52042365e-01 -1.04189265e+00 7.74225891e-01 1.35348201e+00 -3.34066033e-01 -2.19494045e-01 1.35473952e-01 -3.31673950e-01 9.10384655e-01 7.20659316e-01 -3.02380294e-01 -1.59127265e-02 -7.68481731e-01 3.17496479e-01 4.69161086e-02 1.85506448e-01 -1.01332104e+00 6.52133107e-01 1.84565075e-02 5.34642756e-01 -1.51965928e+00 4.55503017e-02 -7.21602261e-01 -6.36731207e-01 2.98155367e-01 -2.53129870e-01 -1.74954653e-01 7.37726390e-02 6.72112107e-01 -3.37288133e-03 -3.85111123e-01 6.62907660e-01 3.22832614e-01 -4.12638187e-01 3.59693885e-01 -4.90750015e-01 -1.24001103e-02 7.57339120e-01 -4.73095685e-01 -4.65063483e-01 3.27224955e-02 -3.57147217e-01 5.17267644e-01 5.09352624e-01 4.51450765e-01 7.88406730e-01 -7.48641849e-01 -8.58756125e-01 2.89982319e-01 1.12931065e-01 5.32672405e-01 9.33928266e-02 9.31681216e-01 -9.07923520e-01 5.71291625e-01 4.85539556e-01 -1.09219861e+00 -1.17627442e+00 2.63863325e-01 4.87501025e-01 -4.00226086e-01 -1.06507576e+00 7.46168375e-01 3.62263471e-01 -6.20412707e-01 5.47412038e-01 -3.97787362e-01 -1.36607438e-01 -2.15967670e-01 2.35452861e-01 2.24326983e-01 2.92718470e-01 -5.89968622e-01 -1.76457405e-01 1.12652743e+00 -3.99798751e-01 2.10140139e-01 1.11971676e+00 -3.30417246e-01 -2.67378744e-02 2.50618875e-01 8.86868834e-01 6.40599579e-02 -1.11534214e+00 -3.77167970e-01 2.90293396e-01 -2.15489522e-01 3.37246768e-02 -7.17983544e-01 -9.23147440e-01 1.17302179e+00 4.52033371e-01 3.75328124e-01 1.05697823e+00 -1.42838523e-01 1.25808895e+00 8.43085945e-01 -1.67120442e-01 -1.34407771e+00 2.74392962e-01 7.55760491e-01 6.31897867e-01 -1.23712564e+00 2.64003146e-02 -5.90035200e-01 -6.03125215e-01 1.45650887e+00 6.21541858e-01 -2.56004240e-02 2.50334918e-01 3.33190501e-01 2.51028724e-02 -3.37376386e-01 -4.93943393e-01 -1.85024261e-01 4.62727934e-01 -1.80670433e-02 4.38778996e-01 -1.66974857e-01 -1.30215511e-01 3.57267022e-01 -3.37885600e-03 -3.08644503e-01 3.03399503e-01 8.29618216e-01 -9.13541496e-01 -7.61951923e-01 -6.56377017e-01 9.13174510e-01 -6.51801646e-01 -9.27056819e-02 -6.38173997e-01 5.01246095e-01 -1.56991705e-01 8.23256850e-01 1.00960657e-01 -2.58814514e-01 3.28302681e-01 -1.18650392e-01 7.69562670e-04 -6.13277972e-01 -4.88586366e-01 6.45119488e-01 -2.56413132e-01 -2.34028567e-02 5.30317575e-02 -6.58918858e-01 -1.66158223e+00 -1.57127097e-01 -8.51135433e-01 -2.18148842e-01 6.20513737e-01 1.04096651e+00 8.05998743e-02 5.30745029e-01 6.64542735e-01 -5.43567598e-01 -5.39345026e-01 -1.00717521e+00 -4.78883445e-01 1.18156098e-01 5.53551950e-02 -1.77264780e-01 -2.38224283e-01 4.09665972e-01]
[12.020853042602539, 2.281815528869629]
1d0f2851-7e7c-4f2a-bc28-a7def2ca7f1c
integrating-visuospatial-linguistic-and
2110.10834
null
https://arxiv.org/abs/2110.10834v1
https://arxiv.org/pdf/2110.10834v1.pdf
Integrating Visuospatial, Linguistic and Commonsense Structure into Story Visualization
While much research has been done in text-to-image synthesis, little work has been done to explore the usage of linguistic structure of the input text. Such information is even more important for story visualization since its inputs have an explicit narrative structure that needs to be translated into an image sequence (or visual story). Prior work in this domain has shown that there is ample room for improvement in the generated image sequence in terms of visual quality, consistency and relevance. In this paper, we first explore the use of constituency parse trees using a Transformer-based recurrent architecture for encoding structured input. Second, we augment the structured input with commonsense information and study the impact of this external knowledge on the generation of visual story. Third, we also incorporate visual structure via bounding boxes and dense captioning to provide feedback about the characters/objects in generated images within a dual learning setup. We show that off-the-shelf dense-captioning models trained on Visual Genome can improve the spatial structure of images from a different target domain without needing fine-tuning. We train the model end-to-end using intra-story contrastive loss (between words and image sub-regions) and show significant improvements in several metrics (and human evaluation) for multiple datasets. Finally, we provide an analysis of the linguistic and visuo-spatial information. Code and data: https://github.com/adymaharana/VLCStoryGan.
['Mohit Bansal', 'Adyasha Maharana']
2021-10-21
null
null
null
null
['dense-captioning', 'story-visualization']
['computer-vision', 'computer-vision']
[ 5.31200230e-01 3.58795196e-01 8.61161500e-02 -3.58146071e-01 -8.41696382e-01 -7.42998719e-01 8.73723805e-01 -1.84028372e-01 -6.21621422e-02 6.77559495e-01 8.35519254e-01 -2.47620553e-01 5.35781205e-01 -7.66412795e-01 -1.06919527e+00 -4.99473304e-01 3.65376234e-01 2.80754298e-01 1.42598376e-01 -4.20551389e-01 1.24644637e-01 1.71706066e-01 -1.38344884e+00 1.00928378e+00 5.06009102e-01 6.37028575e-01 5.65950096e-01 8.64825428e-01 -8.25942233e-02 1.21753991e+00 -6.18211627e-01 -5.21177471e-01 2.47234970e-01 -9.83606100e-01 -8.66841495e-01 2.94408530e-01 7.14104712e-01 -4.75428760e-01 -4.84066069e-01 6.58863187e-01 6.45346880e-01 4.85455692e-02 6.44011617e-01 -9.97178972e-01 -1.07313347e+00 7.41788030e-01 -6.81228161e-01 2.16188759e-01 4.14837420e-01 5.43126404e-01 9.88263905e-01 -9.55051720e-01 1.15129733e+00 1.38601232e+00 2.31875241e-01 4.95526999e-01 -1.47729838e+00 -5.88323534e-01 7.35067576e-02 1.84365511e-01 -9.05457139e-01 -5.53259492e-01 7.51249015e-01 -5.67525148e-01 1.13052118e+00 1.45435184e-01 5.92117131e-01 1.49685168e+00 -5.42654470e-02 8.80319655e-01 1.10896802e+00 -7.02441990e-01 -1.70641601e-01 2.67183393e-01 -4.10979718e-01 6.55650139e-01 -2.09546983e-01 1.88432366e-01 -6.17878973e-01 5.02166629e-01 8.75423968e-01 -3.90367568e-01 -1.73101634e-01 -4.75287616e-01 -1.13029337e+00 9.09150302e-01 6.75815701e-01 2.58458048e-01 -1.94389790e-01 5.44559658e-01 4.74656999e-01 1.15216658e-01 5.05403817e-01 5.99005938e-01 5.96197732e-02 -9.11885202e-02 -9.87158418e-01 3.62583905e-01 2.73559451e-01 8.23459387e-01 4.38126653e-01 3.06480974e-01 -6.16075993e-01 8.27613056e-01 -7.13084191e-02 4.11001265e-01 2.42671818e-01 -1.04195404e+00 7.90740490e-01 2.71029413e-01 2.22013649e-02 -1.04029191e+00 -4.75334600e-02 -3.34390223e-01 -5.17582774e-01 5.13550401e-01 3.19431514e-01 -1.60144776e-01 -1.07009375e+00 1.85630059e+00 4.85210791e-02 -6.54545352e-02 5.12756929e-02 1.02644384e+00 9.15180087e-01 1.00121415e+00 1.58126399e-01 2.31968060e-01 1.36641753e+00 -1.24183750e+00 -7.28409827e-01 -5.88385165e-01 4.86124545e-01 -8.56949389e-01 1.50176740e+00 1.16120659e-01 -1.54774165e+00 -3.68477911e-01 -9.37704682e-01 -5.36916375e-01 -3.08377385e-01 1.61078557e-01 3.57661784e-01 2.00468495e-01 -1.19490016e+00 4.42580968e-01 -5.47557890e-01 -2.85667300e-01 6.29886806e-01 -1.57010719e-01 -2.99973518e-01 -1.94189683e-01 -1.05687749e+00 1.11818218e+00 4.12092566e-01 -2.71043211e-01 -9.43498373e-01 -7.73414135e-01 -1.13149488e+00 -1.42636849e-02 2.89870918e-01 -8.79930437e-01 1.31164968e+00 -1.35726964e+00 -1.22640741e+00 9.21772718e-01 -1.00773968e-01 -4.32489961e-01 6.13497853e-01 -6.70453012e-02 7.66158327e-02 4.01814193e-01 1.95795938e-01 1.23894560e+00 8.19943428e-01 -1.46337974e+00 -4.60051060e-01 -1.18699707e-02 2.12519750e-01 5.10942280e-01 -1.17258385e-01 1.41916648e-01 -4.05336857e-01 -1.20354176e+00 -5.30400276e-01 -8.32134783e-01 -2.30436400e-03 1.05929404e-01 -3.68274063e-01 1.79984838e-01 7.79997468e-01 -1.12755013e+00 9.58078325e-01 -2.00716352e+00 2.83055454e-01 -2.10036069e-01 -1.52639477e-02 5.43854060e-03 -4.58860189e-01 5.40370703e-01 -2.06588835e-01 2.43320614e-01 -4.47193444e-01 -5.30289829e-01 -1.82806417e-01 -4.25760485e-02 -7.04602063e-01 6.92506209e-02 4.76601839e-01 1.29890645e+00 -8.79029751e-01 -4.69631493e-01 4.20994014e-01 8.00615668e-01 -6.63845837e-01 1.85608432e-01 -4.17658836e-01 3.60528290e-01 5.86758628e-02 2.82499045e-01 2.22782761e-01 -3.75068516e-01 -5.70023954e-02 -1.93464428e-01 -3.01659331e-02 5.34441054e-01 -7.09826171e-01 1.93220603e+00 -7.91237473e-01 1.20138717e+00 -2.97039539e-01 -6.24661565e-01 5.14870226e-01 2.18142077e-01 -8.99952184e-03 -1.11264265e+00 7.84908235e-02 -1.13722026e-01 -1.31283179e-01 -4.16649014e-01 6.43462241e-01 -2.32536271e-01 -8.20703283e-02 6.97735786e-01 4.63536941e-02 -2.71243393e-01 3.44697684e-01 4.14191037e-01 9.01970923e-01 5.12611628e-01 -2.87645534e-02 2.23977089e-01 -5.26091345e-02 3.06027621e-01 -1.12988062e-01 4.53473747e-01 4.16730046e-01 1.28465068e+00 5.07985592e-01 -1.50421798e-01 -1.64422810e+00 -1.06835294e+00 2.01472670e-01 1.21639597e+00 -2.69705355e-02 -2.72319913e-01 -8.21997583e-01 -3.76717478e-01 -3.68965864e-01 1.19848692e+00 -8.04375231e-01 -1.02081396e-01 -6.99130893e-01 -2.98075378e-01 5.35810113e-01 6.88856363e-01 3.75879705e-01 -1.55023539e+00 -1.12454331e+00 7.87204951e-02 -3.88864040e-01 -1.22649217e+00 -6.51202202e-01 -1.50049990e-02 -5.46089768e-01 -6.89485192e-01 -9.00667846e-01 -7.05816388e-01 6.15297079e-01 2.61939764e-01 1.32873726e+00 -2.55044222e-01 -4.59511369e-01 3.21208388e-01 -4.19526786e-01 -2.25671411e-01 -6.07097387e-01 -3.58893164e-02 -6.85778558e-01 -1.83654189e-01 -3.36648285e-01 -5.07008135e-01 -6.24764800e-01 -2.58038170e-03 -1.06757402e+00 9.73879755e-01 6.47850871e-01 8.18480492e-01 3.50372910e-01 -4.55373198e-01 2.26673439e-01 -8.19006920e-01 7.93221116e-01 -3.28560174e-01 -2.49218404e-01 1.36721417e-01 -1.86141089e-01 2.93706566e-01 4.85904276e-01 -3.08663666e-01 -1.20020747e+00 6.93851262e-02 1.06618844e-01 -5.46017289e-01 2.06746198e-02 3.42647612e-01 -9.90581792e-03 3.66904616e-01 7.39122212e-01 1.89749360e-01 -5.50513994e-03 -1.60291716e-01 8.16709995e-01 3.44854146e-01 6.85324609e-01 -4.26584542e-01 6.39861226e-01 5.20422578e-01 -1.96645766e-01 -5.95484734e-01 -8.16510677e-01 9.39795822e-02 -3.71902198e-01 -2.40049079e-01 1.04728639e+00 -1.02472925e+00 -1.82263270e-01 1.23650886e-01 -1.41802394e+00 -8.13256919e-01 -5.83649039e-01 4.94226366e-02 -8.14851940e-01 3.68635617e-02 -5.04584312e-01 -5.43453574e-01 -3.29607785e-01 -1.15463006e+00 1.15607798e+00 -9.24538672e-02 -4.75164473e-01 -9.00960028e-01 -3.94275859e-02 4.68320996e-01 3.49859327e-01 6.27293229e-01 8.33135247e-01 -7.85955414e-02 -8.07232082e-01 2.11043820e-01 -5.24130464e-01 1.28036395e-01 -8.60400423e-02 -1.94451541e-01 -1.01453602e+00 -6.83532506e-02 -3.77605498e-01 -6.69581234e-01 9.83878076e-01 4.34781253e-01 9.60293770e-01 -5.36558330e-01 -5.53908348e-02 5.87016106e-01 1.28877008e+00 2.84599029e-02 9.18137848e-01 3.83779198e-01 9.66654778e-01 9.47607934e-01 4.00318444e-01 2.41398111e-01 3.97112072e-01 8.64990413e-01 4.01775002e-01 -5.24961352e-01 -8.94975185e-01 -7.24745750e-01 5.76009572e-01 2.07192704e-01 2.01642796e-01 -5.01444399e-01 -8.81150961e-01 8.97643209e-01 -1.70871735e+00 -1.34826255e+00 3.19622457e-02 1.74819744e+00 1.08180761e+00 -1.42991990e-02 1.30181879e-01 -3.81643772e-02 6.42986119e-01 3.89948398e-01 -4.43168432e-01 -6.84031844e-01 -3.14287275e-01 1.51558831e-01 5.01350522e-01 6.62552714e-01 -8.26791286e-01 1.17495763e+00 5.84849644e+00 7.83041477e-01 -1.29102588e+00 1.18553601e-01 9.61968482e-01 -5.07107377e-01 -6.35821342e-01 6.14213292e-03 -3.31707478e-01 3.34854752e-01 7.57556796e-01 3.90124246e-02 5.29766679e-01 5.27688622e-01 4.34166998e-01 -2.96339095e-01 -1.01621294e+00 9.66079295e-01 2.82633394e-01 -1.74635887e+00 2.46189117e-01 3.19364257e-02 8.22018921e-01 -1.36668399e-01 3.75745088e-01 1.89964473e-02 4.72878337e-01 -1.33674824e+00 1.36602914e+00 3.13143551e-01 1.16960788e+00 -7.28238285e-01 2.77358860e-01 2.29795575e-02 -8.55499744e-01 9.16946009e-02 -4.17421758e-02 -1.29582047e-01 4.63927448e-01 2.58848459e-01 -1.10187435e+00 1.19407140e-01 6.06795073e-01 7.24429131e-01 -7.35138118e-01 6.67762935e-01 -5.24040222e-01 5.25459886e-01 -1.09582230e-01 1.39926866e-01 4.47437912e-01 1.36414230e-01 4.79561687e-01 1.49298596e+00 3.07976127e-01 6.19157590e-02 -2.23412186e-01 1.24221122e+00 -1.61562532e-01 3.98761742e-02 -8.63314271e-01 -1.25085980e-01 1.23271555e-01 9.29781914e-01 -7.40530014e-01 -3.20208848e-01 -2.06484675e-01 1.26660097e+00 3.08922708e-01 4.73277539e-01 -8.56619835e-01 -2.89430112e-01 4.15113807e-01 6.12889171e-01 6.04550362e-01 -6.41666204e-02 -6.38365030e-01 -9.34317112e-01 1.38568312e-01 -9.01954651e-01 1.60232961e-01 -1.48553061e+00 -8.72260630e-01 6.30709350e-01 1.34224147e-01 -9.01489675e-01 -5.61953962e-01 -3.16453636e-01 -5.14552772e-01 9.26454723e-01 -1.30681264e+00 -1.36818290e+00 -2.77552485e-01 3.16694170e-01 9.23160255e-01 2.99487114e-02 5.20668924e-01 1.77103672e-02 -1.87526286e-01 6.12408996e-01 -1.96433008e-01 2.25510076e-01 7.68095195e-01 -1.12227070e+00 7.07419872e-01 9.91143405e-01 3.13224077e-01 1.27746046e-01 9.32390988e-01 -6.65986955e-01 -9.54206228e-01 -1.19283938e+00 8.09400916e-01 -6.68999076e-01 3.52156341e-01 -6.64116502e-01 -6.66887999e-01 6.86280310e-01 9.10435677e-01 -1.80729061e-01 4.42929953e-01 -3.05198669e-01 -5.59512496e-01 3.11100304e-01 -9.41223741e-01 9.87855256e-01 1.11121333e+00 -5.31922460e-01 -5.08228898e-01 1.61541939e-01 9.64606404e-01 -5.28443635e-01 -3.17872018e-01 7.13108703e-02 5.35245717e-01 -1.10694921e+00 1.07692385e+00 -5.36436439e-01 1.35669279e+00 -2.57697910e-01 -6.81077465e-02 -1.32599247e+00 -2.94276893e-01 -4.55117315e-01 1.76031336e-01 1.19380200e+00 6.28214240e-01 -3.36242244e-02 6.21604741e-01 5.16853213e-01 -2.29711175e-01 -7.11952567e-01 -4.81677443e-01 -3.34316880e-01 1.64035544e-01 -4.10695642e-01 2.78891683e-01 8.44294071e-01 -1.29467979e-01 7.12856114e-01 -5.65443218e-01 -3.41820329e-01 2.45519742e-01 1.62992120e-01 7.55125821e-01 -4.61225808e-01 -3.44192088e-01 -5.43518960e-01 -1.23332940e-01 -7.90160954e-01 9.19321924e-02 -1.03550112e+00 6.98308786e-03 -2.13173771e+00 2.80313820e-01 -7.15583339e-02 1.77355558e-01 7.22960413e-01 -3.51173766e-02 6.08371198e-01 6.94202006e-01 1.04198955e-01 -4.75790739e-01 5.62036455e-01 1.53899038e+00 -1.39788806e-01 -9.67509151e-02 -7.29489923e-01 -8.21193159e-01 3.75131160e-01 7.19612718e-01 -3.93346786e-01 -5.89927733e-01 -8.29520583e-01 3.03530216e-01 1.52284622e-01 7.37012327e-01 -8.05987716e-01 -1.63602427e-01 -1.36686176e-01 6.27174139e-01 -5.07342577e-01 5.47647893e-01 -4.94147688e-01 1.56374291e-01 2.14509368e-01 -8.09869111e-01 2.26838186e-01 3.96460146e-01 3.20590675e-01 -5.44128604e-02 -1.48570407e-02 8.33118320e-01 -2.68091261e-01 -6.04043007e-01 -1.79637760e-01 -2.65118510e-01 3.37176859e-01 9.40820873e-01 -4.02977079e-01 -3.06492567e-01 -7.60901213e-01 -5.36003947e-01 1.12037271e-01 7.86903501e-01 6.00323021e-01 7.44148672e-01 -1.41502130e+00 -1.03373516e+00 -1.52072668e-01 9.23076347e-02 -9.20742005e-02 2.29414299e-01 3.91861200e-01 -5.93300521e-01 3.54428321e-01 -4.49729085e-01 -4.95795488e-01 -1.18063450e+00 5.06473243e-01 1.97101668e-01 -3.03082556e-01 -7.63783872e-01 8.12147021e-01 5.99354386e-01 -2.69945674e-02 1.32660806e-01 -2.19590187e-01 3.75818834e-02 1.76696107e-01 5.70459485e-01 -3.08776069e-02 -2.41484150e-01 -7.53224015e-01 5.55604473e-02 4.74269301e-01 -6.36286987e-03 -7.56412089e-01 1.38520849e+00 -1.33839279e-01 2.42096931e-01 4.04196143e-01 1.19685507e+00 -1.40129045e-01 -1.62980080e+00 -4.90483418e-02 -2.86790192e-01 -5.19710600e-01 2.81856097e-02 -1.19200420e+00 -1.03370702e+00 1.11511254e+00 4.43313807e-01 -1.75871491e-01 1.06847572e+00 1.64015993e-01 7.93661296e-01 -6.32844642e-02 -7.02015162e-02 -9.58281040e-01 5.38463414e-01 4.31865960e-01 1.48163486e+00 -1.18120813e+00 -6.84285387e-02 -1.87871277e-01 -1.24057198e+00 7.74759650e-01 4.55125451e-01 -1.59440532e-01 2.93886922e-02 2.60224879e-01 7.99766481e-02 -9.68002900e-02 -9.02262926e-01 -2.73067176e-01 3.24627072e-01 7.19217777e-01 5.78200579e-01 -1.32201582e-01 8.47918242e-02 3.31580639e-01 -5.55488408e-01 -9.72063839e-02 7.07645178e-01 7.06461728e-01 -1.73495561e-01 -8.73906136e-01 -3.48180920e-01 2.34077036e-01 -3.78746867e-01 -4.52020854e-01 -5.76579213e-01 5.95248103e-01 -6.16421085e-03 7.32311010e-01 2.32870236e-01 -1.01872422e-01 2.48179048e-01 9.85869323e-04 7.57044017e-01 -8.24543536e-01 -6.42395556e-01 7.12732896e-02 2.82350838e-01 -4.26648617e-01 -2.71738529e-01 -6.28566980e-01 -1.29043472e+00 -1.31426811e-01 1.84601501e-01 -3.29323947e-01 7.13031232e-01 6.43557191e-01 5.15205562e-01 8.29326034e-01 2.42231995e-01 -1.07580841e+00 8.53355676e-02 -8.41266811e-01 4.42552045e-02 5.64461172e-01 3.11369926e-01 -5.11135697e-01 -1.27443731e-01 4.87350374e-01]
[11.167121887207031, 0.7110996246337891]
8a4160ed-739c-48e7-8664-a4c4d4849bbf
the-use-of-mutual-coherence-to-prove-ell1ell0
1901.02783
null
http://arxiv.org/abs/1901.02783v1
http://arxiv.org/pdf/1901.02783v1.pdf
The Use of Mutual Coherence to Prove $\ell^1/\ell^0$-Equivalence in Classification Problems
We consider the decomposition of a signal over an overcomplete set of vectors. Minimization of the $\ell^1$-norm of the coefficient vector can often retrieve the sparsest solution (so-called "$\ell^1/\ell^0$-equivalence"), a generally NP-hard task, and this fact has powered the field of compressed sensing. Wright et al.'s sparse representation-based classification (SRC) applies this relationship to machine learning, wherein the signal to be decomposed represents the test sample and columns of the dictionary are training samples. We investigate the relationships between $\ell^1$-minimization, sparsity, and classification accuracy in SRC. After proving that the tractable, deterministic approach to verifying $\ell^1/\ell^0$-equivalence fundamentally conflicts with the high coherence between same-class training samples, we demonstrate that $\ell^1$-minimization can still recover the sparsest solution when the classes are well-separated. Further, using a nonlinear transform so that sparse recovery conditions may be satisfied, we demonstrate that approximate (not strict) equivalence is key to the success of SRC.
['Chelsea Weaver', 'Naoki Saito']
2019-01-09
null
null
null
null
['sparse-representation-based-classification']
['computer-vision']
[ 5.64745367e-01 1.90321475e-01 -4.03314292e-01 -1.54796988e-01 -1.04881930e+00 -5.00231922e-01 4.08850200e-02 -1.72083825e-01 4.98485379e-02 7.99429357e-01 2.01763347e-01 -2.90202737e-01 -5.84288895e-01 -6.78228498e-01 -7.27203906e-01 -9.87528622e-01 -5.49444258e-01 4.44246344e-02 -5.66397250e-01 -2.11666271e-01 1.80133298e-01 3.75965595e-01 -1.61055720e+00 2.34025896e-01 5.25336921e-01 1.23086286e+00 -2.05389753e-01 4.67524827e-01 2.37463728e-01 8.04915488e-01 -4.34182197e-01 -1.87973157e-02 6.50815189e-01 -7.97276735e-01 -7.43701696e-01 1.03375427e-01 5.78349829e-01 1.78393930e-01 -2.13138387e-01 1.57679474e+00 6.60894290e-02 1.01154879e-01 6.79018617e-01 -1.30325949e+00 -5.85579038e-01 3.90810758e-01 -5.73465645e-01 2.18379647e-01 5.35993457e-01 -3.79389048e-01 1.24094260e+00 -8.86318803e-01 5.84117413e-01 7.37005770e-01 9.12177145e-01 2.13685967e-02 -1.33035946e+00 -6.47726357e-01 -1.14557624e-01 4.98960260e-03 -1.76767921e+00 -7.35849380e-01 8.18287492e-01 -4.89086270e-01 6.27029419e-01 5.28495312e-01 5.82249284e-01 5.47631085e-01 4.64479886e-02 6.65991783e-01 9.77652669e-01 -7.49667823e-01 3.41863692e-01 1.34071810e-02 2.41565242e-01 8.27337921e-01 4.29638207e-01 1.97924882e-01 -5.98226964e-01 -2.90484339e-01 6.06527448e-01 1.68055501e-02 -7.75704384e-01 -2.74119914e-01 -1.09655654e+00 1.21047115e+00 1.96489602e-01 6.16068542e-01 -2.31218323e-01 -1.62207037e-01 7.69779980e-02 7.15354443e-01 2.01112509e-01 5.74463129e-01 5.43216392e-02 2.20018893e-01 -1.34013784e+00 9.80795454e-03 9.38985348e-01 9.38154817e-01 8.68106186e-01 4.74294037e-01 2.79015243e-01 6.53582573e-01 -8.35911185e-02 7.87062824e-01 2.57539421e-01 -1.25188351e+00 3.50149900e-01 1.92098081e-01 1.31716607e-02 -1.44351888e+00 -2.39505805e-02 -1.02769554e+00 -1.18923676e+00 -1.03727460e-01 3.26702923e-01 1.21816881e-01 -3.33964556e-01 2.08034515e+00 4.07088101e-02 3.58444810e-01 3.46393615e-01 8.40537190e-01 2.92233795e-01 6.04829431e-01 -5.89099228e-01 -8.73358190e-01 8.92537892e-01 -4.03345734e-01 -4.30696309e-01 -2.68858016e-01 6.41312957e-01 -8.02682996e-01 6.21726930e-01 4.46231812e-01 -1.20504546e+00 -3.14589351e-01 -1.32164776e+00 2.30388850e-01 2.25929812e-01 -1.16798699e-01 7.09475517e-01 6.58103883e-01 -8.11738014e-01 4.29391712e-01 -3.72444630e-01 5.68711339e-03 3.04398209e-01 3.76474291e-01 -7.70286560e-01 -7.64835477e-01 -7.75865674e-01 5.80029607e-01 -1.89401656e-01 1.18430443e-02 -6.15199625e-01 -7.43623137e-01 -9.61267769e-01 8.83773640e-02 9.30750892e-02 -3.29313695e-01 7.25515664e-01 -9.92064774e-01 -5.85374594e-01 9.01418328e-01 -6.80231333e-01 -3.74805182e-01 1.09268930e-02 1.73914745e-01 -4.94402230e-01 4.06207621e-01 5.65977335e-01 -1.78650208e-02 1.11920714e+00 -1.21135831e+00 -6.23546064e-01 -4.85877872e-01 -2.76989818e-01 -1.03141777e-01 -1.73262566e-01 -3.90968323e-01 1.37970850e-01 -7.08018124e-01 9.08402383e-01 -7.28560209e-01 -1.85227320e-01 -3.52413386e-01 -2.66505659e-01 2.17925400e-01 5.71780384e-01 -4.92365539e-01 1.06354296e+00 -2.60032725e+00 2.17370197e-01 8.17876935e-01 3.68536025e-01 -1.45125955e-01 -3.58204246e-01 4.48825926e-01 -5.26908517e-01 -1.47594854e-01 -5.51754534e-01 -1.17565192e-01 -1.74602568e-01 1.37520745e-01 -5.78123808e-01 1.14598882e+00 -6.30291477e-02 3.51545185e-01 -7.56411433e-01 -1.99533448e-01 -4.18576188e-02 4.79425877e-01 -6.23622417e-01 -1.34439304e-01 1.22668274e-01 3.15868288e-01 -2.55424052e-01 9.22525883e-01 7.58708715e-01 -4.62429017e-01 2.90047854e-01 -3.59332234e-01 1.68738991e-01 -1.92931280e-01 -1.50529778e+00 1.40332854e+00 -1.83002368e-01 7.59807348e-01 6.50631905e-01 -1.78164935e+00 8.75844479e-01 3.13484460e-01 9.65919733e-01 -8.12846243e-01 -4.06799875e-02 5.08065164e-01 -6.45458251e-02 -2.96544760e-01 1.20520785e-01 -7.76472688e-01 -2.62683164e-02 4.99636918e-01 -1.03270359e-01 -1.84858143e-01 2.43135095e-01 2.03877598e-01 1.30494177e+00 -5.64667225e-01 4.34081644e-01 -7.51865268e-01 4.52966809e-01 2.68892378e-01 7.44688511e-01 7.90516376e-01 2.50618644e-02 6.18175924e-01 3.44294965e-01 -2.79069245e-01 -8.91745687e-01 -9.44494665e-01 -4.49822247e-01 7.16596365e-01 3.04346392e-03 -2.19600707e-01 -4.95796680e-01 -2.53477871e-01 5.93005642e-02 4.76853848e-01 -6.12074912e-01 -2.06423044e-01 -4.57059801e-01 -6.94766819e-01 4.46556836e-01 2.80515999e-01 3.72822285e-01 -2.81463027e-01 -4.76503909e-01 1.42292842e-01 -3.67856503e-01 -1.03547132e+00 -3.77867281e-01 4.61386979e-01 -9.34307158e-01 -1.14592528e+00 -7.71811843e-01 -8.13432634e-01 9.79424596e-01 6.02884293e-01 1.12836516e+00 2.28494227e-01 -2.05292240e-01 5.66284716e-01 -4.88723129e-01 -3.05460151e-02 -1.15981631e-01 -3.48528773e-01 1.00240692e-01 2.05412388e-01 1.45838305e-01 -1.08772695e+00 -3.01406294e-01 1.91594586e-01 -8.43810260e-01 -3.18257987e-01 4.29725975e-01 1.00454628e+00 9.04827058e-01 3.80096972e-01 4.93793011e-01 -7.10137129e-01 3.79528970e-01 -5.93900502e-01 -4.43733752e-01 2.12624505e-01 -5.79122424e-01 1.66785926e-01 7.77564287e-01 -2.95770168e-01 -1.16951905e-01 1.09694421e-01 6.62459573e-03 -7.40905523e-01 3.53213429e-01 8.44979942e-01 -7.93387890e-02 -3.61327350e-01 8.27156544e-01 7.19676495e-01 -8.02724995e-03 -3.19713295e-01 1.03492759e-01 4.01080638e-01 7.35067189e-01 -6.96896672e-01 8.51952791e-01 7.85039186e-01 4.36122209e-01 -1.14715636e+00 -1.20182288e+00 -6.14985645e-01 -1.99639991e-01 1.75512627e-01 1.42525584e-01 -1.12672544e+00 -5.70864558e-01 -3.51303577e-01 -6.28945768e-01 1.06511623e-01 -8.28229249e-01 8.16465914e-01 -6.68908775e-01 4.50327218e-01 -3.51614505e-02 -9.53063071e-01 -9.85358581e-02 -9.15098965e-01 9.00767326e-01 -2.71600693e-01 -3.35486531e-01 -7.30091989e-01 3.73918004e-02 3.84581059e-01 3.74336988e-01 3.22438300e-01 1.07377768e+00 -4.91418004e-01 -3.40533733e-01 -5.20900249e-01 -2.11659506e-01 4.96574730e-01 3.05724628e-02 -4.74249065e-01 -6.32721901e-01 -6.50964677e-01 7.85682917e-01 -1.85179904e-01 7.96679318e-01 5.58007419e-01 7.82252789e-01 -6.94314539e-01 -1.21894933e-01 8.93906295e-01 1.67296350e+00 1.36158159e-02 5.84761739e-01 -7.93006679e-04 3.59539449e-01 4.56948727e-01 1.40152618e-01 5.94393194e-01 -4.73973155e-02 2.75892138e-01 2.31257126e-01 -3.46896015e-02 -1.91600788e-02 1.24415517e-01 3.04920077e-01 9.21607792e-01 1.03682674e-01 1.87294886e-01 -5.23006380e-01 6.71929538e-01 -1.36801004e+00 -1.43666399e+00 1.24499528e-03 2.38817501e+00 7.83122838e-01 -2.07739025e-01 -9.34291333e-02 8.50498021e-01 5.72659492e-01 2.66244501e-01 -4.87100363e-01 -1.33336872e-01 -5.07112741e-01 8.36198270e-01 4.16526079e-01 8.13972056e-01 -8.03918064e-01 3.43644500e-01 6.72425842e+00 9.06982243e-01 -1.22271967e+00 3.62951830e-02 4.02577937e-01 -1.73648894e-01 -6.10392869e-01 4.27257977e-02 -4.52588946e-01 4.24652398e-01 6.06957555e-01 -3.32290977e-01 5.59476674e-01 8.11749578e-01 -1.09207727e-01 -1.03997640e-01 -1.36508715e+00 1.30925941e+00 5.15390694e-01 -1.56814992e+00 -2.21326381e-01 3.14970821e-01 9.97307718e-01 -1.54707879e-01 1.69155225e-01 3.25059285e-04 2.28003725e-01 -1.34907269e+00 7.15864420e-01 2.13775635e-01 1.12455392e+00 -6.02489531e-01 4.55223709e-01 4.74244326e-01 -1.32686663e+00 -1.87791586e-01 -3.73199016e-01 -1.32365614e-01 -1.17862083e-01 1.32451379e+00 -3.98361146e-01 6.18271410e-01 3.99356365e-01 8.62566173e-01 -8.21739202e-04 7.67494738e-01 1.60702348e-01 5.44890106e-01 -5.39258778e-01 5.74278235e-01 -5.02501391e-02 -4.04819191e-01 6.78232312e-01 1.15491319e+00 7.40794659e-01 5.94020903e-01 4.42852825e-01 6.68850958e-01 -9.21658799e-03 7.22816633e-03 -7.67168641e-01 -7.93240145e-02 3.93664658e-01 6.15544438e-01 -5.40843248e-01 -1.93107158e-01 -2.36171409e-01 6.46620452e-01 -1.19184986e-01 5.28517365e-01 -2.47616321e-01 -2.06036687e-01 8.20201516e-01 8.08204636e-02 3.44190478e-01 -3.16765428e-01 -5.67599177e-01 -1.39754784e+00 1.28465578e-01 -1.42175758e+00 5.27196527e-01 -3.48347008e-01 -1.31321800e+00 5.22265732e-01 -2.13749871e-01 -1.48126948e+00 -3.26163709e-01 -2.96772361e-01 -2.54943430e-01 8.96255255e-01 -1.40767884e+00 -5.94180286e-01 -2.48827673e-02 9.98148561e-01 1.85635522e-01 -2.29958981e-01 1.04496396e+00 1.91945851e-01 -3.51100087e-01 5.98440588e-01 3.81247789e-01 1.61766633e-01 3.68677005e-02 -6.07239366e-01 -7.28449106e-01 1.06967890e+00 4.80117500e-01 7.36582518e-01 8.17760050e-01 -2.90699542e-01 -1.82868326e+00 -6.82094395e-01 9.05752838e-01 2.41885055e-02 3.45413536e-01 1.40750200e-01 -6.16241693e-01 7.30111778e-01 -2.26133451e-01 3.30964088e-01 1.06490910e+00 6.87920675e-02 -8.03901851e-01 -2.34046355e-01 -1.36020839e+00 1.77004978e-01 1.02662551e+00 -7.87615657e-01 -5.11727750e-01 4.46460694e-01 2.21971750e-01 -5.00762388e-02 -9.17493701e-01 5.43838263e-01 4.72248226e-01 -1.06300402e+00 1.12303174e+00 -6.57833278e-01 4.50961649e-01 -2.69043505e-01 -1.23411000e+00 -8.13976824e-01 -3.22634518e-01 -5.88837206e-01 -1.97136194e-01 3.57193559e-01 2.78861046e-01 -5.87491095e-01 9.73437607e-01 2.05446199e-01 -1.38014376e-01 -6.36064172e-01 -1.42959499e+00 -9.99519289e-01 5.02856411e-02 -4.32728112e-01 2.58754730e-01 1.40736198e+00 2.89448470e-01 7.41281137e-02 -3.95619005e-01 3.29089344e-01 8.89121950e-01 5.98422706e-01 4.08979505e-01 -1.27335060e+00 -6.50955915e-01 -4.50947225e-01 -4.80878115e-01 -9.19148505e-01 2.10123971e-01 -1.07599127e+00 -9.72615927e-02 -1.18062639e+00 1.57667756e-01 -6.21570945e-01 -1.35681972e-01 3.71827751e-01 4.25740749e-01 3.41019452e-01 1.91736326e-01 6.27377033e-01 -2.22446144e-01 3.83252501e-01 9.73732293e-01 -4.33751613e-01 -2.93596834e-02 -1.38227910e-01 -1.07113457e+00 5.25081396e-01 4.37640518e-01 -4.47130114e-01 -3.46297264e-01 -3.25944573e-01 4.95814353e-01 4.15195167e-01 2.23992705e-01 -1.25386429e+00 2.39446357e-01 -1.18291855e-01 1.82519600e-01 -1.23930849e-01 5.78141689e-01 -8.56510282e-01 4.38642412e-01 4.71871585e-01 -4.04681861e-01 -1.72841057e-01 -1.79846779e-01 4.64525402e-01 -4.18675482e-01 -3.94470304e-01 7.81197488e-01 -1.94948480e-01 -2.20610887e-01 1.35202065e-01 -4.78583723e-02 3.21858168e-01 8.65119100e-01 -4.95992035e-01 -7.36821741e-02 -6.52234197e-01 -7.67530560e-01 -2.59467781e-01 3.22451651e-01 -5.02189159e-01 6.41728461e-01 -1.26687503e+00 -9.66788292e-01 8.00004542e-01 -1.01808123e-01 -1.63803324e-01 1.42166093e-01 1.01976693e+00 -1.62457585e-01 3.81913662e-01 1.34967953e-01 -7.73637176e-01 -8.80602002e-01 4.32402760e-01 3.04733425e-01 -6.44583479e-02 -4.85397249e-01 1.11819935e+00 -9.58941206e-02 1.51033849e-01 1.58781588e-01 -8.30897875e-03 3.81853640e-01 -2.81874603e-03 6.08346939e-01 3.03951263e-01 2.20155254e-01 -8.20609450e-01 -3.74682963e-01 7.84602523e-01 2.23630145e-01 -3.78577858e-01 1.42718112e+00 8.15728679e-02 -3.96791399e-01 5.76706193e-02 1.64437854e+00 4.18611944e-01 -7.28601754e-01 -3.31095964e-01 -4.24491882e-01 -6.52450383e-01 1.31987007e-02 -2.49683708e-01 -1.25986660e+00 7.80365407e-01 2.16341957e-01 3.93528521e-01 1.38100183e+00 1.92682669e-02 6.67305529e-01 4.98593688e-01 6.17310345e-01 -7.32348859e-01 -1.93387643e-01 3.55213076e-01 8.37018967e-01 -9.83160257e-01 2.33473808e-01 -5.79327166e-01 -5.64131029e-02 8.68474722e-01 -2.39649236e-01 -4.23815876e-01 8.28276813e-01 3.77087116e-01 -3.17318052e-01 -2.64815807e-01 -4.16446775e-01 -1.07650571e-02 2.54031807e-01 5.79860210e-01 3.30257386e-01 2.31254823e-03 -1.60336748e-01 4.90524352e-01 -6.32498741e-01 8.35720077e-03 2.54299402e-01 8.95094931e-01 -5.52837670e-01 -8.34535003e-01 -7.60647774e-01 6.11194551e-01 -2.89246678e-01 -1.81377575e-01 4.70846444e-02 5.57410061e-01 2.31446877e-01 1.16983581e+00 -1.52412683e-01 -4.05234963e-01 2.91649122e-02 2.04877853e-02 7.00785339e-01 -5.66819370e-01 -9.13621411e-02 1.52719453e-01 -1.52278706e-01 -6.16609871e-01 -6.21742845e-01 -7.43014514e-01 -1.16029477e+00 -6.60568655e-01 -9.25969705e-02 7.05979288e-01 3.42990637e-01 1.10605419e+00 2.81841189e-01 -1.00863807e-01 9.75058377e-01 -4.74860758e-01 -8.93637836e-01 -5.81998110e-01 -1.00678229e+00 3.16303164e-01 6.50119662e-01 -3.48927677e-01 -9.12350297e-01 3.20447624e-01]
[7.119048118591309, 4.47360372543335]
0c2bf049-9232-4f22-b79f-d076e788819a
structure-aware-slam-using-quadrics-and
1804.09111
null
http://arxiv.org/abs/1804.09111v3
http://arxiv.org/pdf/1804.09111v3.pdf
Structure Aware SLAM using Quadrics and Planes
Simultaneous Localization And Mapping (SLAM) is a fundamental problem in mobile robotics. While point-based SLAM methods provide accurate camera localization, the generated maps lack semantic information. On the other hand, state of the art object detection methods provide rich information about entities present in the scene from a single image. This work marries the two and proposes a method for representing generic objects as quadrics which allows object detections to be seamlessly integrated in a SLAM framework. For scene coverage, additional dominant planar structures are modeled as infinite planes. Experiments show that the proposed points-planes-quadrics representation can easily incorporate Manhattan and object affordance constraints, greatly improving camera localization and leading to semantically meaningful maps. The performance of our SLAM system is demonstrated in https://youtu.be/dR-rB9keF8M .
['Mehdi Hosseinzadeh', 'Niko Suenderhauf', 'Trung Pham', 'Yasir Latif', 'Ian Reid']
2018-04-24
null
null
null
null
['camera-localization']
['computer-vision']
[-9.86325666e-02 -1.50424480e-01 -2.38338470e-01 -5.09590209e-01 -5.64280570e-01 -7.43848383e-01 6.07529819e-01 1.34050772e-01 -2.84498364e-01 5.80511332e-01 -1.93081617e-01 -1.68544933e-01 -1.43177003e-01 -8.41838300e-01 -8.54271293e-01 -1.87133834e-01 1.97968423e-01 8.26584995e-01 6.71446502e-01 -1.80000022e-01 4.30233926e-01 6.92832708e-01 -1.44760859e+00 -2.91329682e-01 9.32333291e-01 6.68977976e-01 7.95141160e-01 5.13340890e-01 -1.96447641e-01 4.31242347e-01 -1.75335944e-01 -2.51970887e-01 3.52362752e-01 1.58509403e-01 -4.66856390e-01 1.07014447e-01 8.58557343e-01 -3.50240052e-01 -3.71689528e-01 1.32140958e+00 2.68126279e-02 6.98109642e-02 3.43612254e-01 -1.55385554e+00 -3.72415543e-01 -4.03288864e-02 -6.38809800e-01 -2.80519933e-01 7.64273703e-01 -2.63548732e-01 9.01260376e-01 -1.30492079e+00 7.61370540e-01 1.29958820e+00 7.31413066e-01 8.84191245e-02 -9.92755949e-01 -5.33825040e-01 7.75730237e-02 9.79854912e-02 -1.83007777e+00 -3.75232249e-01 8.22163343e-01 -3.71243715e-01 8.05743277e-01 2.38301963e-01 6.84896231e-01 4.84078974e-01 2.48188779e-01 6.54975414e-01 6.21442914e-01 -2.89835632e-01 1.49749249e-01 2.49849409e-01 -2.92932987e-03 9.38671350e-01 9.47485268e-01 -2.11721241e-01 -6.72270715e-01 -8.51230323e-02 1.17567527e+00 5.39704323e-01 -2.08209723e-01 -1.20213413e+00 -1.58401740e+00 6.97858274e-01 9.82393861e-01 -1.18814588e-01 -2.81525731e-01 4.01471078e-01 -2.00259164e-01 -1.91270739e-01 2.80521978e-02 4.44067925e-01 1.69484898e-01 9.99580994e-02 -6.26046360e-01 3.86398584e-01 4.37946856e-01 1.87461340e+00 1.37261927e+00 -1.67712972e-01 5.95031679e-01 4.71577704e-01 5.03113329e-01 9.47165370e-01 1.40970945e-03 -1.08326352e+00 4.41843420e-01 8.26911926e-01 5.73539019e-01 -1.39577496e+00 -4.22536850e-01 -2.52917767e-01 -2.08606362e-01 1.47431970e-01 1.85641155e-01 3.17790240e-01 -8.08869481e-01 1.22209358e+00 4.94992554e-01 8.92316550e-02 -1.21083064e-02 1.18631208e+00 6.28379226e-01 3.28855932e-01 -2.86825359e-01 5.00972152e-01 1.36982191e+00 -1.06390929e+00 -5.91906369e-01 -6.09786272e-01 5.47178328e-01 -7.27905452e-01 6.81852043e-01 5.40443845e-02 -6.45593882e-01 -3.56338799e-01 -1.09695911e+00 -5.11257112e-01 -2.97835618e-01 1.82106689e-01 6.55903220e-01 2.44079739e-01 -1.04696882e+00 5.47553338e-02 -7.99337685e-01 -6.51697397e-01 1.50670752e-01 3.51465493e-01 -6.93787873e-01 -3.60783547e-01 -6.64244354e-01 1.11872387e+00 4.50478673e-01 -7.03821853e-02 -8.03603709e-01 -3.16663086e-01 -1.23946953e+00 -3.38988453e-01 4.73394871e-01 -7.86643803e-01 1.12239039e+00 -4.39606071e-01 -9.35333669e-01 7.15147316e-01 -6.44782960e-01 -3.75256211e-01 5.82746565e-01 -4.91730660e-01 -5.42057231e-02 2.80248433e-01 6.39209569e-01 8.80627811e-01 4.22027707e-01 -1.49837613e+00 -1.03049457e+00 -5.25728822e-01 3.06487650e-01 6.72622442e-01 2.86507756e-01 -2.71791756e-01 -6.96293354e-01 2.89362110e-03 9.87908661e-01 -1.08867419e+00 -3.43931437e-01 3.54215741e-01 -3.57219577e-01 1.00113980e-01 9.84800577e-01 -2.56531239e-01 5.81672251e-01 -2.03413010e+00 2.07538139e-02 1.42406151e-01 2.16819003e-01 -4.58272845e-01 -1.90710351e-02 4.44688618e-01 5.55646956e-01 -1.01810575e-01 2.95957308e-02 -6.01758957e-01 2.84538995e-02 3.20584267e-01 -2.51088440e-01 8.60957682e-01 1.99289881e-02 9.55084503e-01 -1.15589273e+00 -4.08770442e-01 6.17184162e-01 5.35159707e-01 -2.93262899e-01 -1.07726820e-01 2.84621902e-02 3.21581215e-01 -4.23384190e-01 9.38793600e-01 1.01309395e+00 4.89146337e-02 -4.57460545e-02 -3.15447152e-02 -3.96670669e-01 3.10074389e-01 -1.41176057e+00 2.04476261e+00 -1.79696620e-01 7.13771999e-01 6.70834631e-02 -4.18796808e-01 1.13369358e+00 -6.80520460e-02 3.09352785e-01 -4.67811406e-01 -9.06244665e-02 5.38307786e-01 -4.06632483e-01 1.28999084e-01 1.15338564e+00 1.74489737e-01 -2.45766625e-01 -2.04978302e-01 -2.04288885e-02 -5.14951646e-01 -4.92508896e-02 3.03069830e-01 8.02193463e-01 3.80491078e-01 6.32374227e-01 -3.65579486e-01 1.78443521e-01 7.09348798e-01 6.49180233e-01 6.75367236e-01 -1.55403793e-01 5.76150894e-01 -1.44668639e-01 -4.45232838e-01 -1.15835512e+00 -1.30905688e+00 -2.46050164e-01 4.33012843e-01 1.29088855e+00 -5.73878467e-01 -3.28702539e-01 -2.42211103e-01 3.94811243e-01 4.67642456e-01 -1.93545803e-01 2.20010757e-01 -3.78202438e-01 -1.24900118e-01 1.46313950e-01 4.43633288e-01 3.97706777e-01 -5.09037793e-01 -1.04834795e+00 1.85564443e-01 -2.88426340e-01 -1.31525433e+00 -2.38954768e-01 5.01196040e-03 -7.81729043e-01 -1.15517056e+00 -5.41822791e-01 -8.24688673e-01 9.64020729e-01 1.28653598e+00 7.72639632e-01 -9.86000523e-02 -3.17564696e-01 4.76100951e-01 -2.68448532e-01 -4.61255193e-01 -3.76199782e-02 -2.11561874e-01 5.48830092e-01 -1.21355258e-01 4.03888166e-01 -3.48349512e-01 -6.46922290e-01 4.95493650e-01 -2.44655326e-01 3.95671397e-01 4.65722948e-01 3.08647275e-01 9.52286541e-01 -2.62769997e-01 1.15634866e-01 -5.13042092e-01 -9.58595276e-02 -2.88731545e-01 -9.85010386e-01 3.82178575e-02 -5.22189021e-01 -1.96537778e-01 9.39499959e-02 -1.65979564e-01 -7.95699537e-01 5.95421433e-01 3.89028549e-01 -4.89841998e-01 -1.99429244e-01 3.52661014e-01 -1.95784166e-01 -5.93608141e-01 5.72448432e-01 3.08380038e-01 -1.36907160e-01 -4.38322604e-01 5.29797733e-01 4.88237768e-01 7.21977592e-01 -3.89355689e-01 9.20038998e-01 9.73014295e-01 1.52869120e-01 -8.15881073e-01 -4.31234479e-01 -1.11481500e+00 -1.03021848e+00 -3.01102757e-01 5.75704277e-01 -1.30525458e+00 -3.02467823e-01 4.24558893e-02 -1.22014248e+00 1.25418082e-01 -8.60336795e-02 7.11878896e-01 -6.60588920e-01 5.21638274e-01 -1.08375579e-01 -8.82856309e-01 1.10421799e-01 -1.21354771e+00 1.35981548e+00 1.97012454e-01 -5.03301062e-02 -6.77074075e-01 -4.47303541e-02 2.29046538e-01 -8.16057026e-02 3.20728123e-01 7.15452805e-02 -2.58974969e-01 -1.29888844e+00 -3.67667347e-01 -3.94934535e-01 -4.97120708e-01 1.93294391e-01 -2.60156244e-01 -6.94719732e-01 -2.35541359e-01 -1.67857602e-01 2.90647030e-01 6.51665628e-01 1.83350027e-01 1.56485066e-01 -4.60734032e-02 -7.81285286e-01 7.58106768e-01 1.76091945e+00 1.30400375e-01 4.85873729e-01 6.99108422e-01 9.46407735e-01 4.95178670e-01 9.50548768e-01 2.22417161e-01 8.29796374e-01 8.56118321e-01 8.88266563e-01 4.75673750e-03 -2.52106469e-02 -6.04650021e-01 1.66053757e-01 3.39338124e-01 1.52431667e-01 9.72832441e-02 -1.24314237e+00 6.79498255e-01 -2.17735362e+00 -5.50645351e-01 -6.53259516e-01 2.11568165e+00 1.64792195e-01 -1.59500062e-01 -2.22778946e-01 -2.65413374e-01 7.69763231e-01 -1.53809845e-01 -4.37143415e-01 2.62371540e-01 -1.67937040e-01 -6.29771829e-01 1.04591131e+00 8.84500206e-01 -1.07761073e+00 1.25282931e+00 5.57396841e+00 3.40538591e-01 -8.34110856e-01 9.59959030e-02 -6.06098592e-01 1.34958401e-01 -3.47506195e-01 5.23490727e-01 -1.13596761e+00 2.62815058e-02 3.71055871e-01 -1.21605627e-01 2.00154886e-01 1.27121913e+00 1.90547928e-02 -4.86891985e-01 -9.39950287e-01 1.30061364e+00 1.63109645e-01 -1.35935104e+00 1.44500837e-01 2.89365798e-01 6.73185825e-01 4.03641820e-01 -1.50249660e-01 -7.43068904e-02 2.77376175e-01 -6.21303737e-01 1.29038095e+00 3.50625008e-01 7.52330363e-01 -5.45256317e-01 5.30245841e-01 6.56416416e-01 -1.49751747e+00 -1.77143887e-02 -7.16552377e-01 -1.95003688e-01 3.70109200e-01 3.93924713e-01 -1.37114382e+00 6.84370518e-01 6.64359450e-01 8.24992180e-01 -6.79098725e-01 1.43028569e+00 -2.03627959e-01 -3.47979844e-01 -5.94849169e-01 -4.34260517e-02 2.27116436e-01 -3.87147337e-01 9.89668012e-01 1.03185391e+00 5.81269979e-01 -3.04709315e-01 4.29984689e-01 9.97222543e-01 9.75659192e-02 4.09335718e-02 -9.51004922e-01 4.84761536e-01 8.17181945e-01 1.22552192e+00 -9.92416024e-01 -2.48459801e-01 -4.81571645e-01 1.00184453e+00 3.12575310e-01 2.63494879e-01 -8.09890568e-01 -2.88999379e-01 7.05414593e-01 2.72618949e-01 9.06743407e-02 -8.65289390e-01 -2.95640647e-01 -1.42018020e+00 1.91165790e-01 -1.65277779e-01 -2.05947563e-01 -1.15809631e+00 -6.82377815e-01 2.88915932e-01 -3.70913325e-03 -1.49580240e+00 -7.75243491e-02 -6.19203866e-01 -2.67039817e-02 8.24412525e-01 -1.62632787e+00 -1.44062793e+00 -7.40669072e-01 5.07420480e-01 5.88351846e-01 1.52825028e-01 7.20275640e-01 7.23865181e-02 1.39749348e-01 -6.56272471e-02 2.59469241e-01 -1.03647327e-02 5.97954571e-01 -1.30239713e+00 5.59649825e-01 1.05739224e+00 5.99986255e-01 8.51202667e-01 6.74405158e-01 -1.00825202e+00 -1.73304713e+00 -1.19693935e+00 6.69881761e-01 -7.81323016e-01 3.88105512e-01 -7.01475024e-01 -6.37752533e-01 8.87839198e-01 -4.25335020e-01 7.42033450e-03 -1.94792636e-02 9.89214238e-03 -3.98901522e-01 6.63288832e-02 -9.92188931e-01 7.00108469e-01 1.10699379e+00 -6.39041901e-01 -4.28882957e-01 2.62545794e-01 6.79138482e-01 -8.30052674e-01 -4.68295932e-01 4.37587470e-01 5.80844939e-01 -8.36292684e-01 1.15014648e+00 1.21893197e-01 -4.26154017e-01 -1.01901639e+00 -5.58985531e-01 -9.88287091e-01 -2.34446928e-01 -3.04489166e-01 -5.95134078e-03 7.74990499e-01 1.17685817e-01 -6.31967843e-01 7.65623868e-01 4.43694890e-01 -1.77734837e-01 -1.92600831e-01 -9.92191553e-01 -9.26172554e-01 -5.50750494e-01 -5.04854798e-01 5.98749816e-01 7.15115249e-01 -1.07662357e-01 2.07722560e-01 -2.75430053e-01 7.87485600e-01 8.31094980e-01 2.23466426e-01 1.19054949e+00 -1.42346454e+00 3.57468605e-01 -2.61544287e-01 -9.87962127e-01 -1.38523340e+00 -1.14094459e-01 -7.73342133e-01 2.98802406e-01 -1.97190166e+00 1.65020317e-01 -6.79333329e-01 1.51609659e-01 2.35203758e-01 1.90194681e-01 4.18848395e-01 2.65833557e-01 7.27658153e-01 -9.35109437e-01 3.42467159e-01 8.19011927e-01 6.62739128e-02 -5.56587009e-03 -2.60120958e-01 -3.89531881e-01 8.95601332e-01 7.85607100e-01 -4.13223892e-01 -2.92316079e-01 -7.18284965e-01 2.19976529e-01 -2.13746354e-02 8.17014456e-01 -1.16149604e+00 6.44507349e-01 -1.80331752e-01 1.84219509e-01 -1.02228940e+00 8.68424177e-01 -1.05237293e+00 4.88211870e-01 3.93083185e-01 2.73748696e-01 8.32654238e-02 3.49860713e-02 9.23554718e-01 -1.75891832e-01 -2.69107550e-01 3.94036233e-01 -2.93946594e-01 -1.34744763e+00 3.23882401e-01 1.57853998e-02 -5.01694500e-01 1.14274859e+00 -6.18380666e-01 -2.44861245e-01 -4.27210748e-01 -3.91658336e-01 3.70948136e-01 1.14806736e+00 5.91861546e-01 1.00962532e+00 -1.35769510e+00 -4.49244738e-01 1.71348736e-01 5.69657207e-01 5.63812792e-01 4.66232970e-02 7.57126629e-01 -1.09070778e+00 6.59083724e-01 -2.02093899e-01 -1.21132588e+00 -1.17042542e+00 4.84062910e-01 2.50966638e-01 5.41234076e-01 -7.74233282e-01 5.50910175e-01 4.49110866e-01 -6.91547513e-01 7.84894004e-02 -2.91399211e-01 2.15010464e-01 -3.31665754e-01 3.54551941e-01 3.49714935e-01 -2.04321101e-01 -1.17672646e+00 -7.58342803e-01 9.08098996e-01 2.80881345e-01 -2.96283007e-01 1.01316452e+00 -6.81828141e-01 -1.82479307e-01 5.15695810e-01 8.61284316e-01 3.26032937e-01 -1.28662753e+00 -3.17241967e-01 3.05229902e-01 -8.05694044e-01 -2.29616880e-01 -3.11238885e-01 -2.02077821e-01 7.96709657e-01 3.94144803e-01 -2.04238892e-01 4.02406216e-01 1.29063502e-01 3.68736118e-01 7.21784651e-01 1.38421476e+00 -6.81901157e-01 -1.43435583e-01 5.82390010e-01 8.06274593e-01 -1.47095263e+00 2.16039523e-01 -8.29675257e-01 -5.56685150e-01 1.17091525e+00 5.49982607e-01 -2.64013410e-01 1.73077345e-01 1.43527567e-01 2.96709146e-02 -2.18693912e-01 -2.93139387e-02 -2.84524083e-01 2.79331952e-01 8.33619297e-01 -1.20213076e-01 2.30181277e-01 1.57289609e-01 1.66710168e-01 -2.77877778e-01 -2.96890557e-01 6.48595572e-01 1.07806349e+00 -8.97707582e-01 -7.69127429e-01 -6.71672881e-01 -1.00152172e-01 1.73194200e-01 8.95895239e-04 -2.16887370e-01 9.62438524e-01 6.20586276e-02 7.81136811e-01 1.86413169e-01 -2.77637243e-01 2.77763844e-01 -1.93715587e-01 4.45167720e-01 -7.66769946e-01 3.77072930e-01 1.67481616e-01 -8.76046345e-02 -7.49249458e-01 -4.15595293e-01 -6.71407580e-01 -1.55734003e+00 1.29562384e-02 -5.57134748e-01 1.21340998e-01 1.09858716e+00 4.92315263e-01 5.74218750e-01 5.36453016e-02 1.39191866e-01 -1.08875930e+00 -1.42159566e-01 -6.21479750e-01 -7.08860636e-01 1.30864590e-01 5.03683865e-01 -1.07660902e+00 -2.31511015e-02 -1.81928724e-01]
[7.327858924865723, -2.2907612323760986]
ac5409c6-8f95-4073-968f-17cb5af5e15c
singgan-generative-adversarial-network-for
2110.07468
null
https://arxiv.org/abs/2110.07468v4
https://arxiv.org/pdf/2110.07468v4.pdf
SingGAN: Generative Adversarial Network For High-Fidelity Singing Voice Generation
Deep generative models have achieved significant progress in speech synthesis to date, while high-fidelity singing voice synthesis is still an open problem for its long continuous pronunciation, rich high-frequency parts, and strong expressiveness. Existing neural vocoders designed for text-to-speech cannot directly be applied to singing voice synthesis because they result in glitches and poor high-frequency reconstruction. In this work, we propose SingGAN, a generative adversarial network designed for high-fidelity singing voice synthesis. Specifically, 1) to alleviate the glitch problem in the generated samples, we propose source excitation with the adaptive feature learning filters to expand the receptive field patterns and stabilize long continuous signal generation; and 2) SingGAN introduces global and local discriminators at different scales to enrich low-frequency details and promote high-frequency reconstruction; and 3) To improve the training efficiency, SingGAN includes auxiliary spectrogram losses and sub-band feature matching penalty loss. To the best of our knowledge, SingGAN is the first work designed toward high-fidelity singing voice vocoding. Our evaluation of SingGAN demonstrates the state-of-the-art results with higher-quality (MOS 4.05) samples. Also, SingGAN enables a sample speed of 50x faster than real-time on a single NVIDIA 2080Ti GPU. We further show that SingGAN generalizes well to the mel-spectrogram inversion of unseen singers, and the end-to-end singing voice synthesis system SingGAN-SVS enjoys a two-stage pipeline to transform the music scores into expressive singing voices. Audio samples are available at \url{https://SingGAN.github.io/}
['Zhefeng Wang', 'Baoxing Huai', 'Zhou Zhao', 'Jinglin Liu', 'Yi Ren', 'Chenye Cui', 'Rongjie Huang', 'Feiyang Chen']
2021-10-14
null
null
null
null
['singing-voice-synthesis']
['speech']
[-2.18553673e-02 -1.15353204e-01 -3.84357851e-03 1.34058028e-01 -1.19355786e+00 -5.44097364e-01 1.75693169e-01 -7.73280561e-01 2.88023919e-01 5.61108351e-01 5.77999592e-01 -9.12074745e-02 3.29655379e-01 -5.94853163e-01 -7.04305112e-01 -6.44183695e-01 1.08266048e-01 1.53454781e-01 -1.87145531e-01 -3.31093132e-01 -4.40756202e-01 1.59036025e-01 -1.76032400e+00 4.06579196e-01 7.92560220e-01 7.06303239e-01 2.52688408e-01 1.28211582e+00 2.99254149e-01 6.18411958e-01 -8.98368776e-01 8.62124655e-03 3.35893720e-01 -1.08147192e+00 -4.03235704e-01 -2.49185264e-01 7.26384938e-01 -6.17959797e-01 -6.49424553e-01 8.86407852e-01 1.19428229e+00 3.62806439e-01 3.17832768e-01 -9.70786035e-01 -6.71634793e-01 8.01376879e-01 5.00668623e-02 -8.63740072e-02 1.37245551e-01 7.42451847e-01 1.22032404e+00 -1.17569196e+00 3.28610331e-01 1.27390635e+00 6.94954455e-01 9.48008358e-01 -1.07222688e+00 -1.00042641e+00 -5.92918456e-01 -1.41631171e-01 -1.20109785e+00 -9.99359071e-01 9.58007395e-01 -1.15813851e-01 1.07370102e+00 6.81466460e-01 7.69433737e-01 1.24703944e+00 -1.80354178e-01 6.71381235e-01 8.39411199e-01 -2.78946161e-01 -3.96928452e-02 -2.99140424e-01 -5.61026990e-01 4.46777523e-01 -5.11288702e-01 6.20823622e-01 -9.56446648e-01 -7.82966167e-02 1.00696921e+00 -4.51146990e-01 -4.69836414e-01 4.64862794e-01 -1.18918550e+00 7.00473249e-01 2.60352761e-01 3.28170687e-01 -2.80730814e-01 6.49588287e-01 5.15379012e-01 3.95307720e-01 3.24112833e-01 6.90241039e-01 -1.26829714e-01 -5.13753891e-01 -1.35385787e+00 5.47891557e-01 7.84960806e-01 6.54507577e-01 2.32915580e-01 1.21464562e+00 -4.94685918e-01 1.09209263e+00 2.89115845e-03 8.25763345e-01 8.12165678e-01 -1.20191216e+00 2.92062759e-01 -3.67876500e-01 -4.13713485e-01 -5.01710534e-01 5.09619080e-02 -8.30329299e-01 -9.31981504e-01 2.02332839e-01 1.34605184e-01 -3.57620895e-01 -7.57573724e-01 1.76430786e+00 2.36684933e-01 5.59297502e-01 6.59475550e-02 1.21622956e+00 1.20566726e+00 1.12933326e+00 -5.02530456e-01 -1.72712788e-01 1.12683058e+00 -1.38558626e+00 -1.04980028e+00 -2.78954506e-02 2.47138795e-02 -1.16586387e+00 1.55513692e+00 3.91301453e-01 -1.53494024e+00 -1.03808963e+00 -1.08305025e+00 -3.79499257e-01 2.79160589e-01 2.72520095e-01 4.27801162e-01 6.03365242e-01 -1.08634746e+00 9.36462402e-01 -7.41732359e-01 3.59193772e-01 1.90809682e-01 3.00671697e-01 4.61457111e-02 4.41389680e-01 -1.28875172e+00 1.48547590e-01 -2.28561019e-03 -1.75137118e-01 -1.31961167e+00 -1.10576296e+00 -6.18644655e-01 2.54875660e-01 8.82587358e-02 -8.20230901e-01 1.46475291e+00 -9.14108694e-01 -2.19573426e+00 3.43650132e-01 -7.14202002e-02 -5.11396468e-01 2.37761065e-01 -2.81827182e-01 -7.33025551e-01 1.18786626e-01 -1.28517523e-01 5.27213693e-01 1.38342178e+00 -8.98235142e-01 -2.96404839e-01 1.56971931e-01 -5.43225348e-01 3.08943659e-01 -5.35463870e-01 -8.77161324e-02 -2.42283389e-01 -1.37201762e+00 -2.34240100e-01 -7.95153379e-01 6.48474395e-02 -3.76691133e-01 -3.37366551e-01 1.01944521e-01 9.09654796e-01 -8.55928957e-01 1.50350034e+00 -2.32227492e+00 2.06518129e-01 -2.05045000e-01 2.35716119e-01 6.59695387e-01 -3.08708280e-01 3.18497509e-01 -6.02531321e-02 -3.74335013e-02 -1.89921677e-01 -4.67633069e-01 1.16589420e-01 -2.90430468e-02 -7.78189659e-01 1.92085072e-01 8.93256739e-02 9.47194874e-01 -8.60014141e-01 -1.55273139e-01 2.17894211e-01 8.23912203e-01 -1.12471282e+00 5.55015028e-01 -8.41429755e-02 5.87190509e-01 7.50334039e-02 6.37510061e-01 3.53308320e-01 2.75020987e-01 -5.10905981e-02 -1.75518140e-01 -6.87863529e-02 8.34382653e-01 -9.94535446e-01 1.83134878e+00 -7.23561049e-01 6.42693102e-01 4.64480311e-01 -2.80015111e-01 9.85442221e-01 6.90059483e-01 2.86853313e-01 -3.17967862e-01 5.70920072e-02 7.43286133e-01 2.07511336e-01 -2.10742608e-01 7.19988108e-01 -4.91865605e-01 1.29965767e-01 2.09708348e-01 4.85541523e-01 -9.20686543e-01 -2.00416952e-01 -2.53032774e-01 7.91525662e-01 6.01955540e-02 -2.49425285e-02 -4.96923225e-04 2.58498728e-01 -4.89122003e-01 5.35401881e-01 4.03561234e-01 8.52247998e-02 9.29771781e-01 9.75212827e-02 7.04354569e-02 -1.27255583e+00 -1.25032890e+00 -2.99447775e-02 1.11335003e+00 -5.59912145e-01 -6.54085279e-01 -1.12548923e+00 -8.29459205e-02 -1.48128599e-01 6.59591258e-01 -4.17995490e-02 -2.32991323e-01 -9.26754534e-01 -7.82540515e-02 1.25448501e+00 4.05656159e-01 2.39057928e-01 -1.41604435e+00 -2.43374050e-01 4.47415531e-01 -1.81553304e-01 -7.51155496e-01 -1.29317749e+00 3.58292013e-02 -7.72702038e-01 -4.06342596e-01 -9.39516842e-01 -9.25172389e-01 -6.47586361e-02 -7.29250237e-02 1.10461617e+00 -9.79734212e-02 -1.48563564e-01 -7.63932541e-02 -1.18349217e-01 -2.76046991e-01 -7.79530883e-01 1.34939015e-01 5.61887562e-01 -2.60111302e-01 -3.02772760e-01 -1.02793384e+00 -6.04963362e-01 1.49971336e-01 -7.58725762e-01 -4.68714461e-02 1.42060965e-01 1.38603938e+00 8.59496653e-01 7.19199516e-03 8.00244868e-01 -3.57676029e-01 8.27031374e-01 -1.80406183e-01 -3.75208586e-01 -4.83688712e-01 -3.11149091e-01 -3.08149546e-01 1.31229794e+00 -6.95904374e-01 -7.10116386e-01 -1.66675419e-01 -8.43284249e-01 -1.00163758e+00 1.54301032e-01 6.49678260e-02 -1.91318765e-01 1.83160171e-01 7.92587817e-01 2.93744087e-01 1.55220091e-01 -5.89363277e-01 4.33044076e-01 9.29828107e-01 1.04766428e+00 -5.57697654e-01 8.65722299e-01 5.78075312e-02 -1.51305869e-01 -1.18696284e+00 -4.75489587e-01 -1.50426850e-01 1.50921643e-01 -8.94470289e-02 5.98357439e-01 -1.20082700e+00 -7.88396060e-01 5.92442453e-01 -9.42047477e-01 -8.45396578e-01 -9.46378112e-01 6.75527990e-01 -8.28470051e-01 2.06795305e-01 -9.81407106e-01 -7.84274638e-01 -9.99124706e-01 -1.17800593e+00 1.06114721e+00 1.44105017e-01 -2.25112155e-01 -5.72305620e-01 1.93246141e-01 4.17949587e-01 6.49735868e-01 1.03983358e-01 4.02349979e-01 -3.70273739e-02 -2.49043137e-01 2.13877618e-01 4.51801032e-01 8.44099104e-01 1.64255917e-01 -2.91135255e-02 -1.42256689e+00 -5.03597200e-01 1.93916172e-01 -4.56912488e-01 8.15414071e-01 4.97210413e-01 1.18302238e+00 -6.37496889e-01 3.83958071e-01 1.18001783e+00 6.96310937e-01 1.33334517e-01 5.79621255e-01 -5.63922286e-01 9.47059870e-01 3.25586885e-01 4.62145895e-01 4.84775245e-01 -1.23512417e-01 8.45076323e-01 2.00081050e-01 -3.07493210e-01 -1.15803075e+00 -8.09005320e-01 7.67120600e-01 1.74636531e+00 -1.11366794e-01 -3.25051576e-01 -3.07479858e-01 5.50982654e-01 -1.29871249e+00 -1.13653326e+00 -6.44951314e-02 2.13789606e+00 1.11225975e+00 -2.71469206e-01 3.63429338e-01 4.00075346e-01 5.77437937e-01 5.78247428e-01 -6.63031220e-01 -5.36850333e-01 -1.90266475e-01 9.87462044e-01 1.25842497e-01 6.70957804e-01 -7.68902719e-01 1.13150561e+00 5.81994057e+00 1.45787191e+00 -1.46426618e+00 4.26072240e-01 2.95028567e-01 -6.84567690e-01 -5.27530611e-01 -2.09118068e-01 -6.84883237e-01 4.15300459e-01 1.08772647e+00 -2.83414014e-02 1.23242939e+00 7.63706744e-01 2.06163660e-01 8.76179457e-01 -7.45593309e-01 1.09596860e+00 4.84973937e-02 -1.43542683e+00 -1.10028170e-01 1.79753602e-02 8.05904925e-01 4.65549603e-02 5.87214112e-01 3.74080479e-01 -9.67902467e-02 -1.39108849e+00 1.00995994e+00 1.35736644e-01 1.66708302e+00 -9.64079261e-01 1.37946352e-01 1.27726927e-01 -1.41107118e+00 1.40898123e-01 -1.50768816e-01 -8.65695402e-02 2.72263408e-01 4.21899617e-01 -1.06112278e+00 2.48326793e-01 4.42627072e-01 5.16409039e-01 2.22083434e-01 4.77458864e-01 -3.81508529e-01 1.35231590e+00 -3.09767067e-01 6.92738593e-02 1.37762889e-01 1.06916517e-01 1.06839895e+00 1.11854398e+00 7.02105403e-01 1.47326468e-02 -2.29712531e-01 1.08674228e+00 -4.02581960e-01 8.11658800e-02 -3.86783242e-01 -4.19668376e-01 5.66372931e-01 9.90428746e-01 4.97090481e-02 -1.92273661e-01 8.38551447e-02 1.04322731e+00 -1.67539299e-01 2.73965210e-01 -8.95762861e-01 -6.22058272e-01 1.10921049e+00 3.41783762e-01 2.48011410e-01 -2.81491250e-01 -2.34718099e-01 -1.16081095e+00 -1.82741016e-01 -1.47560728e+00 -8.07701424e-02 -6.41842902e-01 -9.96698022e-01 8.21995795e-01 -7.15591311e-01 -1.30998588e+00 -6.12862885e-01 -2.58801550e-01 -7.66484022e-01 1.05477285e+00 -1.17543530e+00 -1.13774371e+00 -9.05616581e-02 6.52413070e-01 9.36908722e-01 -4.75715548e-01 1.10954750e+00 5.89136183e-01 -2.95802534e-01 8.69513690e-01 5.83447814e-02 -3.53689566e-02 8.36893260e-01 -1.16378868e+00 8.85252953e-01 8.77952397e-01 3.71097386e-01 4.21309650e-01 7.93975592e-01 -4.31386381e-01 -1.60761249e+00 -1.19304526e+00 7.15131760e-01 -1.08403660e-01 4.60025340e-01 -4.90396678e-01 -8.35273325e-01 2.15831578e-01 3.11749160e-01 2.17531070e-01 6.30504906e-01 -1.95767879e-01 -2.65948653e-01 -1.86039075e-01 -9.11905468e-01 7.29924142e-01 1.08068228e+00 -8.79916966e-01 -3.06891322e-01 1.36565953e-01 1.00116169e+00 -8.30673635e-01 -9.41274464e-01 2.94549584e-01 5.95613420e-01 -9.98981118e-01 1.11369717e+00 -2.71451652e-01 6.44729018e-01 -4.62237865e-01 -2.70400017e-01 -1.50005031e+00 -2.32834339e-01 -1.46371567e+00 -4.30120558e-01 1.43949330e+00 2.15003148e-01 -3.84737700e-01 4.74192977e-01 -4.90211487e-01 -7.47815549e-01 -6.31924093e-01 -9.30237055e-01 -8.44061911e-01 2.54407078e-01 -4.57534909e-01 9.57358778e-01 7.82666564e-01 -1.33851198e-02 4.90321904e-01 -1.05960798e+00 -2.34112963e-02 4.23311830e-01 2.01205924e-01 9.68547106e-01 -4.67206687e-01 -1.04058170e+00 -4.08726960e-01 1.05087712e-01 -1.18476498e+00 1.73247755e-01 -1.01708126e+00 2.56017745e-01 -9.27908480e-01 -3.89456958e-01 -3.81592095e-01 -2.07202192e-02 3.81615371e-01 -1.19922452e-01 6.17110312e-01 4.96492893e-01 7.68708065e-02 1.61517963e-01 8.77477288e-01 1.76111603e+00 3.84151600e-02 -5.22361577e-01 1.65696628e-02 -3.51307750e-01 6.39113069e-01 8.42739284e-01 -3.47534627e-01 -3.29888642e-01 -2.81592071e-01 -3.13709527e-01 5.87265015e-01 3.61200005e-01 -1.29742169e+00 -1.99740101e-02 2.03316167e-01 2.72383150e-02 -4.25854206e-01 8.44387293e-01 -2.49899194e-01 3.90153795e-01 5.07046282e-01 -2.73007274e-01 -4.23217773e-01 2.35285386e-01 2.85050683e-02 -6.11651480e-01 1.68899000e-01 9.97926772e-01 1.25642717e-01 -1.23270631e-01 3.21380496e-01 -2.53995121e-01 3.35193127e-01 3.15956861e-01 1.37601227e-01 -1.03071861e-01 -7.15367675e-01 -4.84779716e-01 -3.26449335e-01 2.72091269e-01 4.34230775e-01 6.53193474e-01 -1.62645924e+00 -1.00617862e+00 5.39689302e-01 -4.18824583e-01 -4.96357344e-02 6.55358851e-01 5.51434219e-01 -4.77425933e-01 1.65064752e-01 -1.50548220e-01 -3.41183394e-01 -1.31213677e+00 2.60465592e-01 4.07489777e-01 -2.82866526e-02 -7.13409483e-01 1.04087818e+00 2.29853168e-01 -4.41450983e-01 1.46848291e-01 -2.79741108e-01 2.13788688e-01 -1.51211351e-01 4.46092516e-01 6.41997099e-01 -5.68693550e-03 -7.00736523e-01 -2.29458567e-02 4.56518531e-01 6.54706419e-01 -3.04974407e-01 1.15789902e+00 2.47435570e-01 1.35201916e-01 3.07933658e-01 1.25070834e+00 7.46640921e-01 -1.30364490e+00 1.08005755e-01 -9.83872771e-01 -4.19792771e-01 3.06978583e-01 -5.48126459e-01 -1.27650976e+00 1.07765198e+00 3.63340646e-01 5.34529751e-03 1.38691485e+00 -2.46471867e-01 1.60727012e+00 -1.37809560e-01 -5.42783961e-02 -9.20850575e-01 2.49497548e-01 7.09864259e-01 1.23191535e+00 -6.39920175e-01 -3.04444224e-01 -2.22095057e-01 -7.07136095e-01 1.05562627e+00 3.39973480e-01 -3.72171223e-01 3.00506741e-01 6.89750552e-01 2.47814402e-01 2.25266635e-01 -7.97271967e-01 -1.40714839e-01 4.59809810e-01 4.08199310e-01 6.73762083e-01 4.09111410e-01 -1.69116527e-01 8.71498287e-01 -1.04122627e+00 -1.58616111e-01 1.55494645e-01 -2.43782997e-02 -1.98735967e-01 -1.09701443e+00 -5.65838993e-01 2.09599301e-01 -8.54319036e-01 -5.21792471e-01 -2.27603301e-01 2.40644991e-01 2.09035099e-01 1.11195767e+00 9.13469344e-02 -7.56946564e-01 3.65020782e-01 -1.00852728e-01 3.92638803e-01 -5.23978412e-01 -1.21200800e+00 7.25963414e-01 2.51799107e-01 -4.44693655e-01 2.62345791e-01 -3.77427280e-01 -1.34381592e+00 -4.87656713e-01 -3.12337786e-01 1.62258372e-01 6.09295428e-01 2.07990512e-01 4.73371416e-01 1.14919245e+00 9.76731718e-01 -1.07504427e+00 -8.59190524e-01 -1.04551852e+00 -7.18132377e-01 1.70995131e-01 6.77186430e-01 -5.94321936e-02 -5.94358146e-01 -2.32031457e-02]
[15.49459171295166, 6.151684761047363]
fd67a336-dccb-4f76-8412-b9201fdd70d3
spatiotemporal-self-attention-modeling-with
2207.13259
null
https://arxiv.org/abs/2207.13259v1
https://arxiv.org/pdf/2207.13259v1.pdf
Spatiotemporal Self-attention Modeling with Temporal Patch Shift for Action Recognition
Transformer-based methods have recently achieved great advancement on 2D image-based vision tasks. For 3D video-based tasks such as action recognition, however, directly applying spatiotemporal transformers on video data will bring heavy computation and memory burdens due to the largely increased number of patches and the quadratic complexity of self-attention computation. How to efficiently and effectively model the 3D self-attention of video data has been a great challenge for transformers. In this paper, we propose a Temporal Patch Shift (TPS) method for efficient 3D self-attention modeling in transformers for video-based action recognition. TPS shifts part of patches with a specific mosaic pattern in the temporal dimension, thus converting a vanilla spatial self-attention operation to a spatiotemporal one with little additional cost. As a result, we can compute 3D self-attention using nearly the same computation and memory cost as 2D self-attention. TPS is a plug-and-play module and can be inserted into existing 2D transformer models to enhance spatiotemporal feature learning. The proposed method achieves competitive performance with state-of-the-arts on Something-something V1 & V2, Diving-48, and Kinetics400 while being much more efficient on computation and memory cost. The source code of TPS can be found at https://github.com/MartinXM/TPS.
['Lei Zhang', 'Xian-Sheng Hua', 'Xihan Wei', 'Biao Wang', 'Chao Li', 'Wangmeng Xiang']
2022-07-27
null
null
null
null
['action-classification']
['computer-vision']
[ 3.04622501e-02 -4.38752800e-01 -6.59805909e-02 1.81016158e-02 -6.33776069e-01 -1.82073653e-01 4.48519468e-01 -2.37301022e-01 -3.25986356e-01 1.14303686e-01 1.34289384e-01 -1.51834488e-01 1.66263565e-01 -5.48312485e-01 -8.05297375e-01 -7.87773132e-01 1.08473919e-01 1.46960586e-01 6.36862218e-01 -1.51281565e-01 3.20715368e-01 4.55589622e-01 -1.62800574e+00 4.22309011e-01 5.51704586e-01 1.31639862e+00 4.77817088e-01 6.25011265e-01 -1.20724343e-01 9.54423308e-01 -2.86912709e-01 -1.23816825e-01 5.06650269e-01 -3.74842525e-01 -6.59791231e-01 1.81505665e-01 6.91856742e-01 -4.35456336e-01 -9.64267135e-01 8.15556705e-01 6.86043084e-01 2.78342396e-01 4.02231902e-01 -1.30985785e+00 -8.48898351e-01 1.39824236e-02 -8.57311726e-01 7.77792573e-01 2.23832458e-01 3.37229609e-01 6.90043986e-01 -1.13871229e+00 4.14875120e-01 1.11003113e+00 6.80315256e-01 4.24214453e-01 -1.03240561e+00 -4.98242378e-01 2.43115664e-01 9.34918165e-01 -1.40185058e+00 -3.58508885e-01 9.86033320e-01 -3.72584790e-01 1.41846895e+00 2.30308920e-01 1.13927627e+00 8.39648664e-01 2.47643158e-01 1.27205241e+00 8.33151579e-01 -9.18958858e-02 1.04982138e-01 -3.45813274e-01 -1.44413561e-01 6.29796267e-01 -4.49811101e-01 -2.09435776e-01 -7.90763557e-01 4.57029790e-01 1.16243923e+00 3.33361536e-01 -2.69728214e-01 -4.50561762e-01 -1.11699247e+00 4.23564255e-01 4.86359477e-01 2.47026414e-01 -4.86254781e-01 4.47907388e-01 5.13577163e-01 1.70598835e-01 6.81474626e-01 -4.93321568e-03 -4.92031217e-01 -5.17723680e-01 -7.45893478e-01 2.22122163e-01 -7.95652643e-02 9.62814391e-01 6.05942965e-01 2.07827076e-01 -4.04736191e-01 7.76329935e-01 -2.01876029e-01 6.46181405e-01 6.63855910e-01 -9.97006834e-01 2.57819712e-01 7.75017500e-01 -6.09221049e-02 -1.09992909e+00 -2.69314677e-01 -3.30350220e-01 -8.83819938e-01 9.35800448e-02 2.73063868e-01 3.75612199e-01 -1.01648355e+00 1.44361532e+00 5.84929645e-01 5.92102170e-01 -3.57573241e-01 1.09906638e+00 7.91282117e-01 7.44229376e-01 -1.46550804e-01 -1.85736895e-01 1.24853957e+00 -1.18215013e+00 -6.13802791e-01 -2.03964844e-01 5.02080321e-01 -5.12060285e-01 1.22402096e+00 1.62816226e-01 -1.32856083e+00 -7.43098795e-01 -7.18137264e-01 -5.95844626e-01 -2.18299761e-01 -1.19000226e-02 6.04217231e-01 1.92244217e-01 -1.05667210e+00 4.44550812e-01 -1.09072089e+00 -3.76093388e-01 8.02540183e-01 2.35463858e-01 -4.90333408e-01 -3.52654308e-02 -9.82482553e-01 6.97217882e-01 -1.22736588e-01 6.30032346e-02 -9.50046301e-01 -1.05149245e+00 -8.54938984e-01 1.48493424e-01 4.79380280e-01 -5.44314325e-01 1.26034725e+00 -7.64935195e-01 -1.44346035e+00 9.48329806e-01 -2.97683835e-01 -4.72936600e-01 3.58361483e-01 -3.81620765e-01 -4.60604392e-02 4.23146814e-01 2.83586174e-01 5.54965436e-01 1.13608110e+00 -5.52957416e-01 -5.89854240e-01 -6.90093219e-01 1.14868738e-01 4.61554438e-01 -4.05558765e-01 -1.12477541e-01 -1.00612724e+00 -8.73618305e-01 2.51685858e-01 -9.18875873e-01 1.34296417e-02 3.67092550e-01 1.44042984e-01 -3.04908007e-01 1.13866401e+00 -6.02380276e-01 1.05211914e+00 -2.30019379e+00 3.96748632e-01 -5.62500775e-01 2.15446353e-01 4.24053431e-01 -1.28257379e-01 1.91553980e-01 -1.44642204e-01 -2.10745558e-01 -8.85666385e-02 -3.23158473e-01 -2.63627261e-01 1.44105637e-02 -1.69625670e-01 4.27780271e-01 2.45553657e-01 1.25930679e+00 -8.56110275e-01 -4.43646818e-01 5.45240998e-01 5.94195664e-01 -7.57558405e-01 1.37428507e-01 4.49742004e-02 4.32259560e-01 -4.54052716e-01 7.00354636e-01 6.15945876e-01 -3.93156737e-01 -3.18675846e-01 -5.00381887e-01 -2.47897834e-01 6.91247135e-02 -8.03785264e-01 2.06703544e+00 -5.05922437e-01 8.27716947e-01 -9.48294923e-02 -1.25966263e+00 6.76332593e-01 1.23209171e-01 9.34015930e-01 -1.18556678e+00 1.05792664e-01 -1.22260615e-01 -3.86913300e-01 -7.42721975e-01 3.28591138e-01 4.21740003e-02 8.30840394e-02 3.10656309e-01 1.18164830e-01 -1.50504276e-01 6.69691339e-02 1.04031637e-01 1.25711048e+00 4.02463853e-01 7.59434104e-02 -1.01249166e-01 4.79525506e-01 7.98873603e-02 6.93109214e-01 4.52272862e-01 -4.58963126e-01 5.68077087e-01 3.50839198e-01 -7.37682521e-01 -1.03421783e+00 -9.39385831e-01 1.87858701e-01 9.54190850e-01 2.81316876e-01 -4.17301118e-01 -6.13522232e-01 -5.98955452e-01 5.41015603e-02 2.65414745e-01 -7.68219948e-01 -2.58050025e-01 -6.95695877e-01 -2.44855613e-01 2.91192800e-01 8.50003839e-01 9.07097518e-01 -1.02308726e+00 -8.90828311e-01 1.67406827e-01 -3.00754756e-01 -1.09348881e+00 -9.60359693e-01 1.19607663e-02 -9.19022620e-01 -9.88991082e-01 -9.34206665e-01 -6.97035372e-01 4.14032698e-01 9.28445876e-01 7.52865195e-01 -1.49871960e-01 -5.18500924e-01 6.21249557e-01 -4.72134531e-01 -2.58501887e-01 4.55893815e-01 -2.03314334e-01 6.77391365e-02 1.64184794e-01 4.60512102e-01 -8.10161650e-01 -8.57648969e-01 5.15305281e-01 -8.80798340e-01 2.18701541e-01 4.71197605e-01 9.05406356e-01 8.64482284e-01 2.27280930e-02 2.03791380e-01 -1.62220478e-01 8.06894302e-02 -2.33138904e-01 -4.45067137e-01 -2.08033025e-02 -6.41932636e-02 -1.66368887e-01 4.58477765e-01 -6.08552694e-01 -9.17944312e-01 1.96875766e-01 -4.10151482e-02 -1.22383165e+00 1.62947625e-01 2.97079355e-01 -9.91609171e-02 -1.63705960e-01 2.52167374e-01 6.95702732e-01 2.10042849e-01 -6.59010649e-01 1.69905543e-01 4.55989510e-01 3.99968088e-01 -2.42974252e-01 4.96064454e-01 6.26541972e-01 -2.11482178e-02 -1.03313315e+00 -8.60652924e-01 -4.49537337e-01 -6.46042347e-01 -3.47320318e-01 1.02420306e+00 -9.88603950e-01 -8.25249434e-01 9.92651999e-01 -9.11529422e-01 -7.78944910e-01 -4.33518857e-01 4.19030100e-01 -7.92845547e-01 4.34584111e-01 -5.04310369e-01 -4.71968919e-01 -3.39428693e-01 -1.11997616e+00 1.22789598e+00 1.55508071e-01 1.96987152e-01 -6.57627106e-01 -4.76021208e-02 6.33234024e-01 4.02421921e-01 -6.35576099e-02 6.63902223e-01 -3.10846027e-02 -8.34929943e-01 8.76815710e-03 -3.03442508e-01 3.06079924e-01 5.88558540e-02 -4.32482898e-01 -7.66910076e-01 -2.24080101e-01 1.61219925e-01 -2.12173790e-01 8.98013532e-01 6.80360675e-01 1.51292419e+00 -1.34098157e-01 -1.52472273e-01 8.79002810e-01 1.15794063e+00 4.22360927e-01 7.24768698e-01 1.75423697e-01 9.02480423e-01 3.01608145e-01 7.53072977e-01 5.65978527e-01 3.94501626e-01 1.19292545e+00 3.88878077e-01 4.08794964e-03 -4.50520009e-01 -2.43086904e-01 3.86834532e-01 7.79692471e-01 -2.81976283e-01 5.96480677e-03 -9.58778679e-01 8.21067572e-01 -2.06401992e+00 -1.33432198e+00 1.71460193e-02 2.00267506e+00 5.78973770e-01 -1.20222688e-01 1.54143587e-01 2.25926697e-01 5.67369819e-01 2.75491834e-01 -7.95874059e-01 -7.74016082e-02 -1.68862063e-02 2.29176432e-01 3.34697306e-01 2.57430792e-01 -1.16134584e+00 1.01795506e+00 5.26921558e+00 1.09544313e+00 -1.27335560e+00 2.88195848e-01 5.54102659e-01 -5.38448989e-01 1.62151441e-01 -2.84274131e-01 -5.50561905e-01 5.39468944e-01 4.01289821e-01 -2.85494000e-01 4.03113335e-01 8.22765410e-01 3.70071352e-01 -7.32788891e-02 -1.04280555e+00 1.53627276e+00 2.20466539e-01 -1.51947904e+00 1.41367525e-01 2.85070501e-02 4.97255087e-01 8.03576857e-02 1.03567883e-01 3.06908578e-01 -4.80652928e-01 -6.92076504e-01 6.91830575e-01 3.67222935e-01 8.31336141e-01 -5.76089740e-01 3.91486108e-01 1.28719032e-01 -1.52907550e+00 -3.23544890e-01 -2.89020002e-01 -2.26820245e-01 1.78411663e-01 3.78877342e-01 -1.20584510e-01 3.12001169e-01 1.32854533e+00 1.20795202e+00 -3.79456460e-01 9.87572968e-01 1.55877009e-01 2.27044329e-01 -2.85622299e-01 1.29875198e-01 3.23969811e-01 -7.86503777e-02 6.16722584e-01 9.38767791e-01 5.48203766e-01 4.99893129e-01 4.45659906e-02 4.89258826e-01 1.04465455e-01 -3.47704291e-02 -6.73669338e-01 4.79067303e-02 1.92742854e-01 9.20601785e-01 -5.38830221e-01 -4.61658269e-01 -5.91541111e-01 1.42883730e+00 2.48733297e-01 1.76834702e-01 -1.08412635e+00 -3.00684452e-01 8.91084552e-01 4.68303174e-01 8.62328410e-01 -3.90492588e-01 -8.33158940e-02 -1.24047637e+00 3.06560308e-01 -7.97193468e-01 2.82647163e-01 -1.11856639e+00 -1.06512296e+00 4.41483766e-01 -6.34636506e-02 -1.50751960e+00 3.54953349e-01 -5.68072438e-01 -4.44541454e-01 4.42567021e-01 -1.16219318e+00 -1.14789033e+00 -5.98689020e-01 1.05277705e+00 9.59336936e-01 -8.27927142e-02 4.47459877e-01 5.11703968e-01 -5.66260397e-01 5.31904161e-01 -1.45207241e-01 3.19082029e-02 5.33103764e-01 -1.01820087e+00 5.86602986e-01 7.61584163e-01 2.12165669e-01 1.57382220e-01 2.32735708e-01 -5.17001748e-01 -1.83357716e+00 -1.06864691e+00 6.92327857e-01 -4.99398023e-01 6.57284558e-01 -2.79823333e-01 -9.22386646e-01 5.64135432e-01 8.64361078e-02 4.36751664e-01 2.65724778e-01 -2.80748755e-01 -2.63596714e-01 -3.52188677e-01 -8.19724858e-01 6.41040146e-01 1.58127844e+00 -6.17017925e-01 -4.03953671e-01 3.60601515e-01 6.29916251e-01 -5.46620727e-01 -9.08455789e-01 2.94527888e-01 6.06933594e-01 -1.06454074e+00 1.19339311e+00 -3.46825749e-01 3.90788585e-01 -5.12232304e-01 -1.93507284e-01 -1.05978811e+00 -6.04605138e-01 -6.38484418e-01 -3.68394524e-01 8.17984641e-01 -1.31794155e-01 -4.34047371e-01 9.11162674e-01 4.23466384e-01 -3.32261384e-01 -9.89357531e-01 -1.24711549e+00 -8.06967378e-01 -2.56275117e-01 -5.14529586e-01 3.02867562e-01 7.42874205e-01 6.38214350e-02 2.92675823e-01 -4.78308082e-01 -1.30494516e-02 6.17872715e-01 1.65243134e-01 8.41532290e-01 -7.08914161e-01 -3.45101655e-01 -4.34724212e-01 -8.89832437e-01 -1.67225027e+00 -2.37626135e-01 -6.62059069e-01 -1.27116665e-01 -1.36220825e+00 2.53529727e-01 -2.03946710e-01 -3.41274142e-01 6.94467843e-01 -9.48242322e-02 6.13525391e-01 4.18532938e-01 2.65656739e-01 -6.49459660e-01 1.05074990e+00 1.55157435e+00 -3.58453333e-01 -2.62172818e-01 -1.85422987e-01 -3.27771693e-01 5.10682106e-01 7.59999335e-01 -2.70113438e-01 -6.31138742e-01 -7.39432752e-01 -2.08333686e-01 2.64876872e-01 6.16872668e-01 -1.14164257e+00 3.64061624e-01 -1.53159022e-01 3.54774117e-01 -8.25502157e-01 7.19663143e-01 -8.10020447e-01 2.51215249e-02 3.43076140e-01 -2.54278593e-02 3.04077655e-01 5.14147043e-01 5.56793511e-01 -2.98571676e-01 2.68387109e-01 6.22181058e-01 -2.69279778e-01 -1.12676883e+00 7.71127760e-01 -3.27596813e-01 1.03255697e-01 1.21825135e+00 -4.42571223e-01 -1.61481798e-01 -3.25231522e-01 -7.67196238e-01 1.38891935e-01 5.00426650e-01 5.49349248e-01 9.09424722e-01 -1.50569129e+00 -5.26601255e-01 4.05653507e-01 5.94805218e-02 -2.05826424e-02 1.01378572e+00 1.25471950e+00 -4.40757275e-01 4.61006761e-01 -6.46028459e-01 -9.17279840e-01 -1.37331200e+00 7.69743443e-01 3.10535967e-01 -5.04804403e-02 -9.64707732e-01 9.70909595e-01 5.92375338e-01 1.93946902e-02 1.21719904e-01 -2.86728710e-01 5.27827889e-02 -1.06418192e-01 6.90303385e-01 4.00244087e-01 -5.48312999e-02 -6.30614698e-01 -5.98640501e-01 1.09069264e+00 -3.02300118e-02 1.95154384e-01 1.43527281e+00 -1.67049915e-01 1.24867111e-01 2.41212368e-01 1.22682154e+00 -4.83358353e-01 -1.75508583e+00 -3.15666348e-01 -5.98324001e-01 -8.58329475e-01 4.32099223e-01 -4.41547424e-01 -1.45504653e+00 1.10267472e+00 6.37850702e-01 1.03143156e-01 1.40972030e+00 3.91544178e-02 9.11266446e-01 1.86205953e-01 3.80108446e-01 -1.14462721e+00 4.63511109e-01 4.15586680e-01 1.22311413e+00 -1.15672302e+00 -3.45510319e-02 -2.38774449e-01 -8.15706253e-01 6.30508840e-01 8.53854120e-01 -1.87782049e-01 7.94965446e-01 1.67001724e-01 -1.41212031e-01 -2.61754960e-01 -8.30480456e-01 -2.10139140e-01 2.55538404e-01 7.59864688e-01 1.61972880e-01 -2.70621777e-01 -2.96575185e-02 2.96595246e-01 3.92445505e-01 1.60968781e-01 2.20797166e-01 8.97801757e-01 -6.42986447e-02 -8.07057977e-01 -1.04789175e-02 3.15250397e-01 -2.39350647e-01 -1.30252659e-01 1.54401828e-02 6.07284009e-01 4.46653403e-02 6.31345153e-01 3.68080080e-01 -5.38421452e-01 4.80458677e-01 -5.09823337e-02 7.61174619e-01 -3.29894543e-01 -3.21546286e-01 1.29510432e-01 -2.24475086e-01 -1.03814280e+00 -5.70194125e-01 -8.07287335e-01 -1.05000699e+00 -3.13924134e-01 9.05202404e-02 -2.34226108e-01 3.88566852e-01 6.35796726e-01 9.44189847e-01 5.96967220e-01 4.71663296e-01 -1.29333234e+00 -2.80425906e-01 -8.53180528e-01 -4.27086741e-01 3.68729889e-01 1.73276156e-01 -1.00754082e+00 -4.53434773e-02 2.30871871e-01]
[8.68067455291748, 0.3186199963092804]
4f16fb02-443b-4cb9-8f74-9600f7347b06
anything-3d-towards-single-view-anything
2304.10261
null
https://arxiv.org/abs/2304.10261v1
https://arxiv.org/pdf/2304.10261v1.pdf
Anything-3D: Towards Single-view Anything Reconstruction in the Wild
3D reconstruction from a single-RGB image in unconstrained real-world scenarios presents numerous challenges due to the inherent diversity and complexity of objects and environments. In this paper, we introduce Anything-3D, a methodical framework that ingeniously combines a series of visual-language models and the Segment-Anything object segmentation model to elevate objects to 3D, yielding a reliable and versatile system for single-view conditioned 3D reconstruction task. Our approach employs a BLIP model to generate textural descriptions, utilizes the Segment-Anything model for the effective extraction of objects of interest, and leverages a text-to-image diffusion model to lift object into a neural radiance field. Demonstrating its ability to produce accurate and detailed 3D reconstructions for a wide array of objects, \emph{Anything-3D\footnotemark[2]} shows promise in addressing the limitations of existing methodologies. Through comprehensive experiments and evaluations on various datasets, we showcase the merits of our approach, underscoring its potential to contribute meaningfully to the field of 3D reconstruction. Demos and code will be available at \href{https://github.com/Anything-of-anything/Anything-3D}{https://github.com/Anything-of-anything/Anything-3D}.
['Xinchao Wang', 'Xingyi Yang', 'Qiuhong Shen']
2023-04-19
null
null
null
null
['3d-reconstruction']
['computer-vision']
[ 2.58107811e-01 -5.49190417e-02 1.51625782e-01 -3.05731893e-01 -8.37449014e-01 -6.05069816e-01 5.58790684e-01 -2.04273850e-01 7.94036463e-02 2.98102319e-01 7.23465383e-02 -4.17620331e-01 -1.24876708e-01 -6.50082827e-01 -5.81312954e-01 -6.03764474e-01 7.70539865e-02 3.76179665e-01 2.11163431e-01 -2.77376771e-01 2.33635202e-01 1.03860354e+00 -1.60068738e+00 2.47305378e-01 5.46817362e-01 1.19054592e+00 5.24677515e-01 5.32623351e-01 -1.38966769e-01 5.86386681e-01 -1.89242333e-01 -1.93470955e-01 7.35631883e-01 -4.30830836e-01 -5.67847490e-01 3.44270796e-01 6.11527801e-01 -6.02684319e-01 -2.80439049e-01 7.86681473e-01 5.55993021e-01 -3.94915938e-02 8.18538904e-01 -9.00839508e-01 -7.26562738e-01 -2.03304127e-01 -6.15179479e-01 1.16895944e-01 4.85160768e-01 3.36354494e-01 8.31563711e-01 -1.14162767e+00 7.36231327e-01 1.04638040e+00 6.61871493e-01 5.33113301e-01 -1.30156302e+00 -6.04886711e-01 8.25543255e-02 -3.13344330e-01 -1.41384149e+00 -4.31519926e-01 1.01049161e+00 -5.71173370e-01 9.84915078e-01 3.96291643e-01 8.64658117e-01 9.95475292e-01 1.68694988e-01 9.14956093e-01 1.38070762e+00 -3.31293046e-01 1.67622447e-01 6.86650053e-02 -1.53582856e-01 6.80510104e-01 1.36323916e-02 1.92839205e-01 -6.62874103e-01 1.50659140e-02 1.04407620e+00 5.45355789e-02 -3.80637795e-01 -4.31572348e-01 -1.21834409e+00 4.31876570e-01 6.72491372e-01 -2.99460385e-02 -3.72988462e-01 3.30223620e-01 -1.42710224e-01 -1.36322901e-01 6.85488105e-01 2.10788548e-02 -1.30568460e-01 2.25892261e-01 -9.54108238e-01 3.74393225e-01 3.57974529e-01 1.07885253e+00 6.84522092e-01 2.49416664e-01 2.83330917e-01 7.85048664e-01 5.07214427e-01 7.49028504e-01 -9.54011083e-02 -1.34936416e+00 2.04252586e-01 4.77867305e-01 1.20209798e-01 -7.11746216e-01 -4.40987885e-01 -4.27194685e-01 -6.70211315e-01 7.37013340e-01 1.48908973e-01 2.40325928e-01 -1.08617115e+00 1.32367635e+00 6.30676270e-01 -1.46710947e-01 -1.92920551e-01 1.16898108e+00 1.04152811e+00 5.75866342e-01 -1.35228768e-01 1.45809174e-01 1.14461815e+00 -4.84291852e-01 -2.66878515e-01 -4.40773606e-01 1.00550011e-01 -7.17748702e-01 1.17926610e+00 4.54544604e-01 -1.37483895e+00 -3.21105123e-01 -9.23314035e-01 -2.50749588e-01 -2.40122586e-01 -2.64554888e-01 5.64328790e-01 3.93369853e-01 -1.17534649e+00 3.49392414e-01 -8.36959183e-01 -4.24050927e-01 8.05248559e-01 1.80719644e-01 -3.42607141e-01 -3.24759364e-01 -4.56418127e-01 7.73608923e-01 3.37684453e-01 1.29403546e-01 -9.72541988e-01 -6.15399778e-01 -7.65376151e-01 -4.16124672e-01 1.88837886e-01 -1.02471232e+00 1.05886555e+00 -5.64540863e-01 -1.16388023e+00 1.13035321e+00 2.63479422e-03 -2.19977777e-02 5.47672153e-01 -1.69378281e-01 -3.44413407e-02 4.24117208e-01 3.89593802e-02 7.90651560e-01 8.10539365e-01 -1.98455918e+00 -3.56360674e-01 -6.17164612e-01 1.19120451e-02 5.00097454e-01 1.49710938e-01 -1.57819226e-01 -7.01619685e-01 -7.97055066e-01 6.34227574e-01 -7.92399049e-01 -1.15737058e-01 4.44398135e-01 -5.01805842e-01 3.36969912e-01 7.18536377e-01 -6.47683144e-01 6.94550097e-01 -2.16085696e+00 1.76266447e-01 6.11262918e-02 2.91492611e-01 -1.24327078e-01 2.91291904e-02 5.64862847e-01 1.96479976e-01 1.69427887e-01 -5.80749393e-01 -3.82243603e-01 -2.95077208e-02 2.48399135e-02 -2.34969199e-01 6.42704964e-01 2.71486312e-01 9.95992839e-01 -6.44895017e-01 -3.19247395e-01 6.41022742e-01 8.67009640e-01 -3.94438654e-01 1.50973469e-01 -4.27366585e-01 7.08532989e-01 -6.32129729e-01 1.10622633e+00 7.98547506e-01 -2.49391541e-01 -2.03760684e-01 -1.79066703e-01 -1.18829742e-01 -5.98718598e-02 -1.10356712e+00 1.97394085e+00 -3.76507163e-01 4.36310440e-01 3.87009889e-01 -5.46318948e-01 9.40969646e-01 -9.30814222e-02 5.28069019e-01 -8.56957734e-01 1.73862666e-01 4.51963663e-01 -5.38309097e-01 -4.88979697e-01 4.79189962e-01 -4.41709101e-01 -4.77481931e-02 4.30159450e-01 -1.16904434e-02 -9.29917037e-01 -2.75322884e-01 3.00646454e-01 8.26744974e-01 7.46833265e-01 1.53887197e-01 -1.56715080e-01 5.09024598e-02 1.80620611e-01 1.03881948e-01 6.48742259e-01 -9.15938690e-02 1.21582103e+00 -1.17060475e-01 -2.56656080e-01 -1.04595649e+00 -1.50103962e+00 -3.68955880e-01 4.99285072e-01 5.23635209e-01 -1.95905901e-02 -3.91894162e-01 -4.06669825e-01 1.68035150e-01 9.46473300e-01 -5.52624047e-01 9.92469043e-02 -2.91939050e-01 -6.60759270e-01 2.84102678e-01 3.35526168e-01 3.83838028e-01 -8.90301466e-01 -9.02354658e-01 -7.12509677e-02 -2.80048162e-01 -1.11846399e+00 -7.83411488e-02 4.64942038e-01 -9.90032911e-01 -7.30695426e-01 -8.56626928e-01 -4.68102992e-01 7.12968588e-01 4.93800044e-01 1.11939478e+00 1.80477947e-01 -5.17602861e-01 7.65024006e-01 -3.95418614e-01 -4.14845318e-01 -2.06863239e-01 -4.97594625e-01 -1.67374521e-01 -2.79358923e-01 9.00959820e-02 -7.83495605e-01 -9.75273848e-01 3.10795695e-01 -1.17169392e+00 3.76183659e-01 4.04532194e-01 3.01277786e-01 8.51118803e-01 -3.15150291e-01 2.10032046e-01 -3.93194109e-01 2.57527828e-01 -4.53464031e-01 -5.17291486e-01 -1.09968893e-01 -4.68822896e-01 -1.92706481e-01 2.44667217e-01 -2.12480530e-01 -9.35586393e-01 1.34916365e-01 -2.97285289e-01 -4.81399029e-01 -4.77574259e-01 1.17609844e-01 -1.54158384e-01 -6.66085407e-02 7.05498874e-01 4.81717020e-01 -7.37750754e-02 -6.78709686e-01 5.31087935e-01 6.97199404e-01 5.40830493e-01 -4.93004173e-01 9.67526615e-01 1.01058638e+00 4.21999469e-02 -9.01002526e-01 -8.21956098e-01 -3.90403122e-01 -8.81461799e-01 -6.41125381e-01 8.36359203e-01 -9.26091552e-01 -3.20325434e-01 5.02501547e-01 -9.13897455e-01 -7.38788784e-01 -5.36563933e-01 4.23758239e-01 -8.43513906e-01 2.77701080e-01 -3.92867982e-01 -1.00518715e+00 -1.69997334e-01 -1.06360459e+00 1.37068236e+00 1.85201988e-01 -1.06171355e-01 -9.14798498e-01 -2.02597544e-01 7.36399531e-01 3.00452560e-01 6.03531301e-01 8.83644819e-01 -6.77419156e-02 -1.01566780e+00 3.41249928e-02 -3.06336671e-01 3.73588532e-01 -1.19892217e-01 -2.52585471e-01 -1.15886700e+00 -1.45062417e-01 -5.63262440e-02 -4.82867718e-01 7.93178380e-01 4.03084219e-01 1.01834190e+00 1.20894380e-01 -1.17456980e-01 9.28591013e-01 1.73287129e+00 -3.82150970e-02 6.51795745e-01 5.12225807e-01 6.43615305e-01 4.97323483e-01 5.19252658e-01 5.75343370e-01 5.46489298e-01 6.80440784e-01 9.11568761e-01 -2.17014924e-01 -7.46042371e-01 -2.80465066e-01 1.01106822e-01 5.60830057e-01 -2.46092305e-01 -4.02133644e-01 -9.73041594e-01 4.89012569e-01 -1.33519590e+00 -7.12822795e-01 -2.86269695e-01 2.19604540e+00 6.20474696e-01 -9.60138068e-03 -1.06044002e-02 -7.04975426e-02 2.24667549e-01 2.74669051e-01 -7.01243818e-01 -1.58414945e-01 -3.13562304e-01 3.85826677e-02 4.20339257e-01 5.89334726e-01 -6.48725986e-01 9.45885003e-01 5.85363054e+00 7.18933344e-01 -1.07023323e+00 2.70914491e-02 6.06506765e-01 -2.88254321e-01 -6.34212196e-01 -1.68835819e-02 -7.87098110e-01 1.38100147e-01 3.98656636e-01 2.07444325e-01 5.54076672e-01 4.01484132e-01 4.27881628e-01 -5.11949897e-01 -8.69941115e-01 1.01508677e+00 3.47833395e-01 -1.19780183e+00 1.54644102e-01 2.10967734e-01 6.84693456e-01 2.72978961e-01 1.44731209e-01 -3.03159297e-01 2.63268054e-01 -9.52170074e-01 1.39767528e+00 7.25252271e-01 9.64803636e-01 -3.19088966e-01 1.66338459e-01 4.62297827e-01 -1.03058374e+00 2.90999208e-02 -1.51648626e-01 9.01514366e-02 4.61003870e-01 6.63380980e-01 -7.36775279e-01 7.28223503e-01 1.13807571e+00 6.26031458e-01 -5.18020689e-01 9.94154871e-01 -1.69225037e-01 2.28822544e-01 -4.97325033e-01 1.85932577e-01 8.46233144e-02 -3.23005855e-01 6.36990488e-01 1.17619860e+00 5.22484958e-01 4.30108607e-01 1.45849381e-02 1.11695766e+00 -3.63006666e-02 3.63398753e-02 -7.90378332e-01 2.50347435e-01 2.47057259e-01 1.07340074e+00 -1.06054461e+00 -2.79256757e-02 -3.39097917e-01 1.23688173e+00 2.36541584e-01 5.82372427e-01 -6.94910645e-01 -1.58915430e-01 2.19016060e-01 4.98521239e-01 4.77388918e-01 -6.20723426e-01 -6.49443567e-01 -1.00113809e+00 1.58006668e-01 -6.26740575e-01 9.38139483e-03 -1.56302285e+00 -1.24361491e+00 5.64651072e-01 3.05471212e-01 -1.37465739e+00 2.24511877e-01 -5.56909978e-01 -1.23390868e-01 7.96932578e-01 -1.51779151e+00 -1.48538685e+00 -6.69210076e-01 7.16729462e-01 6.95258260e-01 3.11167151e-01 5.15995622e-01 1.08045757e-01 -2.18548715e-01 -1.42763898e-01 1.29495069e-01 -3.14599633e-01 3.69522095e-01 -1.03562331e+00 3.66664588e-01 6.65518939e-01 2.17720136e-01 3.62220138e-01 6.10859513e-01 -8.07354212e-01 -1.56657422e+00 -7.83518255e-01 2.93245852e-01 -7.62827873e-01 4.29451495e-01 -5.58146536e-01 -6.32785916e-01 6.30036294e-01 8.26180950e-02 5.54619655e-02 5.17173827e-01 -5.00065029e-01 -2.91491777e-01 1.37924850e-01 -1.29279375e+00 6.60980880e-01 1.29106593e+00 -5.06101370e-01 -4.83318120e-01 2.31689513e-01 4.66610342e-01 -5.29200196e-01 -9.11243618e-01 4.35600847e-01 6.29554033e-01 -1.44561672e+00 1.28589165e+00 5.01905419e-02 4.34853941e-01 -3.93938869e-01 -6.45734370e-01 -9.79923308e-01 -1.78078160e-01 -3.78748357e-01 -1.63160473e-01 1.11839652e+00 2.58698732e-01 -5.15817642e-01 5.50992370e-01 5.94092548e-01 -3.23051900e-01 -7.59470463e-01 -9.17559981e-01 -6.10280037e-01 1.43253341e-01 -9.25897777e-01 3.57914716e-01 7.56198883e-01 -5.36982298e-01 -6.37552664e-02 -1.82556406e-01 3.22737575e-01 9.79793251e-01 2.89742023e-01 6.89886630e-01 -1.06594718e+00 -1.83259398e-01 -4.63197947e-01 -2.40069419e-01 -1.38742316e+00 -2.46498555e-01 -1.17298079e+00 8.04853532e-03 -2.09533858e+00 1.84170723e-01 -6.64733469e-01 1.40573680e-01 3.55701655e-01 2.36241534e-01 8.54718864e-01 4.27744150e-01 4.47541893e-01 -3.38650107e-01 6.46507680e-01 1.35452414e+00 -6.36241660e-02 -1.49279490e-01 -1.29403830e-01 -7.77493954e-01 7.55453825e-01 6.90554857e-01 -3.68990093e-01 -2.59139895e-01 -7.11440384e-01 1.19127542e-01 -3.50814797e-02 8.72007370e-01 -8.03229511e-01 -6.12541102e-02 -1.91809878e-01 6.86518788e-01 -8.14385056e-01 8.82306635e-01 -9.64259148e-01 3.26309055e-01 1.30518496e-01 6.21376745e-02 -1.86021432e-01 4.04377818e-01 5.70558369e-01 2.35595003e-01 -9.60776880e-02 7.60753512e-01 -5.27654052e-01 -7.94023931e-01 2.81826943e-01 -9.46793556e-02 3.50885913e-02 1.04611635e+00 -6.66310012e-01 -2.45490357e-01 -2.85457462e-01 -8.23171675e-01 -3.03353779e-02 9.91926193e-01 3.79290342e-01 1.06872129e+00 -1.19792140e+00 -6.98497772e-01 2.33103812e-01 1.79795295e-01 3.89658839e-01 3.33143324e-01 8.73098910e-01 -7.84546375e-01 1.77126601e-02 -1.38323382e-01 -9.73199546e-01 -1.02643871e+00 2.57394373e-01 4.35860217e-01 3.79830748e-01 -1.01461864e+00 8.12557518e-01 2.75682181e-01 -6.79490268e-01 1.24279484e-01 -2.14384422e-01 4.62258786e-01 -1.85439840e-01 2.41136760e-01 2.18554303e-01 -2.20572408e-02 -8.36720347e-01 -3.16427678e-01 9.75444794e-01 3.07198882e-01 -3.69849384e-01 1.69471538e+00 -4.91034567e-01 7.96006471e-02 6.32887900e-01 9.73594725e-01 1.47886397e-02 -1.60878718e+00 -9.70878080e-02 -4.08574581e-01 -7.57178187e-01 2.49173164e-01 -1.06283545e+00 -9.09698546e-01 9.51285720e-01 6.40558362e-01 -3.66213880e-02 1.20533872e+00 4.70341533e-01 5.66885054e-01 -7.77588859e-02 6.67501390e-01 -5.82928836e-01 1.63021624e-01 1.89780131e-01 1.05694735e+00 -1.12140846e+00 3.65639359e-01 -3.26138854e-01 -7.00362742e-01 9.53107059e-01 3.07805091e-01 3.37614715e-02 6.45703435e-01 3.68392617e-01 1.78181693e-01 -6.08137548e-01 -2.82091856e-01 -2.91559309e-01 2.17157602e-01 8.61025095e-01 1.29035503e-01 -2.64288843e-01 1.95243239e-01 5.80428503e-02 -7.85104409e-02 -1.09936662e-01 4.48734462e-01 1.21531022e+00 -4.63984787e-01 -7.60536015e-01 -6.20347500e-01 2.50320554e-01 -1.38939381e-01 -1.03356190e-01 -4.85019535e-01 8.27623367e-01 8.07379261e-02 7.56036341e-01 -1.99415550e-01 -2.20970154e-01 3.79634172e-01 -9.52907503e-02 7.56777585e-01 -4.73024130e-01 -4.26830500e-01 3.69681209e-01 -1.07033826e-01 -6.15751386e-01 -5.22957563e-01 -7.16291189e-01 -1.43274415e+00 -1.15175068e-01 -1.50421038e-01 -5.33367634e-01 9.56322193e-01 6.39733255e-01 3.36132675e-01 2.45420694e-01 5.34924448e-01 -1.52326798e+00 -1.86089739e-01 -4.97570634e-01 -8.03770542e-01 3.83523941e-01 3.88316870e-01 -6.10057831e-01 -4.79468197e-01 1.37934878e-01]
[9.016722679138184, -2.9570748805999756]
626a6d33-06c5-4714-ab91-d5d6af6e87a3
re2g-retrieve-rerank-generate
null
null
https://openreview.net/forum?id=_R7UMusdRsc
https://openreview.net/pdf?id=_R7UMusdRsc
Re2G: Retrieve, Rerank, Generate
As demonstrated by GPT-3 and T5, transformers grow in capability as parameter spaces become larger and larger. However, for tasks that require a large amount of knowledge, non-parametric memory allows models to grow dramatically with a sub-linear increase in computational cost and GPU memory requirements. Recent models such as RAG and REALM have introduced retrieval into conditional generation. These models incorporate neural initial retrieval from a corpus of passages. We build on this line of research, proposing Re2G, which combines both neural initial retrieval and reranking into a BART-based sequence-to-sequence generation. Our reranking approach also permits merging retrieval results from sources with incomparable scores, enabling an ensemble of BM25 and neural initial retrieval. To train our system end-to-end, we introduce a novel variation of knowledge distillation to train the initial retrieval, reranker and generation using only ground truth on the target sequence output. We find large gains in four diverse tasks: zero-shot slot filling, question answering, fact checking and dialog, with relative gains of 9% to 34% over the previous state-of-the-art on the KILT leaderboard. We make our code available as open source.
['Anonymous']
2022-01-16
null
null
null
acl-arr-january-2022-1
['zero-shot-slot-filling', 'slot-filling']
['natural-language-processing', 'natural-language-processing']
[ 2.02496320e-01 3.17751318e-01 -2.33058184e-01 -6.01670444e-02 -1.72724223e+00 -7.93420315e-01 7.15298772e-01 3.43388975e-01 -4.97891396e-01 9.70363855e-01 5.03790140e-01 -4.36432183e-01 -8.79651122e-03 -8.68716538e-01 -7.87469089e-01 8.90371948e-02 1.01340592e-01 1.20339894e+00 5.30762076e-01 -5.81442654e-01 2.41588369e-01 -2.15090588e-01 -1.36177182e+00 7.31313169e-01 1.01407671e+00 9.60543454e-01 8.64683464e-02 1.03431571e+00 -3.64245325e-01 1.01724231e+00 -7.76803017e-01 -5.03382146e-01 2.21335694e-01 -4.26516593e-01 -1.08093095e+00 -3.31794977e-01 5.11162281e-01 -4.83112067e-01 -3.35276693e-01 3.83475304e-01 8.18231761e-01 6.09779954e-01 4.40021962e-01 -8.94127011e-01 -1.05752921e+00 1.02556300e+00 -2.33499616e-01 7.07194805e-02 6.69937313e-01 1.97101802e-01 1.31894922e+00 -9.22634304e-01 7.80110359e-01 1.34632659e+00 6.62548542e-01 5.55462718e-01 -1.26522350e+00 -4.96785164e-01 -4.93957028e-02 1.52779832e-01 -1.15233862e+00 -4.71754700e-01 -1.31093543e-02 -2.29906663e-01 1.58110642e+00 3.29223901e-01 3.18482131e-01 8.16976428e-01 -5.56923866e-01 1.10688305e+00 6.62976384e-01 -6.10604942e-01 1.57558098e-01 -4.33338713e-03 2.53027588e-01 5.94467282e-01 -6.61432520e-02 -5.65376468e-02 -7.65490592e-01 -5.54139555e-01 6.10050440e-01 -3.33361715e-01 -1.38170227e-01 2.66960323e-01 -1.22049809e+00 1.02527523e+00 3.48524272e-01 -4.05748375e-02 -3.45742017e-01 1.45489663e-01 3.28732014e-01 3.36397171e-01 5.50087571e-01 9.77114677e-01 -5.57817698e-01 -4.15918022e-01 -1.26450109e+00 6.82392657e-01 1.25950933e+00 9.91546094e-01 7.20770657e-01 -2.80862093e-01 -1.09029579e+00 1.09275019e+00 -6.00432791e-02 3.69515330e-01 7.08456814e-01 -1.09223962e+00 7.55388737e-01 4.52000320e-01 4.45061505e-01 -4.54734594e-01 -2.49138847e-01 -2.95414627e-01 -3.67519259e-01 -4.49382842e-01 3.86485577e-01 -3.21323663e-01 -1.36373913e+00 1.63963175e+00 2.63813674e-01 2.98367858e-01 2.53033131e-01 7.39347756e-01 8.17908943e-01 9.36470926e-01 1.55070513e-01 1.68140829e-01 1.38774407e+00 -1.42723286e+00 -4.19496536e-01 -3.27179134e-01 8.33688855e-01 -9.82795000e-01 1.17373192e+00 1.33749261e-01 -1.21905148e+00 -3.29247564e-01 -7.50997901e-01 -5.45144320e-01 -4.00508583e-01 1.75132174e-02 7.58132577e-01 4.57672179e-01 -1.25227213e+00 5.85170746e-01 -6.62115276e-01 -2.99614459e-01 1.72526196e-01 6.63557947e-02 2.75645852e-02 -1.48520172e-01 -1.68292332e+00 9.95439172e-01 7.09329545e-01 -1.81985542e-01 -7.19977081e-01 -1.06006300e+00 -6.94415390e-01 6.16095848e-02 5.12965977e-01 -1.08052123e+00 1.83152354e+00 -3.09752673e-01 -1.65356612e+00 5.00355482e-01 -1.92497045e-01 -9.33713853e-01 4.88506854e-01 -6.19518340e-01 -4.57241014e-02 3.91349569e-03 1.87655464e-01 8.97552371e-01 3.52431774e-01 -7.80502558e-01 -6.27331495e-01 1.40070602e-01 1.71870708e-01 4.41594899e-01 -2.09963638e-02 9.24391905e-04 -7.47419536e-01 -5.24433911e-01 -1.44606620e-01 -9.80512500e-01 -3.45984578e-01 -5.36822557e-01 -5.34334242e-01 -4.05665725e-01 3.01265061e-01 -9.95029271e-01 1.39093959e+00 -1.61604178e+00 -7.76066631e-02 -1.37861073e-02 -3.13617811e-02 4.00324255e-01 -4.79519814e-01 8.15602779e-01 4.48514462e-01 2.97815204e-01 -2.30688542e-01 -3.35772008e-01 3.48896861e-01 2.27280315e-02 -5.89281142e-01 -4.16166812e-01 2.59649277e-01 1.34738481e+00 -1.02254534e+00 -4.03948426e-01 -1.73208684e-01 2.17554301e-01 -9.18344378e-01 2.90186852e-01 -8.74154270e-01 1.27608091e-01 -4.06760454e-01 4.60773319e-01 1.91212460e-01 -7.12185562e-01 -1.06283538e-01 2.80601501e-01 2.93063223e-01 7.69095302e-01 -8.70915830e-01 2.09489179e+00 -6.33246958e-01 4.36432511e-01 -3.55933398e-01 -3.64859223e-01 6.74018621e-01 3.87504935e-01 -6.85602007e-03 -7.06632793e-01 -2.02823833e-01 3.47330123e-01 -2.29510903e-01 -2.63659745e-01 1.31566226e+00 -4.17380445e-02 -2.17534259e-01 6.74527466e-01 2.05665499e-01 -3.69538903e-01 6.31697178e-01 7.25140333e-01 1.30410421e+00 6.55930564e-02 -7.85205737e-02 2.82379895e-01 7.95246288e-02 3.22872192e-01 2.07498465e-02 1.25484526e+00 4.51953709e-01 6.56930685e-01 2.75237858e-01 6.87684398e-03 -1.10658157e+00 -1.01664567e+00 1.94094405e-01 1.64641845e+00 -1.08133815e-01 -4.80210543e-01 -8.05549741e-01 -4.04725164e-01 3.32482070e-01 1.03070998e+00 -4.12061691e-01 -7.35203102e-02 -7.00047135e-01 -8.00953746e-01 8.93984735e-01 5.71784317e-01 3.93180072e-01 -1.11501431e+00 -3.67410749e-01 5.97249091e-01 -5.68187892e-01 -1.19398880e+00 -4.36815977e-01 6.20317757e-02 -8.79153967e-01 -5.17380357e-01 -9.48060393e-01 -4.50867057e-01 1.29363865e-01 2.62619518e-02 1.75322306e+00 1.78002194e-01 -2.85124481e-01 3.30057979e-01 -5.35643578e-01 -2.38966122e-01 -5.27958989e-01 6.82607234e-01 -4.20215487e-01 -6.52926683e-01 1.28301829e-01 -3.75987738e-01 -7.71936893e-01 1.30661875e-01 -9.89384770e-01 1.69899359e-01 6.36596382e-01 1.03366029e+00 4.94392067e-01 -5.81558943e-01 8.46348822e-01 -1.10669911e+00 1.03620613e+00 -6.82004273e-01 -3.17846268e-01 4.26966190e-01 -5.91005921e-01 4.10323679e-01 2.61913240e-01 -4.22386259e-01 -1.15389705e+00 -4.88507986e-01 -3.30650538e-01 -3.21796507e-01 2.66767085e-01 7.63507068e-01 4.88691151e-01 4.74671751e-01 1.15442216e+00 1.98974475e-01 -3.01049650e-01 -4.91904587e-01 1.13027954e+00 5.48798859e-01 6.37450039e-01 -8.16452801e-01 5.28917372e-01 -4.46862839e-02 -6.34724081e-01 -3.60105097e-01 -1.09591985e+00 -7.09360003e-01 -1.39613926e-01 3.22387695e-01 5.84666014e-01 -1.08242440e+00 -4.58258688e-01 2.07456499e-01 -1.32025015e+00 -7.51656771e-01 -4.93173689e-01 1.23792253e-01 -5.46193957e-01 2.23429829e-01 -1.05246723e+00 -7.26900280e-01 -9.25611496e-01 -6.46678627e-01 1.40962434e+00 2.82453269e-01 -4.13445681e-01 -8.35754633e-01 2.71911383e-01 6.77726388e-01 7.25184619e-01 -1.52687743e-01 9.61706817e-01 -1.03599250e+00 -1.00857651e+00 -3.21263611e-01 -3.47596645e-01 8.13562348e-02 -3.94635499e-01 -4.71289992e-01 -8.71380568e-01 -1.60145536e-01 -6.38525844e-01 -7.59653747e-01 1.20091057e+00 2.90348157e-02 7.68779933e-01 -3.81681383e-01 -3.03488076e-01 2.52061605e-01 1.13823438e+00 5.11834547e-02 6.03371501e-01 3.69647890e-01 3.96477759e-01 4.99903262e-01 6.45817876e-01 4.01333988e-01 4.80368555e-01 6.66553080e-01 -8.53462815e-02 1.72687054e-01 -2.67512262e-01 -6.24578655e-01 1.50766090e-01 8.79627764e-01 2.08405241e-01 -6.78670704e-01 -8.84096205e-01 7.81589150e-01 -1.99820757e+00 -8.27943206e-01 3.80064994e-01 2.14191079e+00 1.40415263e+00 1.07549824e-01 8.29180703e-02 -4.05446738e-01 5.87959528e-01 6.11940883e-02 -6.81912363e-01 -2.68046767e-01 -3.39413323e-02 5.60124874e-01 3.96141201e-01 7.70289183e-01 -8.14021289e-01 1.46449149e+00 6.49138260e+00 1.17410529e+00 -7.53565669e-01 1.80187747e-01 8.06111395e-01 -2.06010848e-01 -4.32358593e-01 1.57355160e-01 -1.09614897e+00 3.13238442e-01 1.34468162e+00 -3.61462891e-01 5.42510986e-01 6.96316659e-01 -3.01196665e-01 -2.79133141e-01 -9.55480754e-01 6.00180984e-01 3.23405340e-02 -1.61112034e+00 3.31125170e-01 -4.32963014e-01 8.63831162e-01 2.68387884e-01 -2.07903888e-02 9.53884780e-01 9.91627932e-01 -1.06532240e+00 5.58851242e-01 5.52595675e-01 6.80823386e-01 -4.31288511e-01 5.40273249e-01 4.66630697e-01 -7.61118650e-01 6.71908855e-02 -3.88574928e-01 -2.53143813e-02 6.23655379e-01 5.40537417e-01 -1.48375750e+00 5.84874511e-01 2.32615784e-01 3.08783036e-02 -4.53260630e-01 1.06237674e+00 -2.01429650e-01 7.53686905e-01 -5.47397494e-01 -1.85467824e-01 3.08232874e-01 2.75140733e-01 3.55654806e-01 1.41783309e+00 4.22452509e-01 2.93036163e-01 3.00162494e-01 8.64705086e-01 -3.78263354e-01 1.11811608e-01 -6.78367764e-02 -2.44747177e-01 7.70019352e-01 1.07753038e+00 -5.81974685e-01 -8.78770053e-01 -1.03111923e-01 1.09798074e+00 4.69711334e-01 4.15293694e-01 -7.14895129e-01 -5.33265650e-01 4.42161947e-01 -6.80759028e-02 4.15373981e-01 -1.42366633e-01 -1.61708087e-01 -1.15436077e+00 -7.17210919e-02 -8.94053519e-01 5.04451752e-01 -7.36103952e-01 -1.31541562e+00 7.99518287e-01 7.73719624e-02 -6.91186965e-01 -1.02619267e+00 -2.28044868e-01 -2.18380108e-01 1.21414077e+00 -1.52176332e+00 -9.86931860e-01 -5.45257293e-02 2.28896201e-01 6.77412629e-01 3.04007437e-02 1.05991900e+00 3.19160700e-01 -1.10795528e-01 7.76715040e-01 1.60746828e-01 1.47222161e-01 7.70898044e-01 -1.31211174e+00 1.11428750e+00 6.04226470e-01 3.38945746e-01 8.25198472e-01 5.88001013e-01 -7.70016134e-01 -1.17068839e+00 -9.43020105e-01 1.06978583e+00 -9.09167051e-01 8.16274285e-01 -3.75109166e-01 -9.14509594e-01 6.72792494e-01 2.13695601e-01 -2.59269118e-01 6.74090922e-01 4.57691491e-01 -6.21730566e-01 3.91180068e-01 -7.55329907e-01 6.18556321e-01 1.03446329e+00 -6.75850511e-01 -6.95065677e-01 6.12693608e-01 1.32439673e+00 -9.16629493e-01 -8.49486351e-01 2.64896572e-01 4.86924589e-01 -5.90334833e-01 1.09663773e+00 -6.46095812e-01 4.49379891e-01 -1.04921073e-01 -4.89380769e-02 -1.26686084e+00 -1.36322677e-01 -9.81444836e-01 -3.62252355e-01 1.11185980e+00 1.04738164e+00 -5.62595367e-01 1.02230549e+00 7.85956562e-01 -1.33450612e-01 -7.79335976e-01 -7.40943134e-01 -7.41958797e-01 2.80361325e-01 -3.88619542e-01 7.06445754e-01 5.24621487e-01 7.38911778e-02 7.90689707e-01 -3.66003752e-01 -3.62034857e-01 7.62411952e-02 2.50931412e-01 8.20994854e-01 -9.48575675e-01 -7.19915330e-01 -2.91466922e-01 1.34002626e-01 -1.61850011e+00 -1.57334685e-01 -9.87932682e-01 3.06161135e-01 -1.80130506e+00 2.18417645e-01 -6.31347358e-01 -1.75902411e-01 6.32495284e-01 -6.07687473e-01 1.90896183e-01 3.70394170e-01 1.94664061e-01 -9.76247311e-01 5.21714926e-01 1.14530373e+00 -1.44817606e-02 -4.52858686e-01 -2.71824300e-01 -8.59838545e-01 1.72267318e-01 5.93714714e-01 -3.16365987e-01 -5.12940764e-01 -7.49047935e-01 4.09353107e-01 5.19372642e-01 1.11287676e-01 -8.82439733e-01 4.19911325e-01 2.36505225e-01 1.30737126e-01 -6.29854143e-01 6.51624203e-01 -1.55356470e-02 -1.48286298e-01 6.97662383e-02 -8.28632772e-01 6.40361756e-02 5.02052248e-01 5.94492018e-01 -2.56988049e-01 -2.78227806e-01 2.18648553e-01 -3.40291142e-01 -6.29058719e-01 1.32809803e-01 -1.82073146e-01 6.09505355e-01 3.92767459e-01 8.09605569e-02 -8.46985459e-01 -7.12087631e-01 -5.22207975e-01 4.99143094e-01 9.07270834e-02 4.63883251e-01 3.55183601e-01 -1.12059939e+00 -9.87997353e-01 -2.37818867e-01 6.62400424e-02 9.94433984e-02 4.59500015e-01 5.79731286e-01 -4.80965793e-01 8.93553555e-01 2.88062125e-01 -3.64683598e-01 -8.27940643e-01 2.40818650e-01 2.36977078e-02 -9.71907258e-01 -3.86088639e-01 1.22625411e+00 -1.17780909e-01 -7.51043439e-01 1.49305582e-01 -4.15606081e-01 8.11631754e-02 9.07637924e-02 6.06492281e-01 4.23170775e-01 1.50869682e-01 -7.52956718e-02 -2.98461947e-03 3.42873223e-02 -4.61866409e-01 -6.50689304e-01 9.53356862e-01 1.49495274e-01 4.28325534e-02 1.64885014e-01 1.07203352e+00 -1.66382983e-01 -9.89617884e-01 -5.39447963e-01 2.22379014e-01 -2.08608299e-01 -1.82912439e-01 -1.38393664e+00 -4.50896800e-01 6.04765594e-01 2.15037540e-01 1.53801918e-01 9.07298028e-01 2.68570632e-02 1.44911826e+00 7.52636015e-01 4.49569821e-01 -9.16952372e-01 1.70721740e-01 1.04685020e+00 7.78312743e-01 -1.09071231e+00 -2.29906172e-01 -1.52768299e-01 -7.13803351e-01 6.90476120e-01 4.73447055e-01 -2.89937947e-02 1.86249778e-01 1.03011750e-01 4.17261273e-02 4.53107841e-02 -1.20961976e+00 -3.84048820e-01 4.32807267e-01 3.09948772e-01 5.47246754e-01 8.75745341e-02 -1.78575039e-01 5.87020874e-01 -5.98668575e-01 2.75732398e-01 1.37346596e-01 9.04385030e-01 -6.04923666e-01 -1.21684098e+00 -1.01211525e-01 6.57413840e-01 -5.59402823e-01 -7.90054977e-01 -1.96078822e-01 5.90187788e-01 -3.40139061e-01 1.07255316e+00 3.14027295e-02 -3.03403676e-01 2.78441459e-01 4.61428046e-01 2.59773284e-01 -9.07333076e-01 -7.91349113e-01 -1.40422106e-01 6.78403676e-01 -4.80081886e-01 5.23709990e-02 -4.56397355e-01 -1.23302257e+00 -1.98928967e-01 -4.27568108e-01 5.37468016e-01 4.17573869e-01 6.52778864e-01 7.91064620e-01 4.35024709e-01 -7.98899010e-02 -6.21226609e-01 -7.75110841e-01 -1.38406992e+00 -7.50536472e-02 2.18430325e-01 9.29988250e-02 -4.07762647e-01 -6.22163787e-02 -5.74364439e-02]
[11.453539848327637, 8.029667854309082]
3ae86604-c353-4289-a5b5-331c218ec6c7
generate-segment-and-replace-towards-generic
1811.09729
null
https://arxiv.org/abs/1811.09729v3
https://arxiv.org/pdf/1811.09729v3.pdf
Generate, Segment and Refine: Towards Generic Manipulation Segmentation
Detecting manipulated images has become a significant emerging challenge. The advent of image sharing platforms and the easy availability of advanced photo editing software have resulted in a large quantities of manipulated images being shared on the internet. While the intent behind such manipulations varies widely, concerns on the spread of fake news and misinformation is growing. Current state of the art methods for detecting these manipulated images suffers from the lack of training data due to the laborious labeling process. We address this problem in this paper, for which we introduce a manipulated image generation process that creates true positives using currently available datasets. Drawing from traditional work on image blending, we propose a novel generator for creating such examples. In addition, we also propose to further create examples that force the algorithm to focus on boundary artifacts during training. Strong experimental results validate our proposal.
['Mahyar Najibi', 'Bor-Chun Chen', 'Peng Zhou', 'Ser Nam Lim', 'Xintong Han', 'Larry S. Davis', 'Abhinav Shrivastava']
2018-11-24
null
null
null
null
['image-manipulation-detection', 'detecting-image-manipulation']
['computer-vision', 'computer-vision']
[ 7.58914173e-01 9.80249569e-02 -1.97024912e-01 -1.79004848e-01 -6.48325443e-01 -7.41000295e-01 7.67830789e-01 1.33835211e-01 -1.87843055e-01 6.48899674e-01 1.71496589e-02 -1.87427551e-01 3.59394699e-01 -5.85684955e-01 -8.78635108e-01 -2.42973998e-01 1.77362978e-01 2.36971416e-02 3.50218803e-01 -1.59171000e-01 5.69291770e-01 4.14122522e-01 -1.59555316e+00 3.99307907e-01 8.40846002e-01 6.48215234e-01 1.00687198e-01 5.34398258e-01 -3.38144116e-02 8.15006852e-01 -1.04131734e+00 -8.64256680e-01 5.86681187e-01 -5.39276779e-01 -6.39159739e-01 6.79986537e-01 8.83192778e-01 -8.11181068e-01 -1.90346047e-01 1.42759025e+00 3.16864461e-01 -1.03872865e-01 9.88741890e-02 -1.91195071e+00 -9.43448126e-01 5.49437404e-01 -8.91325057e-01 2.78318107e-01 3.72549057e-01 3.41189951e-01 5.21486402e-01 -8.15925360e-01 1.12443805e+00 1.13113904e+00 6.81427479e-01 4.80202913e-01 -1.10474837e+00 -7.41330743e-01 -1.03102423e-01 1.26669770e-02 -1.30407310e+00 -4.09959704e-01 9.11195338e-01 -4.75251257e-01 5.16002238e-01 4.03473735e-01 6.52324140e-01 1.29402363e+00 -2.67280340e-01 7.46527135e-01 1.36269379e+00 -7.01490641e-01 -1.94312930e-02 6.00194454e-01 -3.15648615e-01 7.54871607e-01 6.26571357e-01 4.95161526e-02 -3.57414454e-01 -2.58141637e-01 8.58663857e-01 -1.55689687e-01 -3.02067369e-01 -3.20258737e-01 -9.47880626e-01 6.18306279e-01 3.01774174e-01 2.54718482e-01 -1.87014058e-01 9.42728072e-02 2.83369213e-01 5.86678870e-02 5.22523820e-01 8.06593299e-01 7.24430010e-03 -1.39429912e-01 -1.02060521e+00 4.18187261e-01 5.91948748e-01 1.04003227e+00 8.17617714e-01 -1.41881213e-01 9.51550156e-02 6.57032192e-01 -7.26930723e-02 2.84125239e-01 2.72328228e-01 -1.09482586e+00 4.96369153e-01 6.61546946e-01 3.96315128e-01 -1.71325517e+00 1.41022116e-01 -7.95231536e-02 -4.75215942e-01 3.02985281e-01 4.60348010e-01 8.79071578e-02 -9.06115890e-01 1.31029201e+00 4.09837306e-01 2.66158760e-01 -4.47176397e-01 8.72689366e-01 1.70152679e-01 3.20246994e-01 3.95279117e-02 1.88157950e-02 1.22802174e+00 -1.03744030e+00 -7.50935853e-01 -2.31569618e-01 5.06472409e-01 -1.02414238e+00 1.13727915e+00 6.09578609e-01 -1.02108252e+00 -2.94256508e-01 -1.06899095e+00 -1.42495871e-01 -5.26302934e-01 5.63153364e-02 6.34350240e-01 1.09472084e+00 -8.51489067e-01 5.04673243e-01 -4.18748021e-01 -4.41043556e-01 8.05185020e-01 -8.17850307e-02 -3.90493602e-01 -1.14259973e-01 -7.57611334e-01 1.03311896e+00 3.20476472e-01 -6.00715391e-02 -6.13703966e-01 -5.35544813e-01 -5.98073125e-01 -3.80166292e-01 6.18992269e-01 -3.16800326e-01 1.17053556e+00 -1.42652917e+00 -1.23056841e+00 1.13075387e+00 1.77625120e-01 -2.80067712e-01 9.76430893e-01 -2.66687989e-01 -3.93098205e-01 3.58442217e-01 3.77788395e-01 8.63764703e-01 1.29240930e+00 -1.78517270e+00 -6.19505286e-01 -1.17933422e-01 2.10636809e-01 -9.67924073e-02 -5.61166763e-01 1.79705516e-01 -4.04029667e-01 -9.24907863e-01 -1.53781950e-01 -9.48427796e-01 -1.62698463e-01 2.18027174e-01 -5.49630761e-01 2.28862837e-01 1.15089190e+00 -7.48924792e-01 1.14126801e+00 -2.17160940e+00 -4.88021940e-01 2.62012541e-01 2.56225139e-01 5.23201942e-01 -1.41340896e-01 4.41205055e-01 4.20934446e-02 6.96020782e-01 -2.60864407e-01 -3.81894410e-01 -2.50235647e-01 1.13057591e-01 -6.28147304e-01 4.42669123e-01 3.12965244e-01 7.28584588e-01 -1.18755412e+00 -6.43211365e-01 2.72890389e-01 2.78749943e-01 -4.05755609e-01 -2.58415565e-02 -2.75337994e-01 3.55046004e-01 -1.06949307e-01 8.44077468e-01 8.60595703e-01 -4.24051225e-01 3.96559797e-02 -1.15357369e-01 -1.31044894e-01 1.66796848e-01 -1.08719325e+00 1.36812556e+00 -5.51669225e-02 6.43611193e-01 -4.36481275e-02 -5.81643105e-01 6.00404024e-01 2.97142919e-02 2.20829964e-01 -4.31982845e-01 2.29602009e-01 2.01131016e-01 -2.01353133e-01 -8.02470803e-01 9.73160446e-01 9.15658399e-02 1.56718552e-01 7.11694658e-01 -3.69837821e-01 -3.43277693e-01 3.58970582e-01 4.42982286e-01 1.01574504e+00 4.00500864e-01 2.73411065e-01 3.40957403e-01 1.34462029e-01 4.88113403e-01 2.27630153e-01 8.75884473e-01 -4.35500026e-01 7.45588243e-01 2.16248721e-01 -4.11711544e-01 -1.35546565e+00 -5.99712849e-01 1.05853803e-01 6.46532118e-01 4.45934862e-01 -4.12116170e-01 -8.20297241e-01 -7.32877374e-01 -4.03323360e-02 6.74768269e-01 -3.69061500e-01 3.77742201e-02 -5.10399759e-01 -4.51712400e-01 7.93163955e-01 2.17148796e-01 7.10066736e-01 -1.00455058e+00 -8.25797498e-01 6.64219707e-02 -4.61189687e-01 -1.52541578e+00 -4.14586157e-01 -7.00933218e-01 -5.89946747e-01 -1.20546496e+00 -4.75262165e-01 -6.78741753e-01 1.03810644e+00 8.80441546e-01 9.04536963e-01 5.46442509e-01 -6.97696805e-01 3.40321183e-01 -4.36063319e-01 -4.13443357e-01 -8.47438395e-01 -3.14375132e-01 -3.05230469e-01 1.83726981e-01 1.46207482e-01 -3.06322634e-01 -5.24357677e-01 3.43426913e-01 -1.31263292e+00 3.23134214e-01 4.81276542e-01 5.31191468e-01 1.68138906e-01 2.79545188e-01 4.22095090e-01 -1.15534151e+00 6.85952127e-01 -4.51658666e-01 -6.46167338e-01 2.79082894e-01 -5.46191394e-01 -2.96210796e-01 2.32285455e-01 -7.03102469e-01 -1.08978879e+00 5.90657108e-02 3.63155067e-01 -4.97020453e-01 -2.78568715e-01 1.24883637e-01 2.75861710e-01 -4.41847831e-01 8.63945603e-01 -7.24757686e-02 -7.46402005e-03 -2.60767281e-01 5.47234654e-01 6.71095848e-01 4.57016259e-01 -2.84406245e-01 1.12522876e+00 8.16869736e-01 -3.18713039e-01 -9.29831743e-01 -6.99151397e-01 -2.51747727e-01 -2.05859229e-01 -5.11137605e-01 6.11940324e-01 -6.51245356e-01 -3.87391984e-01 4.95989889e-01 -1.37131107e+00 -1.29317746e-01 -1.58818141e-01 7.60583729e-02 -2.66565859e-01 8.52353096e-01 -5.01273692e-01 -9.25692856e-01 1.73710138e-02 -1.04738045e+00 9.83065307e-01 -4.61688302e-02 -3.64529461e-01 -5.19679666e-01 -2.67513245e-01 7.91956306e-01 4.94456857e-01 5.76006234e-01 4.04551744e-01 -1.32424712e-01 -1.18751943e+00 -4.59818065e-01 -5.27607918e-01 5.18533051e-01 2.50771046e-01 1.86572775e-01 -8.48937154e-01 1.19310888e-02 7.78708886e-03 -4.25083131e-01 4.71013904e-01 -1.54281020e-01 1.24807429e+00 -5.35088122e-01 -4.68948781e-01 1.75258592e-01 1.42150450e+00 -1.01830430e-01 8.38065982e-01 5.52813947e-01 7.67634749e-01 7.79497385e-01 6.63308918e-01 3.94945920e-01 1.98691651e-01 5.63942909e-01 4.81760889e-01 -1.53158866e-02 -3.26053858e-01 -5.21954536e-01 1.53311238e-01 3.28466266e-01 2.87751146e-02 -3.74336928e-01 -8.96644652e-01 5.48323870e-01 -1.72419095e+00 -1.24794686e+00 -3.14825147e-01 2.22199941e+00 7.59556234e-01 9.44838077e-02 3.10426624e-03 2.36804157e-01 1.06121755e+00 1.26216576e-01 -8.05710554e-02 -7.84576312e-02 -4.78013195e-02 1.22619793e-01 7.92414606e-01 3.64750057e-01 -1.18100250e+00 9.63307858e-01 6.49737358e+00 6.15894556e-01 -1.19823408e+00 1.57358989e-01 7.90935934e-01 1.96590289e-01 -2.29210317e-01 2.24860638e-01 -6.40909195e-01 6.26762450e-01 2.18538523e-01 -8.51684138e-02 5.36949694e-01 8.08502197e-01 1.81725204e-01 -4.82853562e-01 -7.16168225e-01 1.15686595e+00 7.27669299e-01 -1.40548635e+00 1.52486369e-01 -1.12352911e-02 9.46401238e-01 -3.09004158e-01 6.94378465e-02 -1.66287735e-01 2.96255916e-01 -7.23664522e-01 8.55775893e-01 2.26577390e-02 6.32514417e-01 -2.31409088e-01 2.27399468e-01 3.65697592e-01 -6.40039086e-01 -3.02010793e-02 -1.03463419e-01 -1.87120721e-01 2.41408750e-01 6.72988236e-01 -1.27771950e+00 1.83810353e-01 4.43899900e-01 4.61817145e-01 -9.76766527e-01 1.19268405e+00 -5.00364482e-01 4.41231251e-01 -3.29420060e-01 9.02305022e-02 -6.16148226e-02 -1.98348746e-01 4.57668573e-01 9.18848097e-01 3.86979848e-01 -2.68945605e-01 1.95783079e-01 1.03073156e+00 -4.08573419e-01 -1.74837932e-02 -1.19483232e+00 -3.13890755e-01 5.19079983e-01 1.36293888e+00 -1.18033969e+00 -4.71524775e-01 -3.13509673e-01 1.44613183e+00 3.03961217e-01 2.38030270e-01 -1.04538310e+00 -1.00143747e-02 1.73356548e-01 4.25125778e-01 1.79123536e-01 -2.90341824e-01 -1.89644158e-01 -1.24125290e+00 3.20071369e-01 -1.19684875e+00 6.46476895e-02 -9.77102578e-01 -1.33409798e+00 3.04942012e-01 2.39215512e-02 -1.32950771e+00 5.88658452e-03 -2.23742440e-01 -2.67535716e-01 4.51499343e-01 -1.45787680e+00 -1.24877679e+00 -6.14167452e-01 3.06215256e-01 4.97160047e-01 3.35234970e-01 6.44155860e-01 5.23875535e-01 -3.98290724e-01 4.41783100e-01 -4.30233210e-01 1.81029603e-01 9.25014257e-01 -8.08012545e-01 6.77421749e-01 1.14962184e+00 2.30685592e-01 6.87620223e-01 7.91637182e-01 -9.19360638e-01 -1.29076886e+00 -9.72101033e-01 8.90772700e-01 -7.28165209e-01 8.05084586e-01 -4.18215096e-01 -7.11309314e-01 7.39963591e-01 2.30380207e-01 6.27918616e-02 4.65887278e-01 -4.31452334e-01 -7.13882625e-01 2.72860825e-01 -1.37455571e+00 7.49489784e-01 1.15201211e+00 -5.83584487e-01 -3.93674821e-01 5.95811069e-01 2.19417408e-01 -3.61467093e-01 -2.75623828e-01 1.71693414e-01 5.10177314e-01 -1.05288923e+00 8.35651934e-01 -3.85463566e-01 6.58151865e-01 -4.88375306e-01 2.23513812e-01 -1.08732259e+00 1.32423162e-01 -9.10934746e-01 1.42885774e-01 1.30363297e+00 9.30585116e-02 -6.22420251e-01 8.67307544e-01 8.94497216e-01 1.22940272e-01 -2.12190762e-01 -4.71466064e-01 -7.33780861e-01 -6.09281480e-01 -2.55304188e-01 3.03731173e-01 1.45053136e+00 -6.60350695e-02 1.01550199e-01 -6.18646324e-01 1.82448924e-01 6.96043015e-01 -8.60618651e-02 1.21434951e+00 -8.38587046e-01 -4.26278412e-01 -3.57967138e-01 -5.64903736e-01 -9.57956374e-01 -2.66384125e-01 -5.71855366e-01 -5.14036939e-02 -1.24388337e+00 2.72502840e-01 -6.40809298e-01 3.57439876e-01 3.01301390e-01 -2.56652772e-01 8.43291700e-01 3.40075642e-01 3.01044196e-01 -4.20019299e-01 -2.23352108e-03 1.23625135e+00 -1.02885924e-02 6.07622787e-02 -5.17531097e-01 -6.29949272e-01 7.41031468e-01 9.98510718e-01 -5.61094820e-01 -3.97116214e-01 -3.84188801e-01 4.29774165e-01 -4.20338660e-01 6.76724851e-01 -1.07069850e+00 -1.58195496e-01 -6.44481778e-02 2.21559882e-01 -3.20796400e-01 2.80511558e-01 -8.86707842e-01 2.64524221e-01 3.52051020e-01 -1.98217452e-01 1.06501929e-01 5.13820127e-02 6.61659420e-01 -1.05796903e-01 -2.44246185e-01 8.23427677e-01 -2.75485605e-01 -6.70712471e-01 -3.42258066e-02 -1.60212696e-01 2.67157666e-02 1.43493450e+00 -4.72302139e-01 -8.15412939e-01 -4.47371960e-01 -2.43789300e-01 -2.19507948e-01 9.89237845e-01 6.39758468e-01 5.83875477e-01 -1.07651389e+00 -2.74279892e-01 1.60931870e-01 1.44050807e-01 -2.94115394e-01 -6.58290321e-03 6.30750120e-01 -8.82976711e-01 -3.16962838e-01 -2.69107819e-01 -3.43696386e-01 -1.40700054e+00 7.30041027e-01 -5.68316355e-02 8.74752998e-02 -6.84806347e-01 7.07977295e-01 -1.64657116e-01 -5.67947961e-02 3.57597657e-02 -6.65540323e-02 3.38869393e-01 -8.12295154e-02 6.99383795e-01 5.48721850e-01 -6.32033944e-02 -5.96907079e-01 5.91200357e-03 -5.79329506e-02 -2.31218621e-01 -1.10861689e-01 1.13334370e+00 -1.98101535e-01 -2.13101476e-01 1.30209476e-01 9.96416688e-01 2.32757062e-01 -1.05838609e+00 1.73759744e-01 1.98761940e-01 -1.18745649e+00 -1.56251773e-01 -7.33293414e-01 -8.88603806e-01 4.38068271e-01 4.46919143e-01 6.01403236e-01 8.44345927e-01 -1.53633595e-01 1.09169114e+00 9.72285643e-02 6.85192704e-01 -1.15424061e+00 4.04128551e-01 3.30890045e-02 8.01239491e-01 -1.52812135e+00 4.29155469e-01 -9.20836270e-01 -5.75828373e-01 8.55187237e-01 5.98609865e-01 -1.42453492e-01 2.70758718e-01 2.24414572e-01 3.19212943e-01 -2.42488474e-01 -1.20178133e-01 4.15414087e-02 -3.49643588e-01 6.01055145e-01 2.25331277e-01 -1.24778263e-01 -5.09576976e-01 -2.74085134e-01 -1.32777467e-01 2.80473500e-01 9.89909768e-01 1.38520837e+00 -1.83330223e-01 -1.36532283e+00 -6.76554084e-01 4.70611662e-01 -5.91340661e-01 -1.17032183e-02 -7.09808528e-01 8.50265980e-01 3.43509316e-01 9.62363243e-01 -1.38475701e-01 -1.55130982e-01 -3.62526737e-02 -5.78524955e-02 6.49133742e-01 -6.16795182e-01 -5.69553614e-01 -3.08676213e-01 2.14953661e-01 -6.11238420e-01 -7.28039086e-01 -5.60657918e-01 -7.26842344e-01 -2.76139289e-01 -6.99239075e-01 -3.41270775e-01 9.90265250e-01 5.25259495e-01 6.06594503e-01 -4.35980596e-02 4.60289508e-01 -1.04524612e+00 -6.31126106e-01 -8.08930933e-01 -2.00676292e-01 1.18981111e+00 1.61059737e-01 -7.44144201e-01 -4.68182504e-01 5.60925961e-01]
[12.394346237182617, 1.0448423624038696]
94fff1c7-5d5a-421e-9326-776d6934bbce
a-glimpse-of-the-whole-path-optimization
2010.13285
null
https://arxiv.org/abs/2010.13285v4
https://arxiv.org/pdf/2010.13285v4.pdf
Malicious Requests Detection with Improved Bidirectional Long Short-term Memory Neural Networks
Detecting and intercepting malicious requests are one of the most widely used ways against attacks in the network security. Most existing detecting approaches, including matching blacklist characters and machine learning algorithms have all shown to be vulnerable to sophisticated attacks. To address the above issues, a more general and rigorous detection method is required. In this paper, we formulate the problem of detecting malicious requests as a temporal sequence classification problem, and propose a novel deep learning model namely Convolutional Neural Network-Bidirectional Long Short-term Memory-Convolutional Neural Network (CNN-BiLSTM-CNN). By connecting the shadow and deep feature maps of the convolutional layers, the malicious feature extracting ability is improved on more detailed functionality. Experimental results on HTTP dataset CSIC 2010 have demonstrated the effectiveness of the proposed method when compared with the state-of-the-arts.
['Jiajie Zhang', 'Bincheng Zhang', 'Wenhao Li']
2020-10-26
null
null
null
null
['traffic-classification']
['miscellaneous']
[ 1.36761926e-02 -8.87818098e-01 -2.33514994e-01 -2.94676483e-01 -4.62436497e-01 -5.99106967e-01 7.43223965e-01 -1.85735270e-01 -4.36493844e-01 7.78730437e-02 -1.44654319e-01 -9.23497856e-01 2.88394928e-01 -6.34226978e-01 -5.07336438e-01 -6.48514628e-01 -3.55548799e-01 -5.56858769e-03 8.47818375e-01 -4.56540763e-01 4.74300891e-01 6.87923610e-01 -7.80115426e-01 8.50709856e-01 2.32908830e-01 1.30547822e+00 -4.56194669e-01 8.42728198e-01 -3.05659294e-01 1.05580378e+00 -6.50408447e-01 -6.52848423e-01 3.79981488e-01 5.38366623e-02 -5.84567130e-01 -3.45838696e-01 1.36716496e-02 -8.51407886e-01 -8.39922547e-01 1.04839027e+00 4.56259131e-01 -2.30781689e-01 1.29734263e-01 -1.24239695e+00 -5.95471919e-01 5.57356298e-01 -6.76082850e-01 6.67653143e-01 3.55541468e-01 3.59594703e-01 7.92399943e-01 -6.19702101e-01 7.79320449e-02 1.20590520e+00 7.13892162e-01 6.32985055e-01 -6.37094319e-01 -9.71968770e-01 3.83811384e-01 6.01790071e-01 -8.91662121e-01 -1.10765494e-01 7.58321285e-01 -2.52559304e-01 1.14515960e+00 1.01800159e-01 1.75415069e-01 1.54210246e+00 1.83210239e-01 9.67100739e-01 7.45425582e-01 -1.98628947e-01 -2.41588876e-01 6.86362991e-03 4.58918363e-01 7.44861305e-01 1.15638502e-01 4.29846048e-01 2.50863563e-02 -6.58729076e-01 3.35821271e-01 6.50627136e-01 -1.17021747e-01 4.94540557e-02 -7.88780510e-01 1.10500419e+00 4.75812972e-01 6.08036518e-01 -2.56783605e-01 2.93432385e-01 1.09799659e+00 7.04537809e-01 5.15343726e-01 -2.12411024e-02 -7.12558448e-01 -1.28574550e-01 -9.50153828e-01 -8.33286196e-02 9.60982919e-01 4.16349471e-01 1.27908811e-01 5.41740537e-01 5.78168966e-02 4.01103377e-01 4.76132691e-01 4.85106498e-01 6.20556295e-01 -9.94858369e-02 4.50977772e-01 4.65850413e-01 -2.12954119e-01 -1.36738312e+00 -3.60989720e-01 -3.33510488e-01 -6.56948566e-01 3.26858722e-02 2.91921288e-01 -1.94780514e-01 -8.28893840e-01 1.32796907e+00 1.90297827e-01 6.63399637e-01 -1.31191462e-01 6.37133360e-01 4.14609104e-01 7.22305655e-01 1.69037551e-01 8.29301216e-03 1.21947932e+00 -7.94529200e-01 -5.17369449e-01 1.73623487e-01 6.89329684e-01 -7.93034732e-01 4.83649611e-01 2.30585903e-01 -4.44941014e-01 -4.20772314e-01 -9.90643620e-01 4.52541113e-01 -8.46322000e-01 -3.78553003e-01 4.56278324e-01 1.03024197e+00 -7.63579845e-01 4.82850879e-01 -7.47712910e-01 -4.13697481e-01 3.09899122e-01 5.06624043e-01 -2.72267405e-02 3.77082556e-01 -1.62804997e+00 6.01046622e-01 3.56212944e-01 2.89090365e-01 -1.17852485e+00 -3.65677923e-01 -4.84466195e-01 1.20658807e-01 2.82966524e-01 2.60325134e-01 1.52199602e+00 -9.36052740e-01 -1.33198130e+00 5.99487603e-01 2.23584071e-01 -8.61466587e-01 4.90430981e-01 -2.67847091e-01 -8.77238750e-01 2.27700248e-01 -3.58567953e-01 -2.69065015e-02 1.24115074e+00 -1.03292775e+00 -6.73212469e-01 -2.85315782e-01 4.04091589e-02 -8.18511069e-01 -7.99999356e-01 8.74847233e-01 -2.34646216e-01 -6.59282982e-01 -3.18361342e-01 -8.48427296e-01 -1.29754499e-01 -3.04695994e-01 -5.41884661e-01 -3.71719003e-01 1.71437275e+00 -1.00493193e+00 1.37482703e+00 -1.94503641e+00 -5.98454952e-01 5.31990528e-01 3.68364960e-01 1.17923200e+00 -2.45420709e-01 5.18982828e-01 -3.06512475e-01 3.09785873e-01 3.32327813e-01 1.18589275e-01 2.99417134e-02 -3.30399536e-02 -8.59788120e-01 6.89172149e-01 1.13137074e-01 9.84348536e-01 -6.21632755e-01 -2.29580060e-01 2.73067534e-01 6.07522726e-01 -4.87946421e-02 2.71626204e-01 -2.12820679e-01 1.03953652e-01 -7.55306005e-01 7.58436799e-01 7.78588891e-01 -3.02666545e-01 1.07455447e-01 1.33711332e-02 1.59755126e-01 3.02851170e-01 -6.28986478e-01 1.01064301e+00 -2.87163436e-01 6.95907295e-01 6.22638799e-02 -1.10362649e+00 7.82057405e-01 6.05905473e-01 5.55832088e-01 -8.52927625e-01 5.37352204e-01 3.35427493e-01 2.40247309e-01 -6.94674611e-01 -8.59676525e-02 2.18026310e-01 -1.50673673e-01 8.13220501e-01 -3.05278748e-01 6.86017811e-01 -1.88362539e-01 3.94101441e-01 1.30028880e+00 -3.65215838e-01 -5.73134944e-02 3.41616303e-01 1.03299940e+00 -2.72695601e-01 3.81122619e-01 9.24871624e-01 -7.84277380e-01 -1.03041224e-01 4.85064417e-01 -8.72307479e-01 -1.02164030e+00 -6.75131738e-01 1.41947076e-01 1.29767942e+00 -1.53992102e-01 -1.99987277e-01 -6.94391370e-01 -1.21629417e+00 1.05771489e-01 3.00959021e-01 -5.19750297e-01 -2.71113247e-01 -1.11613131e+00 -5.39876163e-01 1.15776479e+00 5.69785774e-01 7.16168582e-01 -1.21411169e+00 -3.30641240e-01 4.28639233e-01 9.69359875e-02 -1.53879023e+00 -4.57938135e-01 3.95220295e-02 -4.98782516e-01 -1.23823810e+00 -4.17083323e-01 -9.31884885e-01 1.70334637e-01 5.78002036e-01 7.19327092e-01 7.48923719e-01 -4.97077584e-01 1.58158794e-01 -6.21931851e-01 -1.54849052e-01 -4.38256979e-01 1.95372894e-01 1.15157589e-01 2.59877950e-01 8.29592943e-01 -7.29471445e-01 -5.54270983e-01 2.15311989e-01 -9.86770272e-01 -5.37834048e-01 7.47591197e-01 6.76653266e-01 -2.30285987e-01 -8.37594792e-02 7.02173591e-01 -1.02273762e+00 7.33433008e-01 -6.41689777e-01 -6.07121289e-01 2.89595693e-01 -5.44267595e-01 -7.12105110e-02 1.07247365e+00 -7.18231857e-01 -7.12562740e-01 -1.62111491e-01 -4.35855001e-01 -4.95725483e-01 -1.97736025e-01 1.39737859e-01 -3.26285176e-02 -6.56921268e-01 3.41703027e-01 7.50180542e-01 -1.43639073e-01 -4.23833966e-01 3.16984713e-01 9.65581715e-01 3.13106120e-01 -4.74423617e-01 1.39250505e+00 5.33210635e-01 -3.23220074e-01 -5.15223980e-01 -6.20467007e-01 -6.73656225e-01 -4.51799095e-01 -2.60006130e-01 7.76138008e-01 -3.22490275e-01 -1.27981615e+00 9.36848819e-01 -1.37262285e+00 -6.44111354e-03 5.30942261e-01 8.48596618e-02 -1.38744945e-02 9.94272232e-01 -1.46435034e+00 -8.28209102e-01 -7.87129939e-01 -1.20087826e+00 7.19970822e-01 -4.73796241e-02 3.02227288e-01 -1.17747998e+00 1.54784620e-01 2.00229272e-01 8.84264469e-01 2.44508162e-01 1.02620876e+00 -1.56341922e+00 -4.72231954e-01 -8.94387305e-01 -5.97471356e-01 5.56556344e-01 -1.33566275e-01 3.37663323e-01 -8.95265877e-01 -5.04752219e-01 6.25720993e-02 -3.08547050e-01 9.62218881e-01 -2.30635673e-01 1.35803318e+00 -6.03058994e-01 -3.26606154e-01 5.78089178e-01 1.38439035e+00 5.68933427e-01 6.71181858e-01 6.50617778e-01 7.11558044e-01 2.24563375e-01 9.74046737e-02 5.78441143e-01 -5.82866557e-03 4.75407600e-01 8.05127680e-01 -1.54763699e-01 3.84363443e-01 -2.53386926e-02 6.89123213e-01 8.46159339e-01 4.28337544e-01 -3.03692430e-01 -8.95037711e-01 1.18454851e-01 -1.87757158e+00 -1.35933363e+00 -3.38098496e-01 1.63924146e+00 3.45880032e-01 3.55075508e-01 2.05643535e-01 3.86911869e-01 8.60939682e-01 6.28585279e-01 -3.14814359e-01 -8.08819890e-01 3.69068123e-02 3.29521120e-01 6.07567012e-01 2.66240507e-01 -1.54074514e+00 9.71707880e-01 6.03113031e+00 1.05884385e+00 -1.40023029e+00 1.77537233e-01 5.48945904e-01 1.86753586e-01 1.70459226e-01 -4.22372892e-02 -8.40241969e-01 7.90992379e-01 1.36907864e+00 2.37990141e-01 3.00665200e-01 8.65326047e-01 3.48343641e-01 6.21320248e-01 -6.12915218e-01 7.03282356e-01 1.27818257e-01 -1.14420331e+00 2.47098297e-01 5.53819304e-03 4.29504305e-01 1.06119603e-01 5.17197847e-01 7.49185383e-01 5.38768470e-01 -8.35765541e-01 4.35763896e-01 1.36732683e-01 4.17629898e-01 -6.96338832e-01 1.05561054e+00 3.00009757e-01 -1.23694646e+00 -4.27523613e-01 -2.88969100e-01 2.19051227e-01 6.44367412e-02 3.90268624e-01 -7.21417010e-01 4.30389255e-01 6.93146944e-01 6.65956676e-01 -6.62373185e-01 8.93450260e-01 -7.51085505e-02 9.75955069e-01 -7.21314922e-02 -3.25020015e-01 7.25999355e-01 2.37056896e-01 3.27058822e-01 1.55626142e+00 -8.45438167e-02 -1.41263783e-01 5.94968617e-01 3.20274353e-01 -1.71055526e-01 3.63994278e-02 -5.08299649e-01 -3.76086026e-01 6.80585429e-02 1.36320174e+00 -4.18213159e-01 -4.18892831e-01 -7.28219867e-01 1.09625423e+00 1.17305629e-01 4.23462838e-01 -1.25407732e+00 -5.63621104e-01 6.70474470e-01 3.53714488e-02 4.26815271e-01 -3.83149862e-01 -7.01452270e-02 -1.20890069e+00 3.33747379e-02 -1.05857766e+00 6.12109005e-01 -2.03568235e-01 -1.56830943e+00 7.88221240e-01 -5.31180024e-01 -1.23439038e+00 -1.72915727e-01 -9.60534871e-01 -1.03608239e+00 6.55591071e-01 -1.93940389e+00 -1.28504050e+00 -9.46290717e-02 1.03912449e+00 5.82133591e-01 -6.11179769e-01 6.05740130e-01 6.16966367e-01 -8.38265359e-01 7.63721347e-01 2.52681166e-01 1.02148414e+00 4.03279722e-01 -6.94168508e-01 7.39800572e-01 1.22142744e+00 -1.31754309e-01 7.27956355e-01 3.59109819e-01 -7.49236047e-01 -1.40241706e+00 -1.14142489e+00 5.09887874e-01 -7.69409835e-02 1.26001441e+00 -4.43470091e-01 -1.05138052e+00 6.97371423e-01 2.46152118e-01 1.55749545e-01 7.04188585e-01 -5.78546822e-01 -1.17779362e+00 -1.34497032e-01 -1.11568606e+00 3.19394350e-01 5.23177803e-01 -9.11241472e-01 -1.37792140e-01 4.28199947e-01 9.05188620e-01 7.87457302e-02 -5.18650174e-01 2.93424845e-01 6.41054153e-01 -9.37944591e-01 1.07594800e+00 -1.22817206e+00 -2.05249235e-01 -5.56535739e-03 -1.55300766e-01 -6.46733940e-01 -3.01737130e-01 -7.24006712e-01 -5.39923310e-01 1.05513906e+00 3.27090174e-01 -8.19436312e-01 1.03090370e+00 1.26430526e-01 2.76767462e-01 -6.38552666e-01 -8.36446106e-01 -7.81373084e-01 1.00487880e-01 -5.95355093e-01 4.27246243e-01 9.38657105e-01 -2.59456009e-01 -1.07363664e-01 -6.87599301e-01 2.89306253e-01 6.19611919e-01 -1.07159257e-01 5.24560571e-01 -1.06736696e+00 -2.15610087e-01 -5.30471981e-01 -5.93985796e-01 -1.06476045e+00 4.45995569e-01 -6.73494101e-01 -3.06562334e-01 -1.01452303e+00 2.16929615e-01 -1.04666196e-01 -6.49126172e-01 6.02091491e-01 2.01562680e-02 8.31553619e-03 2.24114791e-01 3.58366907e-01 -8.68929923e-01 1.65337458e-01 6.72815144e-01 -4.26307023e-01 3.10201555e-01 3.49227577e-01 -3.14327538e-01 5.30124962e-01 1.12558365e+00 -6.00274384e-01 3.96317989e-02 -3.16569597e-01 -9.16424245e-02 -9.66707468e-02 3.30189347e-01 -7.69619823e-01 3.60459059e-01 -9.52477660e-03 4.14537489e-01 -6.81614459e-01 1.98779136e-01 -1.03864276e+00 -6.57209754e-01 1.17454970e+00 -4.68399614e-01 4.74085897e-01 1.75886806e-02 7.67053425e-01 -1.56216174e-01 3.11005209e-02 8.02756786e-01 -7.54540488e-02 -8.57077718e-01 7.04198658e-01 -7.08741963e-01 -2.26010308e-01 1.12747788e+00 -4.56232205e-02 -6.32266164e-01 -6.56060398e-01 -3.37839365e-01 7.29535818e-02 -1.22544438e-01 6.35511220e-01 9.46630359e-01 -1.14928377e+00 -6.61571622e-01 2.93308347e-01 -1.11010954e-01 -1.18669271e+00 2.70416792e-02 7.87156999e-01 -6.51412249e-01 8.31995964e-01 -6.47905767e-02 -3.70657682e-01 -1.40086401e+00 1.05097497e+00 4.93925333e-01 -6.84902072e-01 -4.13933367e-01 6.03161454e-01 -3.24107081e-01 -4.42541599e-01 6.06561303e-01 9.14023593e-02 -3.25014323e-01 -3.54395986e-01 8.17319930e-01 2.09424853e-01 8.78622457e-02 -6.71936750e-01 -5.12369335e-01 1.59060717e-01 -7.34207869e-01 2.11361870e-01 1.19877720e+00 2.58461237e-01 -2.47507066e-01 -1.47402197e-01 1.67463076e+00 -3.68431538e-01 -6.51141405e-01 -6.20442092e-01 4.90742326e-01 -4.80819523e-01 -1.31735578e-01 -5.08888900e-01 -1.46888816e+00 1.31438744e+00 8.55021000e-01 6.45669401e-01 8.88434887e-01 -6.26040697e-01 1.79206336e+00 5.43372214e-01 1.94662675e-01 -6.82281017e-01 5.34318209e-01 8.01403940e-01 1.53333008e-01 -1.20360947e+00 -6.17137909e-01 -2.11800084e-01 -1.16672747e-01 1.65610623e+00 8.21263969e-01 -4.09947306e-01 9.44027543e-01 4.18157369e-01 2.87437111e-01 -1.97269067e-01 -8.78996968e-01 -1.12955160e-01 -1.03177622e-01 4.99849439e-01 2.14994475e-01 -2.31449172e-01 -1.33483887e-01 3.87154818e-01 4.94103670e-01 -4.00541455e-01 3.17095786e-01 1.10969782e+00 -6.60139859e-01 -1.22854650e+00 -2.51986921e-01 3.92690688e-01 -1.12446392e+00 -3.22474837e-01 -4.22412634e-01 5.65615356e-01 -3.33631933e-01 1.15521777e+00 -2.73920804e-01 -1.05072629e+00 3.55246328e-02 3.22639011e-02 -2.61195391e-01 -2.30337486e-01 -1.18665171e+00 -1.39420003e-01 -1.86778098e-01 -6.49854481e-01 -3.52014974e-02 -1.69557333e-01 -1.09979594e+00 -8.01528215e-01 -5.33409595e-01 2.10909694e-01 7.06368923e-01 8.85399163e-01 1.39181465e-01 3.79263133e-01 1.26740718e+00 -5.25158703e-01 -1.16212475e+00 -7.71668315e-01 -2.94363976e-01 5.50870240e-01 3.88039976e-01 -2.86500812e-01 -5.41082740e-01 -4.25112069e-01]
[5.170575141906738, 7.293119430541992]
886b6eb8-bc5e-4e1d-ac56-46fae3d36928
pixel-level-reconstruction-and-classification
1806.08037
null
http://arxiv.org/abs/1806.08037v1
http://arxiv.org/pdf/1806.08037v1.pdf
Pixel-level Reconstruction and Classification for Noisy Handwritten Bangla Characters
Classification techniques for images of handwritten characters are susceptible to noise. Quadtrees can be an efficient representation for learning from sparse features. In this paper, we improve the effectiveness of probabilistic quadtrees by using a pixel level classifier to extract the character pixels and remove noise from handwritten character images. The pixel level denoiser (a deep belief network) uses the map responses obtained from a pretrained CNN as features for reconstructing the characters eliminating noise. We experimentally demonstrate the effectiveness of our approach by reconstructing and classifying a noisy version of handwritten Bangla Numeral and Basic Character datasets.
['Supratik Mukhopadhyay', 'Qun Liu', 'Manohar Karki', 'Saikat Basu', 'Robert DiBiano']
2018-06-21
null
null
null
null
['document-image-classification']
['computer-vision']
[ 4.33475971e-01 -2.17734620e-01 2.39751577e-01 -5.05036533e-01 -6.88980103e-01 -4.97001767e-01 3.01714152e-01 -1.62443966e-01 -4.24295038e-01 7.92888880e-01 2.25081056e-01 -5.66470511e-02 -2.24955425e-01 -1.14107931e+00 -7.92911232e-01 -1.00036752e+00 3.34554195e-01 2.31358692e-01 4.47480977e-01 -3.68656428e-03 7.37005949e-01 1.08793962e+00 -1.49013937e+00 1.11596203e+00 5.37004948e-01 1.09368289e+00 3.16794574e-01 1.01939368e+00 -1.23032965e-01 1.17009318e+00 -9.74022627e-01 -2.96596557e-01 3.30188721e-01 -3.28381121e-01 -6.14551425e-01 4.01068360e-01 6.18619502e-01 -3.03901970e-01 -8.88990283e-01 1.37211764e+00 3.10121745e-01 -4.30068970e-02 8.96256685e-01 -5.33263922e-01 -7.70629227e-01 1.00837553e+00 -3.55683237e-01 1.03367142e-01 -1.71084851e-02 -1.39233157e-01 4.05163407e-01 -1.01993251e+00 3.87476563e-01 1.57034481e+00 1.26339424e+00 1.19315088e-01 -1.37409592e+00 -3.33788693e-01 -5.01424491e-01 4.12562996e-01 -1.40201283e+00 -2.57944614e-01 8.61061394e-01 -1.30089253e-01 1.08118248e+00 4.34162498e-01 3.87697250e-01 9.89500046e-01 7.41557658e-01 1.04663324e+00 1.50839055e+00 -8.07307839e-01 2.62891740e-01 4.28653434e-02 4.05464172e-01 8.24649453e-01 1.90707788e-01 5.76759987e-02 -4.71645862e-01 -2.21428618e-01 8.56317461e-01 -5.17464709e-03 1.99685507e-02 -4.80819605e-02 -8.52890790e-01 9.59819257e-01 6.13869369e-01 2.43575543e-01 -8.20754647e-01 3.75688255e-01 2.53794134e-01 1.36774987e-01 -7.74744824e-02 3.13173950e-01 -8.71680751e-02 3.40023860e-02 -1.34454632e+00 1.03723541e-01 9.02395964e-01 6.87331855e-01 7.91544020e-01 6.43156290e-01 -3.68305743e-01 1.00807405e+00 1.87415823e-01 5.47405720e-01 6.33316517e-01 -1.06670558e+00 3.71949561e-02 2.80958742e-01 -2.99789280e-01 -1.30447972e+00 4.33615595e-02 -1.16736725e-01 -1.13120401e+00 6.68327689e-01 1.07411943e-01 1.27157465e-01 -1.73616552e+00 7.57171988e-01 -3.09488595e-01 -2.00764969e-01 1.65937915e-01 5.93966424e-01 5.94409168e-01 1.21211100e+00 -1.79673865e-01 2.17349619e-01 1.07194054e+00 -7.04869449e-01 -9.01607692e-01 -3.76028456e-02 -3.12055767e-01 -1.00583458e+00 5.19115567e-01 1.11748350e+00 -7.79426932e-01 -6.63879275e-01 -1.08936191e+00 -1.23073913e-01 -3.84148061e-01 4.29147154e-01 5.58343828e-01 7.62272418e-01 -7.35527337e-01 1.02538621e+00 -1.03977120e+00 1.18049689e-01 8.12909245e-01 5.25720000e-01 -4.96132016e-01 -2.75404334e-01 -7.05237031e-01 1.07335985e+00 7.02149987e-01 4.28805947e-01 -1.04159701e+00 3.89864706e-02 -8.15286577e-01 1.96371302e-01 -5.21043479e-01 9.39354748e-02 8.75877321e-01 -1.07928908e+00 -1.62527943e+00 4.12819326e-01 7.59035647e-02 -6.35237217e-01 2.97855407e-01 5.58052473e-02 -1.92607954e-01 3.17946196e-01 -1.42988637e-01 6.69104040e-01 1.31600332e+00 -1.33007884e+00 -6.61335528e-01 -3.99772614e-01 -6.79016650e-01 -2.57258713e-01 -2.11321250e-01 -2.30402797e-01 -3.79344970e-02 -1.21003556e+00 9.52003598e-01 -4.45531636e-01 -1.97478130e-01 -2.29280844e-01 -6.64170325e-01 4.93026748e-02 1.30185878e+00 -1.08734179e+00 5.69490016e-01 -2.14794755e+00 -1.10760387e-02 6.03191018e-01 -1.60540923e-01 3.09880763e-01 -2.66536683e-01 1.51932821e-01 6.48469403e-02 -1.19305395e-01 -3.78520727e-01 -9.61440131e-02 -3.28087151e-01 5.94185472e-01 -5.87569892e-01 4.58876908e-01 5.58154345e-01 5.63963115e-01 -3.96412492e-01 -3.91029090e-01 4.41481501e-01 8.71913731e-01 -1.23981044e-01 -2.21269622e-01 2.52239332e-02 -3.40523511e-01 -2.55753428e-01 9.19441581e-01 1.06182158e+00 3.92856061e-01 -2.47845761e-02 -6.27077639e-01 9.64759514e-02 1.15483038e-01 -1.31195688e+00 1.16580844e+00 -1.69400483e-01 1.04462528e+00 -2.19464958e-01 -1.08699059e+00 1.40365624e+00 -1.04386032e-01 -7.02208206e-02 -4.80855465e-01 1.58547282e-01 8.52324367e-02 -2.06792429e-01 -3.18312764e-01 7.32306838e-01 -4.55440916e-02 -6.78689331e-02 6.45634681e-02 5.26470363e-01 -4.13811564e-01 -2.28088856e-01 -1.30978301e-01 1.15515697e+00 1.14583649e-01 6.04645945e-02 -2.55190492e-01 1.48125917e-01 3.78259838e-01 4.82278585e-01 1.29514527e+00 2.31901743e-02 9.99993086e-01 5.47412157e-01 -8.78286958e-01 -1.26973069e+00 -1.13035643e+00 -3.65063190e-01 5.28019667e-01 -2.62552112e-01 2.05150753e-01 -6.65071964e-01 -3.27796638e-01 1.47352248e-01 7.51330793e-01 -6.65326297e-01 6.49579093e-02 -6.96650982e-01 -9.14083004e-01 9.90220904e-01 5.94374835e-01 8.22775602e-01 -1.10524738e+00 -5.05369008e-01 5.29008329e-01 -2.20318407e-01 -9.65800583e-01 3.41942161e-01 1.15493476e+00 -1.08807540e+00 -5.97407758e-01 -5.43767631e-01 -1.13031948e+00 6.36795163e-01 -5.93105517e-02 6.70624435e-01 -3.27745765e-01 -5.51871717e-01 -2.52527386e-01 -4.61874187e-01 -2.84186780e-01 -5.79824567e-01 -2.14459568e-01 -1.95144743e-01 2.49990933e-02 4.10848081e-01 -5.12544513e-01 -2.59640217e-01 -4.49928716e-02 -8.85504723e-01 -3.95345926e-01 9.36648548e-01 1.21067691e+00 8.29796433e-01 7.25464702e-01 -9.52573214e-03 -9.18407679e-01 5.53346694e-01 -1.60067212e-02 -6.38096333e-01 -9.09901876e-03 -1.10821329e-01 2.20263779e-01 6.47755921e-01 -4.72888887e-01 -1.13077736e+00 8.21658313e-01 -2.78940290e-01 -3.11220855e-01 -3.07714731e-01 3.82484168e-01 8.14540908e-02 -4.79650348e-01 1.10359824e+00 6.85769320e-01 -3.39811891e-01 -5.67708254e-01 1.47960812e-01 1.06296718e+00 1.20290065e+00 -4.65958893e-01 5.36266267e-01 6.27788484e-01 5.97108155e-02 -1.04492044e+00 -3.72365564e-01 4.91993781e-03 -7.11506188e-01 -7.78426463e-03 5.97679496e-01 -7.89204836e-01 -3.24583679e-01 9.53828335e-01 -1.31982040e+00 -1.25023261e-01 -4.79301900e-01 3.88259828e-01 -5.09659648e-01 5.50293505e-01 -1.07114112e+00 -8.37743402e-01 -5.88707440e-02 -1.06424081e+00 8.98148060e-01 2.01280430e-01 2.12879166e-01 -5.37722111e-01 -1.57678589e-01 3.98349352e-02 2.49713287e-01 2.66039222e-01 1.05835772e+00 -3.36962610e-01 -6.00357533e-01 -6.67206943e-01 -1.74947575e-01 9.90594566e-01 1.14238344e-01 2.23974749e-01 -1.20879042e+00 1.32606134e-01 3.10911864e-01 -5.31005323e-01 1.57609344e+00 5.50025761e-01 1.37816358e+00 -4.04006153e-01 -4.30548303e-02 7.08100975e-01 1.76912820e+00 5.17659225e-02 1.33329642e+00 3.87921900e-01 7.79047191e-01 3.16920698e-01 2.08467245e-01 5.79227269e-01 -1.89577147e-01 1.43910751e-01 2.54808336e-01 2.03046083e-01 -4.34777915e-01 -2.65981942e-01 2.25302786e-01 6.32841885e-01 1.83087930e-01 -2.62974024e-01 -1.07540917e+00 6.25900269e-01 -1.52844906e+00 -8.13857555e-01 -2.32356369e-01 1.52252781e+00 1.14014006e+00 1.24077916e-01 -6.60978436e-01 6.92681134e-01 7.99724400e-01 -6.15695156e-02 -2.10205451e-01 -7.80070484e-01 -7.22220063e-01 5.55492222e-01 8.79451275e-01 5.67525387e-01 -1.36927605e+00 1.09203768e+00 7.34678650e+00 1.08144653e+00 -1.18337631e+00 -4.76643115e-01 6.89818680e-01 6.81570590e-01 1.15887057e-02 -1.11502625e-01 -9.43045855e-01 3.71610999e-01 8.38795185e-01 6.27045572e-01 3.31801802e-01 9.85160649e-01 -1.67838797e-01 -4.73501772e-01 -9.22088802e-01 8.73830497e-01 2.76464641e-01 -1.55327594e+00 5.31381905e-01 -2.51495987e-01 1.04674423e+00 -1.62419274e-01 2.54045427e-01 4.27643172e-02 7.58572578e-01 -1.49834406e+00 7.93005705e-01 1.01704299e+00 2.74232775e-01 -8.40815365e-01 1.09764063e+00 4.50033635e-01 -4.08357501e-01 -1.61878288e-01 -9.52460349e-01 -1.12401359e-01 -4.55074012e-01 8.61692071e-01 -8.27335179e-01 -5.83124347e-02 8.75248671e-01 3.51991475e-01 -5.98449349e-01 1.12327206e+00 -1.74767092e-01 7.66238093e-01 -5.68309963e-01 -3.73535976e-02 4.83815551e-01 5.64903282e-02 1.73597723e-01 1.58023107e+00 2.69200891e-01 2.51663327e-01 -1.60296336e-01 8.54780555e-01 -3.59802395e-02 -4.45818216e-01 -4.90471572e-01 6.76971003e-02 3.72298002e-01 1.01469493e+00 -1.09761238e+00 -2.69262969e-01 8.74089450e-02 1.25783277e+00 1.05985552e-01 2.51205593e-01 -2.81935424e-01 -1.00792873e+00 -1.04773864e-02 -2.58942872e-01 1.02315414e+00 -1.84286252e-01 -1.00179362e+00 -7.77473271e-01 -1.12663753e-01 -1.20914984e+00 -1.30293548e-01 -9.67694700e-01 -1.28047383e+00 7.09765136e-01 -4.88804162e-01 -9.04409826e-01 1.04402773e-01 -9.47641850e-01 -2.62625813e-01 9.78563070e-01 -1.13740444e+00 -1.13404858e+00 -3.02930474e-01 4.84901875e-01 6.34069085e-01 -4.31606054e-01 1.01287770e+00 -2.59045869e-01 -1.42686412e-01 3.40619326e-01 6.60995245e-01 5.15047371e-01 1.99995816e-01 -1.32967162e+00 4.15921092e-01 1.04984879e+00 3.32051516e-01 2.68474489e-01 7.19739914e-01 -7.61147916e-01 -1.38128459e+00 -1.14703298e+00 6.19513392e-01 -1.87398955e-01 3.56312126e-01 -3.54593486e-01 -1.02958703e+00 7.07517862e-01 4.93902154e-02 9.08775926e-02 2.57261664e-01 -5.10118544e-01 -4.14438277e-01 -2.50336409e-01 -1.52713370e+00 2.82741100e-01 2.62194928e-02 -5.52899539e-01 -1.00767899e+00 3.12456876e-01 -8.55196640e-02 -2.47643515e-01 -6.77259803e-01 1.93599328e-01 5.34775317e-01 -7.93960273e-01 9.75659668e-01 -2.75050342e-01 5.22358418e-01 -1.98059559e-01 -7.14885056e-01 -1.13752139e+00 -5.55221438e-01 -2.01179832e-01 4.02405336e-02 1.01256955e+00 1.53302029e-01 -8.64945278e-02 1.19807470e+00 2.39798389e-02 6.93142042e-02 -3.08320314e-01 -1.11036468e+00 -5.39790332e-01 9.28178951e-02 -2.78373271e-01 -2.52067018e-02 6.29279256e-01 -5.54246426e-01 -2.42520019e-01 -5.94505906e-01 3.99648041e-01 9.24273849e-01 -1.14897221e-01 1.86669201e-01 -1.10099530e+00 -1.47861302e-01 -9.63532105e-02 -1.07892311e+00 -7.77525961e-01 1.57140464e-01 -4.62370276e-01 5.69098592e-01 -1.38107479e+00 9.83022675e-02 -2.69447654e-01 -1.97090611e-01 6.24477088e-01 2.46028602e-01 6.94893956e-01 -1.60047784e-03 1.97983369e-01 -6.74926024e-03 4.31649476e-01 7.99261928e-01 -6.41221583e-01 1.58891797e-01 -1.59724918e-03 -3.45345289e-01 1.01941597e+00 8.26107144e-01 -9.27888751e-01 1.19828232e-01 -6.62685812e-01 -1.69545725e-01 -1.45499766e-01 5.41877568e-01 -1.19397235e+00 5.31258047e-01 2.63635457e-01 1.29530489e+00 -1.05480921e+00 4.80742395e-01 -9.06377733e-01 -6.88136145e-02 8.65131557e-01 -3.26469541e-01 -1.71232060e-01 3.41649026e-01 6.08348489e-01 -6.56717956e-01 -8.81617427e-01 1.16762054e+00 -3.03754061e-01 -9.04494762e-01 -4.53264624e-01 -1.03856385e+00 -7.20256865e-01 5.84912956e-01 -6.58202231e-01 -2.06908584e-01 -1.11417539e-01 -8.49709630e-01 -6.36998177e-01 2.97940895e-02 -8.76719579e-02 1.39092791e+00 -1.21661031e+00 -8.84058297e-01 4.64913458e-01 -5.66025019e-01 -2.96032518e-01 -2.40656715e-02 -3.46137844e-02 -1.45253849e+00 2.57172734e-01 -9.06057596e-01 -6.41434491e-01 -1.21483970e+00 2.11284280e-01 5.21992862e-01 1.11140251e-01 -6.14124656e-01 1.12754965e+00 -7.05233395e-01 -3.24874818e-01 4.94123548e-01 -4.99165565e-01 -1.19512670e-01 -1.51144221e-01 3.85838538e-01 4.98388261e-01 3.30094695e-01 -5.42668104e-01 -3.09369147e-01 4.83512878e-01 -3.36411536e-01 -1.83664307e-01 1.76555109e+00 2.50068516e-01 -4.83603030e-01 3.65612954e-01 1.06773651e+00 -1.98657393e-01 -1.45615160e+00 -3.23658973e-01 6.28152192e-02 -6.90020144e-01 3.89959365e-01 -7.92607129e-01 -9.30846274e-01 8.90054166e-01 1.01855326e+00 -6.23211414e-02 1.17372191e+00 -6.06783032e-01 4.63155240e-01 1.15315795e+00 2.43249878e-01 -1.36033809e+00 9.48484913e-02 6.49145305e-01 8.97286594e-01 -1.11757267e+00 2.90120035e-01 -3.50080162e-01 -4.81016219e-01 1.80313623e+00 8.32022280e-02 -8.73880208e-01 8.27122808e-01 7.77304232e-01 1.42220184e-01 2.08702441e-02 -2.38970026e-01 2.46545807e-01 -9.58041251e-02 1.10612345e+00 -2.93065738e-02 1.35294512e-01 1.60011128e-01 3.66822004e-01 -3.18275630e-01 -1.32955089e-01 9.73818779e-01 1.16559684e+00 -8.08557868e-01 -7.55879760e-01 -8.37113976e-01 5.01595080e-01 -6.27257645e-01 -3.76195371e-01 -3.79842222e-01 2.31466413e-01 1.92260459e-01 6.91778481e-01 2.29802933e-02 -4.72895265e-01 3.70230794e-01 3.72213088e-02 5.74127436e-01 -4.36850220e-01 -6.75597429e-01 -1.60641924e-01 -2.50116795e-01 -2.49775991e-01 -1.38614357e-01 -6.19644284e-01 -8.30570161e-01 -3.95877123e-01 -4.50734019e-01 -1.24652475e-01 1.09688604e+00 5.57519197e-01 -1.99276045e-01 6.68666065e-01 7.48780668e-01 -9.58420992e-01 -9.71173406e-01 -9.47456539e-01 -8.19232464e-01 2.04545120e-03 3.78292948e-01 -9.80225056e-02 -1.67677909e-01 5.04884124e-01]
[11.785651206970215, 2.6096296310424805]
5aafd17b-8164-447a-bfb0-dd86213f89b2
predicting-crime-using-spatial-features
1803.04474
null
http://arxiv.org/abs/1803.04474v1
http://arxiv.org/pdf/1803.04474v1.pdf
Predicting Crime Using Spatial Features
Our study aims to build a machine learning model for crime prediction using geospatial features for different categories of crime. The reverse geocoding technique is applied to retrieve open street map (OSM) spatial data. This study also proposes finding hotpoints extracted from crime hotspots area found by Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN). A spatial distance feature is then computed based on the position of different hotpoints for various types of crime and this value is used as a feature for classifiers. We test the engineered features in crime data from Royal Canadian Mounted Police of Halifax, NS. We observed a significant performance improvement in crime prediction using the new generated spatial features.
['Stan Matwin', 'Amilcar Soares Junior', 'Fateha Khanam Bappee']
2018-03-12
null
null
null
null
['crime-prediction']
['miscellaneous']
[-1.21045753e-01 -3.68852526e-01 1.68260798e-01 -4.37110722e-01 -7.61860490e-01 -8.77399817e-02 6.71722651e-01 9.75364208e-01 -7.88631797e-01 7.19703615e-01 7.92961359e-01 -5.82114875e-01 -6.48763001e-01 -1.53603864e+00 -3.67253482e-01 -2.70418763e-01 -7.36634672e-01 2.19426870e-01 2.30798349e-01 -3.53813842e-02 8.09897244e-01 6.29048228e-01 -1.43912387e+00 7.43112683e-01 9.46488857e-01 5.83694696e-01 3.33363265e-01 6.53793871e-01 -3.18290591e-01 7.94066429e-01 -3.16749424e-01 8.12664162e-03 3.64964962e-01 6.36603981e-02 -9.18610811e-01 -7.80296683e-01 1.20941110e-01 -3.02040994e-01 -3.25024098e-01 6.90939963e-01 3.65370244e-01 3.53599399e-01 8.26802492e-01 -9.61142242e-01 -7.85142422e-01 4.09728318e-01 -8.04562151e-01 1.07361603e+00 7.09280610e-01 -3.10522228e-01 4.46209550e-01 -8.65176380e-01 4.51311857e-01 1.00019240e+00 9.29322362e-01 -3.38702798e-01 -9.32009995e-01 -6.47337019e-01 -5.10927558e-01 5.91353416e-01 -2.10451984e+00 -3.22264493e-01 3.70732158e-01 -9.69904482e-01 1.45038688e+00 2.18734637e-01 5.64848244e-01 2.06202805e-01 3.11897576e-01 1.25706181e-01 1.13220167e+00 -5.40824592e-01 4.80330676e-01 -1.53025538e-01 1.32811964e-01 3.93629432e-01 2.33676121e-01 3.06105446e-02 -4.93231058e-01 -5.98949671e-01 5.14275193e-01 5.51022768e-01 3.36284965e-01 4.25386220e-01 -3.29027265e-01 1.20091152e+00 7.92898953e-01 5.51966250e-01 -6.06107473e-01 -1.87568262e-01 2.83384264e-01 -3.81839797e-02 9.11702454e-01 1.41020510e-02 -1.94555074e-01 -4.93712038e-01 -1.33935142e+00 4.44248825e-01 1.99505284e-01 1.39714673e-01 1.03661919e+00 -2.25101665e-01 8.36533457e-02 1.10175908e+00 5.41723408e-02 3.75385970e-01 4.51522380e-01 -4.62668568e-01 9.57203984e-01 9.79597032e-01 -2.09104031e-01 -1.92734289e+00 -4.14682388e-01 5.60528696e-01 -7.24842668e-01 -6.70488402e-02 3.69683579e-02 -2.30882332e-01 -1.04690838e+00 7.84834385e-01 2.69161552e-01 7.13919461e-01 -2.74940699e-01 5.34479022e-01 6.52807474e-01 7.19698489e-01 6.39118910e-01 1.97735921e-01 7.87814081e-01 1.64563820e-01 -4.25410628e-01 2.95234472e-01 1.12872946e+00 -2.83568025e-01 3.59415293e-01 -6.20224997e-02 -5.22532463e-01 -2.16780365e-01 -3.56556028e-01 2.82791436e-01 -1.15055490e+00 -5.39960146e-01 6.45132959e-01 8.15674961e-01 -1.26035810e+00 6.73472524e-01 -6.11693144e-01 -4.73570585e-01 8.78454924e-01 2.21687257e-01 -7.25211680e-01 -2.33047560e-01 -1.16162133e+00 1.04570925e+00 3.95131290e-01 -4.61497635e-01 6.29962161e-02 -8.18837702e-01 -1.33474123e+00 -9.87987220e-02 -6.91649377e-01 2.66018748e-01 -3.28413658e-02 -2.22626120e-01 -3.35191548e-01 7.00952470e-01 -4.23815250e-01 -3.53522241e-01 4.47794460e-02 2.05235481e-01 -8.49451542e-01 1.17773756e-01 1.11970472e+00 8.37005451e-02 -6.50690962e-03 -6.80418789e-01 -1.21654928e+00 -8.23961973e-01 -2.57923543e-01 -1.45176932e-01 -3.29792112e-01 5.78289270e-01 2.44679287e-01 -6.48054898e-01 3.85818034e-01 -4.59909707e-01 -3.97383839e-01 -9.94917214e-01 -1.06276698e-01 -2.15621963e-01 8.49802017e-01 -1.12457514e+00 1.85792124e+00 -1.81772757e+00 -5.40432572e-01 8.25303316e-01 -1.41428888e-01 2.53610551e-01 2.19986916e-01 6.35201514e-01 -2.98201919e-01 4.12773788e-01 -4.22874123e-01 3.69926900e-01 -5.18515646e-01 1.73202390e-03 -1.79388002e-01 4.75333959e-01 2.20386699e-01 6.43116832e-01 -9.29895282e-01 -5.60092330e-01 7.53288805e-01 3.93855363e-01 -5.12287080e-01 -4.59130138e-01 9.33338463e-01 1.36133268e-01 -5.28809249e-01 5.02529144e-01 1.06011295e+00 2.94408172e-01 -3.53251308e-01 7.00811803e-01 -5.42249978e-01 2.35129178e-01 -1.17345655e+00 1.08325303e+00 -2.00635567e-01 7.76333511e-01 -7.26289690e-01 -8.04041207e-01 1.13643265e+00 4.16605994e-02 6.50360227e-01 -9.79958951e-01 -3.44731450e-01 -6.22344054e-02 -4.05242175e-01 -7.51426101e-01 5.82787156e-01 2.72137135e-01 -4.68295068e-02 5.36063433e-01 -3.79279941e-01 2.35244885e-01 -7.69941807e-02 1.63782001e-01 1.68560123e+00 -4.59204316e-01 3.72678608e-01 -4.73255396e-01 1.92479447e-01 3.15362483e-01 4.90119666e-01 6.55916333e-01 -1.93851948e-01 6.61805093e-01 2.88500637e-01 -1.13381565e+00 -1.01083326e+00 -9.94932353e-01 -6.55817449e-01 9.37765539e-01 -4.94812936e-01 -4.69559461e-01 -6.81849957e-01 -2.55855829e-01 2.84010381e-01 5.87541819e-01 -8.98272455e-01 3.55241984e-01 -4.60668266e-01 -1.19742036e+00 6.12795472e-01 4.23531532e-01 6.83369398e-01 -8.75565290e-01 -8.33389759e-01 1.42699972e-01 1.05826817e-01 -5.92013419e-01 3.78873467e-01 -1.14343546e-01 -6.61421537e-01 -1.17047918e+00 -3.45017850e-01 -5.04565001e-01 4.01230872e-01 3.94654304e-01 4.80056107e-01 2.90191442e-01 -4.91294473e-01 3.42938900e-02 -7.91807890e-01 -4.25897896e-01 1.96962282e-01 -8.64296108e-02 -3.21029648e-02 1.17438510e-01 1.17804933e+00 -7.70865679e-01 -4.87059087e-01 -2.67041683e-01 -5.53582072e-01 -2.30185062e-01 -2.85149217e-01 1.81384742e-01 9.41674486e-02 6.57493949e-01 5.61717272e-01 -5.57667434e-01 7.37674832e-01 -1.40068233e+00 -3.18309635e-01 -9.54371095e-02 -7.22943991e-02 -4.67708617e-01 1.41021192e-01 3.28470498e-01 -7.29737401e-01 1.69183806e-01 -7.94021413e-02 2.02945396e-01 -6.88772559e-01 7.43627667e-01 3.08810174e-01 -1.22994296e-02 8.04483652e-01 -2.69982871e-03 -7.57475615e-01 -3.66345972e-01 -1.66426942e-01 1.12268674e+00 8.73417109e-02 -3.49919736e-01 5.80928266e-01 5.87723970e-01 8.22655261e-02 -1.16556954e+00 9.12612379e-02 -9.20112967e-01 -1.24849820e+00 -2.01669738e-01 1.09093654e+00 -7.28672922e-01 -4.79866356e-01 2.25117564e-01 -1.20219350e+00 6.52234703e-02 2.77043492e-01 5.88675559e-01 -3.28066312e-02 -4.67258431e-02 -4.77774113e-01 -1.30608714e+00 6.40012398e-02 -5.08802474e-01 6.86884463e-01 2.02992976e-01 -4.70919549e-01 -1.13277841e+00 7.04086065e-01 7.35984817e-02 2.52429545e-01 9.01382864e-01 9.40842092e-01 -5.45792103e-01 -9.11028162e-02 -3.83103609e-01 -5.18834114e-01 -3.27755779e-01 5.00875711e-01 4.77877669e-02 -1.03735781e+00 3.35528642e-01 -6.06077492e-01 3.82478625e-01 1.02945960e+00 6.89539075e-01 1.08530939e+00 -3.35416496e-01 -6.28818810e-01 4.58000332e-01 1.74633837e+00 4.67187911e-01 1.09402418e+00 8.64323199e-01 6.22611940e-01 7.70760775e-01 4.35096174e-01 7.79205859e-01 5.82174420e-01 3.08440715e-01 7.98287764e-02 -1.38037391e-02 7.59473979e-01 -3.40831727e-01 -3.42307448e-01 1.46902129e-02 -7.04070926e-01 4.83452916e-01 -2.01191235e+00 1.26700342e+00 -1.80384290e+00 -1.59854770e+00 -7.66510844e-01 1.94840240e+00 3.35622311e-01 -4.66350585e-01 1.61476567e-01 6.03291690e-01 9.81001735e-01 -5.73446117e-02 3.12813222e-01 -6.42355323e-01 -3.79258185e-03 7.64841616e-01 8.45956326e-01 5.91681600e-01 -1.28411925e+00 9.48454499e-01 6.36623812e+00 7.56850004e-01 -7.58371115e-01 2.44853184e-01 7.43673861e-01 -2.09419325e-01 3.57414596e-02 -1.52079687e-02 -3.52987438e-01 8.44362497e-01 1.17981541e+00 -2.04955295e-01 1.49671882e-01 7.67168105e-01 7.96193957e-01 -8.53203714e-01 -1.18114933e-01 1.00906193e+00 -8.92785117e-02 -1.80500507e+00 -9.40321237e-02 3.40779513e-01 7.41071582e-01 1.40045360e-01 9.19658318e-03 2.10302193e-02 8.16513717e-01 -1.18049741e+00 1.01427883e-01 6.26920342e-01 7.77653337e-01 -1.33465254e+00 5.42834222e-01 3.58640373e-01 -1.31925666e+00 -5.69562674e-01 -5.59600413e-01 -7.21720695e-01 1.27575040e-01 5.91933072e-01 -1.05432904e+00 1.67022243e-01 1.26420081e+00 8.31425846e-01 -7.60172248e-01 1.29793942e+00 2.11940646e-01 4.68471229e-01 -2.79874682e-01 3.19763720e-01 5.24497628e-01 -3.44492607e-02 -1.06974408e-01 1.49172890e+00 6.50059879e-01 7.66913354e-01 -2.33521760e-01 6.11531913e-01 6.04551673e-01 1.87320247e-01 -1.34444797e+00 6.16233408e-01 8.90369892e-01 8.06086898e-01 -7.07080841e-01 -4.70901877e-02 -2.78661579e-01 3.00740689e-01 4.97475386e-01 8.23720638e-03 -5.34359157e-01 -6.16629183e-01 5.45183241e-01 8.26060891e-01 -3.45790349e-02 -3.14505845e-01 -5.36993146e-01 -6.86759651e-01 -4.02824968e-01 5.91653325e-02 5.02842247e-01 -6.67048633e-01 -1.14948761e+00 2.75355875e-01 4.32740301e-01 -9.38711584e-01 -3.03754002e-01 -1.11973144e-01 -1.23447466e+00 1.25696564e+00 -1.18848181e+00 -9.56233799e-01 3.30645293e-02 9.96151745e-01 2.37124443e-01 -3.94613832e-01 9.87916529e-01 4.92382735e-01 -3.15984964e-01 -6.04737997e-02 3.81707370e-01 7.68476248e-01 6.30001575e-02 -1.03640759e+00 6.81952059e-01 8.31909478e-01 -2.10275173e-01 7.16582000e-01 7.19287917e-02 -1.30428386e+00 -4.96616453e-01 -1.59500563e+00 1.64964032e+00 -6.88705206e-01 6.06584132e-01 -6.31712675e-02 -9.30628777e-01 5.28184354e-01 -3.39539826e-01 -1.07434951e-01 1.06912637e+00 3.13032061e-01 -2.02233647e-03 2.46669844e-01 -1.76912856e+00 2.36677557e-01 8.08781028e-01 -7.13889360e-01 -6.75619066e-01 3.03889066e-01 1.08894296e-01 2.70455927e-01 -9.25350666e-01 -5.00327325e-04 2.35561863e-01 -1.04891574e+00 9.21409905e-01 -5.15721977e-01 6.51611328e-01 8.28188956e-02 -4.51798975e-01 -1.27759671e+00 -7.06375003e-01 1.86768934e-01 8.09187472e-01 1.28725708e+00 3.36803466e-01 -6.19833350e-01 7.25338876e-01 1.29744017e+00 1.69523790e-01 -2.28701547e-01 -1.48052049e+00 -4.11146015e-01 4.19963539e-01 -9.45883989e-01 1.09864306e+00 1.53549981e+00 6.94624126e-01 -5.21885097e-01 -9.88466367e-02 2.91354984e-01 3.06605786e-01 -5.89898944e-01 4.66314435e-01 -1.13183594e+00 5.90837300e-01 3.92025076e-02 -1.11399472e+00 3.62003535e-01 1.42702222e-01 -8.25570703e-01 -7.85447240e-01 -1.67358339e+00 2.31511116e-01 -9.66740251e-01 -1.44409582e-01 4.54888314e-01 5.52627295e-02 2.51193911e-01 1.07608989e-01 3.48412469e-02 -2.07770944e-01 -1.26275495e-01 1.38615534e-01 1.85359959e-02 -6.55612946e-01 -1.90864339e-01 -3.03193867e-01 7.79828906e-01 1.00517952e+00 -8.31045210e-01 -1.16756568e-02 -6.16470456e-01 1.47498742e-01 1.00627325e-01 5.81225872e-01 -1.40635633e+00 4.47593361e-01 -6.51465774e-01 8.79090965e-01 -9.20899630e-01 1.37481943e-01 -8.87853503e-01 2.84014791e-01 2.56602079e-01 -8.65787789e-02 3.43864918e-01 2.48287305e-01 3.18848103e-01 -3.19318920e-02 -8.94925371e-02 4.58638012e-01 -9.08530056e-02 -8.55586052e-01 3.40279162e-01 -7.55576909e-01 -4.71893430e-01 1.34066594e+00 -8.54545832e-01 -1.93072602e-01 4.58772890e-02 -6.65371716e-01 -6.22257218e-02 2.06271157e-01 2.45165721e-01 9.05056715e-01 -1.46411395e+00 -7.88900793e-01 4.78961051e-01 -1.55462489e-01 -4.94209766e-01 4.33622956e-01 4.38179404e-01 -8.57618213e-01 6.48168683e-01 -5.55292904e-01 -3.11266810e-01 -1.29951537e+00 2.96024173e-01 3.44768465e-02 1.97738651e-02 -5.59559166e-01 4.00021613e-01 -5.97244799e-01 -3.45081210e-01 -5.68724751e-01 3.11243813e-02 -8.11119556e-01 1.61397561e-01 9.85344291e-01 1.07346821e+00 1.86685339e-01 -1.33416152e+00 -6.16956353e-01 5.66296756e-01 4.09870058e-01 -1.92524478e-01 1.93273020e+00 -5.71724921e-02 -2.13361219e-01 1.96429834e-01 1.29345644e+00 -2.37675443e-01 -4.13275063e-01 1.32916003e-01 6.46053433e-01 -1.34229469e+00 1.15954556e-01 -3.15974146e-01 -8.05415571e-01 8.95912647e-01 7.98107922e-01 3.26324999e-01 8.86340261e-01 -1.07616678e-01 3.70969087e-01 6.54135272e-02 6.87510788e-01 -1.41838288e+00 -8.95647407e-01 6.95243895e-01 4.83918548e-01 -1.28909171e+00 7.66210780e-02 -1.51332223e-03 -5.50444186e-01 9.45444643e-01 1.65022254e-01 -4.79464173e-01 1.38799095e+00 3.61464947e-01 -5.90903580e-01 -3.89833152e-01 -1.64649278e-01 -3.75549912e-01 -1.10244118e-01 1.22213554e+00 2.97555596e-01 4.80725765e-01 -2.46341646e-01 3.78005385e-01 -3.19814742e-01 8.65302235e-02 4.09699261e-01 1.12162638e+00 -8.26738417e-01 -4.75264311e-01 -8.14714909e-01 1.05857861e+00 -6.06185019e-01 -4.19788182e-01 -2.77836770e-01 7.22174168e-01 6.35028839e-01 1.25783801e+00 6.72791064e-01 -5.39615631e-01 1.62612349e-01 -4.68041562e-02 -1.90785348e-01 -7.29718029e-01 -7.00090528e-01 -7.13766038e-01 -1.54523507e-01 -3.76028061e-01 -1.49013028e-01 -9.23170388e-01 -1.36470556e+00 -1.00702190e+00 1.14657626e-01 2.75159419e-01 7.66959131e-01 9.02954102e-01 3.48161101e-01 -2.30112329e-01 7.55800366e-01 -8.07169616e-01 8.07349265e-01 -9.07340825e-01 -9.84474003e-01 4.12538826e-01 2.39927799e-01 -6.04188442e-01 7.64793232e-02 1.15840714e-02]
[6.761672019958496, 1.9470819234848022]
315d6bf6-0b80-4814-af9b-dcb4fb874884
bert-based-chinese-text-classification-for
2104.04197
null
https://arxiv.org/abs/2104.04197v1
https://arxiv.org/pdf/2104.04197v1.pdf
BERT-based Chinese Text Classification for Emergency Domain with a Novel Loss Function
This paper proposes an automatic Chinese text categorization method for solving the emergency event report classification problem. Since bidirectional encoder representations from transformers (BERT) has achieved great success in natural language processing domain, it is employed to derive emergency text features in this study. To overcome the data imbalance problem in the distribution of emergency event categories, a novel loss function is proposed to improve the performance of the BERT-based model. Meanwhile, to avoid the impact of the extreme learning rate, the Adabound optimization algorithm that achieves a gradual smooth transition from Adam to SGD is employed to learn parameters of the model. To verify the feasibility and effectiveness of the proposed method, a Chinese emergency text dataset collected from the Internet is employed. Compared with benchmarking methods, the proposed method has achieved the best performance in terms of accuracy, weighted-precision, weighted-recall, and weighted-F1 values. Therefore, it is promising to employ the proposed method for real applications in smart emergency management systems.
['Xiong Luo', 'Chao Huang', 'Long Wang', 'Zhongju Wang']
2021-04-09
null
null
null
null
['text-categorization']
['natural-language-processing']
[-3.90839810e-03 -1.20652318e-01 4.47151251e-02 -4.64233518e-01 -7.88999438e-01 6.87243268e-02 5.28400242e-01 4.78008002e-01 -6.83128178e-01 7.64516354e-01 3.94255340e-01 -3.42838407e-01 -3.73186678e-01 -8.31495941e-01 -2.02672735e-01 -8.08001876e-01 -2.27262340e-02 2.40161106e-01 -1.83650225e-01 -2.49734268e-01 2.87310064e-01 1.75036713e-01 -1.32289362e+00 1.83229685e-01 1.13466227e+00 1.31170297e+00 1.32569715e-01 1.06468700e-01 -2.79421151e-01 9.62391317e-01 -6.02358282e-01 -3.86647642e-01 2.11083278e-01 -2.59832621e-01 -7.53706813e-01 -6.41869456e-02 -5.54113746e-01 -3.94820333e-01 -3.71051133e-01 8.31095338e-01 8.21075141e-01 4.82027352e-01 6.42728388e-01 -1.13022006e+00 -2.55875617e-01 5.33938706e-01 -5.21942258e-01 3.14138710e-01 1.62663072e-01 -4.15870517e-01 9.54188883e-01 -8.78214598e-01 1.68987393e-01 9.77332532e-01 4.39783782e-01 2.04187021e-01 -6.56871736e-01 -6.74544036e-01 4.55449373e-01 3.18891913e-01 -1.51334059e+00 -1.51364561e-02 7.60964394e-01 -3.53244036e-01 8.41339827e-01 1.58058573e-02 2.73765832e-01 7.94955552e-01 4.56866115e-01 9.06523228e-01 8.66379440e-01 -2.54952461e-01 2.19002634e-01 2.89957702e-01 3.79287213e-01 5.29144049e-01 -5.13700880e-02 -1.34130493e-01 -3.33901048e-01 -6.40690774e-02 2.29899213e-02 2.47586235e-01 -2.10066870e-01 1.49693191e-01 -9.65976596e-01 9.05694544e-01 4.70811129e-01 2.82080442e-01 -6.40131593e-01 -3.72366637e-01 8.07390749e-01 1.74063772e-01 7.89326906e-01 -7.09528700e-02 -2.83391774e-01 -3.09959799e-01 -8.26254368e-01 1.37127742e-01 4.76662636e-01 7.57670164e-01 8.95835236e-02 2.49636889e-01 -3.16406637e-01 8.84361327e-01 4.03136536e-02 4.51654762e-01 7.62440085e-01 -1.40373632e-02 9.57647741e-01 8.33166242e-01 9.87759233e-02 -1.26878965e+00 -7.14444637e-01 -4.67720419e-01 -1.14930439e+00 -4.08176303e-01 -9.83292386e-02 -3.32726181e-01 -5.17909467e-01 1.26594853e+00 3.18432450e-01 -1.05923355e-01 1.39244154e-01 1.08739018e+00 6.84612632e-01 9.62621093e-01 3.37502003e-01 -3.62291008e-01 1.19257081e+00 -5.21934450e-01 -7.29924440e-01 1.74417272e-01 5.01403809e-01 -4.06810641e-01 9.67805982e-01 3.52853090e-01 -8.01477075e-01 -4.83177960e-01 -8.64343524e-01 1.93674803e-01 -3.30361634e-01 4.15394038e-01 3.18807513e-01 3.33970487e-01 -3.40366155e-01 2.87863910e-01 -6.90014839e-01 -3.43886018e-01 2.46880680e-01 2.23692864e-01 -8.90685022e-02 3.45495343e-02 -1.53123486e+00 8.47493589e-01 7.36499846e-01 3.00574720e-01 -4.14469302e-01 -3.05055648e-01 -4.62280512e-01 4.11751747e-01 1.40591547e-01 -2.71058142e-01 8.30965817e-01 -6.76463008e-01 -1.37424588e+00 3.97901386e-01 1.74848318e-01 -6.89571023e-01 5.97671807e-01 -3.15625876e-01 -4.13475752e-01 1.87325433e-01 -1.15434185e-01 2.54955739e-01 6.98619723e-01 -6.80427849e-01 -9.24419999e-01 -3.09177697e-01 -2.48653628e-03 3.36664379e-01 -8.33401084e-01 1.55696616e-01 3.52802016e-02 -8.47962976e-01 2.15954948e-02 -6.16404116e-01 1.15118576e-02 -5.11313617e-01 -2.11192638e-01 -4.99448568e-01 6.81172788e-01 -9.90253806e-01 1.56520581e+00 -2.19375968e+00 -7.09664673e-02 3.41398925e-01 -1.24633484e-01 2.70851970e-01 1.94664493e-01 6.16139829e-01 8.41518268e-02 -8.59127119e-02 -4.27409351e-01 -1.63988739e-01 1.50403043e-03 5.76168150e-02 -5.64185441e-01 3.38099599e-01 2.31139153e-01 3.28870982e-01 -6.65452361e-01 -5.58541715e-01 1.92835510e-01 6.04394019e-01 -3.41892332e-01 5.03425181e-01 1.66585073e-01 3.40362549e-01 -7.51140773e-01 2.17017964e-01 6.51131749e-01 -2.03533117e-02 -9.35082138e-03 -1.83325335e-02 -1.77923769e-01 3.34636480e-01 -1.09777772e+00 1.07621646e+00 -6.65992320e-01 2.69810617e-01 -4.55786213e-02 -1.49544859e+00 1.35669947e+00 4.68532473e-01 7.80595541e-01 -8.57234061e-01 4.47869986e-01 -1.37042269e-01 -3.65003973e-01 -7.36420989e-01 5.53024113e-01 -2.47194432e-02 -1.54510543e-01 4.12208438e-01 -2.26237342e-01 1.59757823e-01 1.96586132e-01 6.83249310e-02 7.80485809e-01 -2.70868361e-01 2.00339898e-01 -3.75246733e-01 8.94831419e-01 -5.92225492e-02 6.56136096e-01 4.08166856e-01 -4.83476296e-02 4.43940341e-01 2.23425835e-01 -6.78778708e-01 -7.45628774e-01 -8.48846138e-01 -2.77314484e-01 9.68068123e-01 7.54847154e-02 -2.80588448e-01 -6.07130766e-01 -7.87857056e-01 -2.38817662e-01 8.76117170e-01 -3.16861391e-01 -3.58844101e-01 -4.25040662e-01 -1.11469710e+00 5.27188778e-01 5.63202441e-01 8.47845197e-01 -1.02319551e+00 -7.56985128e-01 3.84403497e-01 -7.89096475e-01 -1.00786424e+00 -3.52056831e-01 7.74797201e-02 -6.35963202e-01 -9.78886187e-01 -6.72114909e-01 -7.18834937e-01 5.87840676e-01 4.40359004e-02 4.31556702e-01 1.86049845e-02 -6.00518398e-02 2.35487834e-01 -8.09539735e-01 -4.60600704e-01 -1.82154551e-01 4.18358177e-01 7.93397352e-02 4.89640474e-01 3.98216248e-01 -4.94575053e-02 -5.98548412e-01 1.27179965e-01 -1.05327022e+00 -2.98057143e-02 1.82165220e-01 1.16679037e+00 1.82277247e-01 4.91131335e-01 1.03183556e+00 -3.39884639e-01 1.13426280e+00 -5.62299430e-01 -6.07903481e-01 2.79565603e-01 -9.57575560e-01 2.11334005e-02 1.01539207e+00 -2.03035891e-01 -1.32723641e+00 -3.13116878e-01 -1.85848042e-01 1.62015498e-01 1.60475165e-01 8.66543651e-01 5.23725674e-02 4.81360286e-01 2.21569732e-01 5.08693635e-01 -1.92775309e-01 -4.69632179e-01 -1.55188844e-01 1.27456558e+00 2.65013665e-01 -3.96617919e-01 3.87189478e-01 2.06553161e-01 -3.58187556e-01 -7.39269793e-01 -7.56090403e-01 -5.46046913e-01 -2.47870535e-01 -1.67769417e-01 8.79978776e-01 -9.67589021e-01 -6.70951307e-01 5.06054759e-01 -1.01028156e+00 2.43279561e-01 6.88154399e-02 7.49569535e-01 -2.24381089e-01 2.88901299e-01 -5.58043718e-01 -1.13983810e+00 -1.03233778e+00 -8.66696358e-01 7.66331136e-01 2.10408658e-01 2.87178662e-02 -8.48657012e-01 -3.45003605e-01 2.14444503e-01 5.55354774e-01 1.53002039e-01 9.75480318e-01 -8.12875152e-01 3.94314937e-02 -5.33261180e-01 -2.63713628e-01 6.22480631e-01 1.96547493e-01 -1.82307571e-01 -6.69163644e-01 -5.06821990e-01 3.04700583e-01 -2.68956184e-01 6.28438592e-01 5.44597208e-02 1.32620394e+00 -5.30253172e-01 -1.30613461e-01 3.89458448e-01 1.24856317e+00 4.89419222e-01 5.34500122e-01 6.89884841e-01 4.05334622e-01 5.11940062e-01 8.49314094e-01 1.13324380e+00 7.10315645e-01 3.87752652e-01 1.77309141e-01 1.25210777e-01 5.36073446e-01 -1.54933751e-01 3.39100093e-01 1.07186711e+00 1.79398999e-01 -3.34830374e-01 -9.10486281e-01 4.73445892e-01 -1.76587641e+00 -8.37575555e-01 2.56394297e-01 2.17503285e+00 5.22748351e-01 2.63106704e-01 -8.10531899e-02 5.35513341e-01 7.22402036e-01 1.30226925e-01 -2.75714040e-01 -5.11005461e-01 -1.18426764e-02 -1.81813501e-02 1.66897237e-01 3.83236632e-02 -1.26734829e+00 6.20580971e-01 4.82283545e+00 8.38208199e-01 -1.37223768e+00 -1.84528790e-02 6.85857415e-01 1.10864334e-01 9.65254307e-02 -3.89833272e-01 -6.69071555e-01 8.57115626e-01 8.33007097e-01 -3.36915404e-01 3.93830091e-01 6.77042842e-01 5.71599007e-01 1.15532298e-02 -5.08127093e-01 9.31467354e-01 3.11150011e-02 -8.52239072e-01 8.30901712e-02 -4.13672537e-01 4.70823139e-01 -2.96340678e-02 2.65671574e-02 3.83439213e-01 -6.70117065e-02 -5.75387716e-01 6.39414668e-01 5.51365316e-01 5.46634853e-01 -1.06460357e+00 1.15517092e+00 5.72438955e-01 -1.04906189e+00 -6.23007298e-01 -3.46186936e-01 1.78694315e-02 2.63353080e-01 6.42901480e-01 -9.06739295e-01 9.27494705e-01 7.38808393e-01 4.39081252e-01 -2.31777757e-01 8.40016723e-01 2.22148281e-02 7.55232692e-01 -2.35440791e-01 -2.04310507e-01 2.98213869e-01 -2.49125630e-01 3.75412703e-01 1.25810122e+00 4.70036000e-01 2.50252277e-01 4.09498811e-01 2.63134956e-01 4.99233864e-02 5.69906950e-01 -3.47059160e-01 -3.25196125e-02 2.84596801e-01 1.07705748e+00 -6.00203454e-01 -2.96620041e-01 -3.50181103e-01 5.41777253e-01 1.59922063e-01 2.61832088e-01 -1.03797233e+00 -7.65475571e-01 2.47306433e-02 1.95582006e-02 2.26621687e-01 -1.08899064e-01 -2.70902067e-01 -1.08339727e+00 1.26058027e-01 -7.11280406e-01 7.09035993e-01 -5.38725376e-01 -1.22210538e+00 9.86270010e-01 2.38353051e-02 -1.55380821e+00 -2.11857647e-01 -2.89980412e-01 -6.52099311e-01 5.49588144e-01 -1.42862368e+00 -8.59492838e-01 -2.71568626e-01 5.97616851e-01 6.13821924e-01 -3.40210915e-01 6.16916835e-01 5.27968824e-01 -7.80402660e-01 6.63117349e-01 4.12384182e-01 2.01123625e-01 4.43110287e-01 -7.73673594e-01 -8.27838928e-02 7.32431471e-01 -3.62236500e-01 3.66363555e-01 4.36130971e-01 -4.32254910e-01 -9.22742009e-01 -1.19165790e+00 1.05108356e+00 2.49691114e-01 4.56168890e-01 -2.03004420e-01 -7.84405649e-01 2.69521415e-01 5.40802889e-02 -4.02978629e-01 4.21217054e-01 -3.19982201e-01 6.10574037e-02 -6.46863818e-01 -1.21439171e+00 2.69676208e-01 3.20079476e-01 -3.08491349e-01 -6.57571554e-01 3.49773794e-01 5.84013224e-01 -2.83679426e-01 -9.04405415e-01 4.53952342e-01 3.33054215e-01 -5.14991879e-01 6.14867091e-01 -5.76615810e-01 4.11421776e-01 -1.10876001e-01 -3.14771384e-02 -1.44837022e+00 -1.00654781e-01 -2.16269523e-01 2.38544047e-02 1.40048265e+00 2.37087235e-01 -8.62228870e-01 1.84499070e-01 5.46931863e-01 -8.29801485e-02 -8.72229338e-01 -1.04280102e+00 -5.57534695e-01 1.30186817e-02 -1.89877853e-01 7.02937663e-01 6.88393354e-01 3.09696287e-01 2.05325171e-01 -4.64893967e-01 -3.68502066e-02 4.00110662e-01 1.18248381e-01 3.78706992e-01 -9.93592978e-01 2.47173324e-01 -1.78204402e-01 -1.64588511e-01 -7.72688091e-01 1.64505795e-01 -9.04160202e-01 5.04878089e-02 -1.46005583e+00 -5.11397645e-02 -4.62797731e-01 -7.71573424e-01 1.56353161e-01 -3.55425656e-01 -2.36199424e-01 1.97559401e-01 -1.42005563e-01 -5.41405678e-01 1.00670707e+00 8.22715819e-01 -3.83121997e-01 -1.86278179e-01 3.62724543e-01 -5.08608580e-01 3.87540609e-01 1.04528308e+00 -5.90031743e-01 -3.92883062e-01 -5.00402093e-01 6.10008910e-02 1.72157049e-01 -4.26009996e-03 -9.14692342e-01 2.93577164e-01 2.67188810e-03 1.55435815e-01 -6.45245552e-01 1.23997303e-02 -9.58892524e-01 -3.22391540e-01 4.28998351e-01 -5.69797575e-01 5.07072866e-01 -8.94351155e-02 4.34225380e-01 -4.93754804e-01 -1.34311229e-01 5.67689538e-01 2.54472584e-01 -3.81884396e-01 2.76027054e-01 -5.62539518e-01 1.49842158e-01 8.06131721e-01 9.34019461e-02 -2.22746432e-01 -4.18466300e-01 -3.07508558e-01 6.07122302e-01 -2.18278497e-01 7.37805545e-01 7.31034100e-01 -1.08773959e+00 -8.27055156e-01 1.87834442e-01 1.56810775e-01 -8.80819112e-02 4.07158524e-01 9.48003411e-01 -4.55223411e-01 3.80882174e-01 8.22647214e-02 -3.17727834e-01 -7.69721925e-01 4.93918300e-01 1.12392895e-01 -4.84485984e-01 -7.65959978e-01 2.92101234e-01 -2.19357595e-01 -3.66824627e-01 4.74640876e-01 -3.05286974e-01 -4.33920175e-01 1.60785094e-01 6.10742927e-01 5.96448421e-01 4.98318374e-01 -5.72325528e-01 -4.10682917e-01 1.61162153e-01 -1.21348619e-01 1.12754636e-01 1.43785977e+00 -2.75923312e-01 3.93964201e-02 2.02334255e-01 1.19618630e+00 -3.34563196e-01 -1.04504383e+00 -9.99146774e-02 1.79515615e-01 -1.40431315e-01 1.86674118e-01 -7.26561546e-01 -1.03977096e+00 1.01357114e+00 8.39129448e-01 2.45395109e-01 1.30101001e+00 -6.01744056e-01 1.13012648e+00 4.32724386e-01 2.68211871e-01 -1.52487719e+00 -2.09497184e-01 6.86863244e-01 8.09174776e-01 -1.17158353e+00 -2.09205493e-01 1.99957713e-01 -1.08410800e+00 1.05602777e+00 4.97493416e-01 -8.60864893e-02 8.05509031e-01 1.64496034e-01 1.31250307e-01 1.46948248e-01 -6.32323027e-01 -2.61521582e-02 4.40113805e-02 1.96611345e-01 3.56458694e-01 2.63256952e-02 -9.14610147e-01 8.39473426e-01 -9.39906836e-02 2.51060389e-02 3.03258359e-01 7.81401336e-01 -2.33516365e-01 -6.08962595e-01 -1.37849241e-01 6.17089510e-01 -8.76406193e-01 3.11381146e-02 1.25369579e-01 5.95030963e-01 -1.95492893e-01 1.24499083e+00 3.62090357e-02 -5.96516013e-01 4.75836515e-01 2.28904918e-01 -1.49348095e-01 3.54651883e-02 -7.29933679e-01 -2.86555618e-01 -1.60205334e-01 -1.30551532e-01 -2.70902306e-01 -4.43563640e-01 -1.56918645e+00 -1.65809199e-01 -5.14186561e-01 6.07223988e-01 7.36918628e-01 1.08934319e+00 3.98540944e-01 5.66432178e-01 1.27667236e+00 -1.90396458e-01 -8.85515034e-01 -1.09049380e+00 -4.81872946e-01 5.32947898e-01 1.03956617e-01 -6.59507394e-01 -3.78534585e-01 -1.52139306e-01]
[9.073366165161133, 3.966703176498413]
a558b6f8-a60c-4dbe-8030-3a8e612322d3
iitransformer-a-unified-approach-to
null
null
https://bmvc2022.mpi-inf.mpg.de/377/
https://bmvc2022.mpi-inf.mpg.de/0377.pdf
iiTransformer: A Unified Approach to Exploiting Local and Non-Local Information for Image Restoration
The goal of image restoration is to recover a high-quality image from its degraded input. While impressive results on various image restoration tasks have been achieved using CNNs, the convolution operation has limited its ability to utilize information outside of its receptive field. Transformers, which use the self-attention mechanism to model long-range dependencies of its input, have demonstrated promising results in various high-level vision tasks. In this paper, we propose intra-inter Transformer (iiTransformer) by explicitly modelling long-range dependencies at the pixel- and patch-levels since there are benefits to considering both local and non-local feature correlations. In addition, we provide a boundary artifact-free solution to support images with arbitrary sizes. We demonstrate the potential of iiTransformer as a general purpose backbone architecture through extensive experiments on various image restoration tasks.
['Tammy Lee', 'Hanul Shin', 'Youngchan Song', 'Soo Min Kang']
2022-11-21
null
null
null
bmvc-2022-11
['color-image-denoising', 'jpeg-compression-artifact-reduction']
['computer-vision', 'computer-vision']
[ 4.23268586e-01 -2.07716957e-01 2.58421093e-01 -3.40833098e-01 -5.93369007e-01 -4.20127772e-02 5.29059708e-01 -3.98427814e-01 -1.71193719e-01 5.79769909e-01 5.97423494e-01 -2.23646134e-01 -8.17618594e-02 -6.73012614e-01 -8.85870457e-01 -8.21587443e-01 3.32650580e-02 -2.95955032e-01 1.95029154e-01 -3.09263170e-01 2.92516321e-01 8.31390738e-01 -1.63280392e+00 8.23165715e-01 9.36504960e-01 1.01281214e+00 5.36838651e-01 5.64033091e-01 3.11187238e-01 9.56510842e-01 -5.72526574e-01 9.27171558e-02 4.66031462e-01 -2.00321555e-01 -7.24208891e-01 4.17183101e-01 1.01094115e+00 -5.23557603e-01 -6.60321474e-01 9.57737267e-01 5.54315329e-01 2.18012691e-01 3.04769278e-01 -8.03433061e-01 -1.17288983e+00 4.31798249e-02 -5.76300025e-01 4.87996727e-01 1.61121320e-02 3.71496797e-01 7.56383717e-01 -1.00232196e+00 3.95322055e-01 1.24131775e+00 6.99808955e-01 3.86384934e-01 -1.61487448e+00 -3.94784719e-01 9.44364890e-02 4.96718109e-01 -8.79634559e-01 -6.98686779e-01 7.53820360e-01 -1.75599068e-01 1.64289653e+00 1.41377121e-01 4.77655560e-01 5.87355077e-01 6.23596668e-01 5.81861377e-01 1.35991681e+00 -3.55091751e-01 3.83497924e-02 -4.50007081e-01 1.57814585e-02 3.84920895e-01 -1.46414667e-01 1.16080642e-01 -7.07460761e-01 9.15123522e-02 1.23229015e+00 9.98649374e-02 -7.56758273e-01 -1.86098397e-01 -1.07005310e+00 5.46810806e-01 1.11606228e+00 3.69726509e-01 -6.82905495e-01 3.50215286e-01 1.05729662e-01 3.54717791e-01 5.07859468e-01 3.74820709e-01 -3.81332666e-01 5.34014285e-01 -8.60714555e-01 -9.52524319e-02 2.64712691e-01 5.34826040e-01 7.84522891e-01 2.88000017e-01 -2.75701314e-01 1.08283103e+00 6.85888752e-02 1.09894812e-01 4.61158216e-01 -1.19164872e+00 1.69308156e-01 3.34071517e-01 6.26436621e-02 -8.59437943e-01 -3.63914818e-01 -8.42213511e-01 -1.21611905e+00 8.96210432e-01 1.93177551e-01 4.90411729e-01 -1.33318269e+00 1.56030917e+00 -7.12559149e-02 1.78542301e-01 3.24441418e-02 1.11260164e+00 8.06376219e-01 6.38774216e-01 1.11091486e-03 -1.42962247e-01 1.19073796e+00 -1.10986078e+00 -6.89553559e-01 -5.04304409e-01 -1.86511700e-03 -7.62766480e-01 1.23161697e+00 4.97865915e-01 -1.21965945e+00 -7.96522260e-01 -1.11715388e+00 -6.05854213e-01 -5.23742773e-02 -7.13258050e-03 5.12586415e-01 1.92747518e-01 -1.67193544e+00 8.85368049e-01 -8.35638583e-01 -2.69479454e-01 8.13868642e-01 4.35660630e-01 -6.78225100e-01 -4.72305477e-01 -6.32100821e-01 9.34068739e-01 -2.10675269e-01 2.69171536e-01 -9.68946218e-01 -7.81995535e-01 -7.27094471e-01 2.92426258e-01 -1.63447604e-01 -9.52199399e-01 8.37444186e-01 -1.06898713e+00 -1.20357585e+00 6.79458141e-01 -3.70520562e-01 -4.68255430e-01 3.99692059e-02 -1.14767700e-01 -3.19874674e-01 2.81010538e-01 1.10441811e-01 8.02198410e-01 1.04746902e+00 -1.19940150e+00 -3.38325351e-01 -3.96720797e-01 8.75595659e-02 2.74187565e-01 -3.70678931e-01 -1.67565029e-02 -3.31369877e-01 -8.86504471e-01 4.51123953e-01 -5.32499254e-01 -3.50104779e-01 2.97836810e-01 -2.47625321e-01 1.07368626e-01 9.58090723e-01 -9.70777810e-01 5.48493564e-01 -2.23718190e+00 5.56178652e-02 -1.90342441e-01 2.42850453e-01 3.92762572e-01 -4.55834508e-01 3.56611252e-01 -4.15186256e-01 -4.52876501e-02 -3.84936839e-01 -3.72756541e-01 -4.95780349e-01 4.29706752e-01 -4.54520196e-01 6.00071251e-01 3.72274339e-01 8.55084956e-01 -4.20784891e-01 -4.19351868e-02 4.69036639e-01 1.02208257e+00 -6.75207138e-01 1.10778280e-01 5.03963009e-02 4.90017444e-01 9.96196941e-02 5.81828356e-01 8.23633611e-01 -4.75606620e-01 -5.30689470e-02 -7.24844098e-01 -1.09113298e-01 2.69264460e-01 -5.61735690e-01 1.83271074e+00 -7.15789080e-01 9.19133604e-01 4.70814973e-01 -1.06347382e+00 5.90598941e-01 2.33717948e-01 3.94191056e-01 -1.12928092e+00 -1.83474883e-01 1.39260739e-02 8.71045813e-02 -2.69651085e-01 4.18993860e-01 -3.05255353e-01 6.42815113e-01 2.70292461e-01 2.41703242e-01 9.44873840e-02 8.58476013e-03 1.42319873e-01 1.34508407e+00 1.86905533e-01 -1.52181359e-02 -6.03640556e-01 3.32273960e-01 -1.29926473e-01 4.45703834e-01 8.10666502e-01 -1.31137282e-01 1.16947055e+00 6.78795576e-02 -6.42321646e-01 -9.93052661e-01 -1.07482755e+00 -3.71890962e-01 7.94722080e-01 1.75468743e-01 -1.43585697e-01 -4.61840183e-01 -3.18294674e-01 -1.71699733e-01 3.83171171e-01 -5.27892768e-01 -1.32551029e-01 -6.08391881e-01 -9.24890518e-01 4.82730754e-02 6.16868496e-01 7.92407870e-01 -1.26978111e+00 -5.03717959e-01 4.18922782e-01 -2.73027718e-01 -1.15611649e+00 -4.51688349e-01 4.48058367e-01 -1.12250686e+00 -9.41609919e-01 -6.39453232e-01 -9.33467090e-01 8.69545281e-01 7.40328372e-01 1.14748359e+00 2.97712058e-01 -5.71042478e-01 1.89527288e-01 -4.39474769e-02 3.15475196e-01 2.60750186e-02 -4.91067022e-01 -1.91473886e-01 2.07005031e-02 -1.89510211e-01 -1.11284876e+00 -9.09184337e-01 3.10782045e-01 -1.04747009e+00 9.48467478e-02 6.38655186e-01 1.18699753e+00 5.53699970e-01 1.84235156e-01 5.54841697e-01 -4.22222644e-01 5.52999496e-01 3.93163227e-02 -1.88690767e-01 7.24929497e-02 -4.18051422e-01 -4.26477231e-02 6.72741950e-01 -3.19303572e-01 -1.15654516e+00 -7.41988793e-02 -2.78265387e-01 -4.13947105e-01 -1.11162893e-01 3.14464509e-01 -2.74189889e-01 -4.10087854e-01 7.53983974e-01 3.46261770e-01 1.50477469e-01 -7.11390853e-01 1.51210785e-01 3.20556879e-01 8.70829701e-01 -4.36207056e-01 5.50819218e-01 7.53153026e-01 -3.77817010e-03 -8.97391260e-01 -7.93588638e-01 -2.24913314e-01 -5.28608084e-01 -4.46917154e-02 7.53997266e-01 -1.21925485e+00 -6.06267869e-01 8.43134165e-01 -1.15854275e+00 -6.46298110e-01 -4.51704681e-01 1.22995973e-01 -5.50932407e-01 3.62322539e-01 -9.81923342e-01 -3.07247519e-01 -3.49051833e-01 -1.23638272e+00 8.71485293e-01 3.21113080e-01 3.52286696e-01 -9.46604133e-01 -4.37964678e-01 4.96045738e-01 8.86346161e-01 -4.40466590e-02 8.90625358e-01 2.33966578e-02 -9.55682278e-01 1.26000360e-01 -6.59578860e-01 7.94554293e-01 2.17641041e-01 -5.30237675e-01 -1.12041998e+00 -5.36214471e-01 1.99343294e-01 -3.07434767e-01 1.40303409e+00 7.32369363e-01 1.36559069e+00 -2.46151239e-01 -4.08284739e-02 9.43817675e-01 1.45968914e+00 -1.91461429e-01 1.18981147e+00 3.57581496e-01 7.26745486e-01 3.18550915e-01 4.50887978e-02 -4.12113294e-02 2.10443765e-01 6.77626431e-01 7.03487694e-01 -7.03060269e-01 -1.06662536e+00 -1.37351770e-02 3.50216508e-01 3.84827346e-01 -6.21667318e-03 -2.85256654e-01 -6.52780533e-01 6.91782594e-01 -1.69745660e+00 -9.10310626e-01 -1.97591349e-01 1.98884714e+00 7.86498725e-01 -1.47831798e-01 -5.89412808e-01 2.09328067e-02 4.92916346e-01 2.54322171e-01 -6.41322076e-01 -1.90747112e-01 -5.76123595e-01 3.95761460e-01 3.90975714e-01 6.06803417e-01 -9.86585915e-01 7.45696068e-01 7.27881527e+00 7.53419936e-01 -1.10245502e+00 2.17812419e-01 8.79912794e-01 -1.20347373e-01 -2.31354684e-01 3.25095057e-02 -3.28875959e-01 1.03007115e-01 4.35803711e-01 3.87735218e-01 6.85076118e-01 2.59819090e-01 2.45880380e-01 -2.71205008e-01 -8.09639275e-01 8.38164210e-01 2.08210170e-01 -1.40395677e+00 8.22764710e-02 1.72337189e-01 8.76691401e-01 3.47152650e-01 2.51116067e-01 -2.44239718e-01 2.05046132e-01 -1.30177164e+00 5.05054951e-01 5.44943511e-01 8.48158062e-01 -6.00699008e-01 4.48018372e-01 1.27871856e-01 -8.46289694e-01 -2.59880006e-01 -4.56990093e-01 -2.30478600e-01 2.15064481e-01 8.70423019e-01 -3.46503079e-01 4.25817251e-01 1.11215734e+00 1.06512129e+00 -5.88868737e-01 1.16149843e+00 -2.81634301e-01 3.79410207e-01 -8.91392976e-02 9.31904435e-01 -5.83247580e-02 -2.13179111e-01 4.93866920e-01 8.21815372e-01 2.84663111e-01 1.75147563e-01 -1.92492381e-02 8.48645747e-01 -3.11420057e-02 -2.74672538e-01 -3.46779048e-01 4.86032963e-01 -1.65715963e-01 1.18265224e+00 -6.53726161e-01 -2.01846376e-01 -4.91082489e-01 1.27978456e+00 3.74756247e-01 7.94411540e-01 -3.95334661e-01 -1.28199443e-01 1.01282954e+00 4.38213110e-01 4.51131493e-01 -2.45179430e-01 -5.62031031e-01 -1.20073855e+00 1.83923349e-01 -8.77844870e-01 8.16623196e-02 -1.24278820e+00 -1.38087964e+00 8.28608811e-01 -4.24102783e-01 -9.98678029e-01 3.17065626e-01 -6.94515467e-01 -5.69129765e-01 1.16371572e+00 -1.76882613e+00 -1.41009712e+00 -3.17499101e-01 8.96684885e-01 5.17023981e-01 7.63424765e-03 7.10790813e-01 3.89987260e-01 -3.24997544e-01 2.22704500e-01 1.36981472e-01 -8.29009265e-02 6.86505556e-01 -9.50254023e-01 3.38457882e-01 1.24110246e+00 5.99440634e-02 7.14023590e-01 7.89727807e-01 -5.40751398e-01 -1.29562151e+00 -9.09657180e-01 6.08723879e-01 2.01387759e-02 3.28508973e-01 -2.45499704e-02 -1.10043466e+00 6.73445106e-01 5.23762047e-01 7.01221943e-01 2.32534930e-01 4.17335704e-02 -6.71877205e-01 -3.66862714e-01 -1.22789991e+00 4.58729655e-01 1.11701536e+00 -7.55878568e-01 -2.39240602e-01 3.34952772e-01 4.63311672e-01 -1.71968162e-01 -7.82810509e-01 4.87754732e-01 2.95458496e-01 -1.42252898e+00 1.36675644e+00 -1.54386595e-01 5.91988981e-01 -5.98990262e-01 -2.95391649e-01 -1.44701087e+00 -7.04066694e-01 -2.97185570e-01 8.35143477e-02 9.87354577e-01 3.68281193e-02 -7.96066105e-01 4.90165085e-01 4.91338134e-01 -6.98114753e-01 -5.52370071e-01 -1.12950683e+00 -6.06344044e-01 -5.25722373e-03 -1.34946585e-01 2.70717800e-01 6.46042526e-01 -2.51322776e-01 1.24214925e-01 -5.43931961e-01 3.18487853e-01 8.35147262e-01 2.34584138e-01 2.78066516e-01 -9.10577953e-01 -3.73828739e-01 -3.56640726e-01 -3.77471417e-01 -1.30272257e+00 2.52389759e-02 -7.63189495e-01 2.63226420e-01 -1.89545679e+00 3.18760097e-01 -2.87177503e-01 -4.42387044e-01 8.18455935e-01 -8.32117908e-03 9.29293811e-01 1.39461994e-01 3.58994544e-01 -2.29988813e-01 6.60136461e-01 1.52021766e+00 -2.49922231e-01 1.02916114e-01 -3.17703694e-01 -7.58423686e-01 4.73653853e-01 7.53704309e-01 -1.99398533e-01 -3.15208614e-01 -8.81987989e-01 -1.74545214e-01 2.41280831e-02 8.67629826e-01 -1.24932277e+00 2.81548500e-01 8.47726837e-02 6.57727420e-01 -3.69167864e-01 6.10372066e-01 -7.24843442e-01 1.86673418e-01 2.27146640e-01 -2.20145196e-01 -5.72156049e-02 3.08931679e-01 4.96935219e-01 -4.50818807e-01 2.25957438e-01 1.13735569e+00 -2.10628688e-01 -6.51859581e-01 2.43746191e-01 -3.98755878e-01 -4.22747850e-01 4.94505167e-01 -3.65075201e-01 -6.30884409e-01 -2.97387123e-01 -8.11327040e-01 -2.16981575e-01 8.42799008e-01 3.80593240e-01 8.98192167e-01 -1.22193968e+00 -7.14735210e-01 6.02618873e-01 -1.32939279e-01 -1.89374492e-01 6.76242411e-01 8.25747311e-01 -4.58913684e-01 1.31221026e-01 -7.88226664e-01 -6.72135293e-01 -1.18115938e+00 4.98395979e-01 6.44379497e-01 -2.61074126e-01 -1.17884278e+00 7.77868807e-01 5.15764594e-01 -1.44433841e-01 8.98585692e-02 -1.75397024e-01 8.55097361e-03 -4.23233032e-01 7.60222852e-01 -4.05775523e-03 4.32596177e-01 -6.95834994e-01 -2.61336952e-01 4.32107329e-01 -2.00000465e-01 -6.08507171e-02 1.66164887e+00 -3.26271504e-01 -5.62310159e-01 -1.67720124e-01 1.07739508e+00 -3.26026469e-01 -1.80340338e+00 -3.35598826e-01 -4.29435730e-01 -7.45630741e-01 6.67700410e-01 -1.14233339e+00 -1.45053995e+00 9.75344479e-01 7.64694095e-01 1.83066819e-02 1.71135664e+00 -2.44948134e-01 7.07873940e-01 -5.46066724e-02 3.27751011e-01 -6.60810828e-01 3.32754344e-01 3.62490743e-01 1.38058019e+00 -1.10411167e+00 1.19905800e-01 -4.00878817e-01 -2.27359861e-01 9.81512427e-01 5.86802840e-01 -3.59028339e-01 6.34896219e-01 3.76836717e-01 1.04620270e-01 -1.63086906e-01 -7.35802531e-01 -2.18011513e-01 3.20708245e-01 9.60806072e-01 5.42431712e-01 -2.43023351e-01 1.11478940e-01 4.02817316e-02 2.14338183e-01 -9.46587250e-02 6.46594346e-01 7.24604189e-01 -5.06198764e-01 -1.11820281e+00 -4.28811908e-01 4.36816275e-01 -5.39782286e-01 -4.26945418e-01 1.33887038e-01 3.24183971e-01 4.88430634e-02 1.26217341e+00 7.32098892e-02 -1.63970694e-01 2.11321533e-01 -4.36368883e-01 6.44014060e-01 -4.88505065e-01 -7.84062982e-01 2.98674285e-01 -1.01327918e-01 -9.14276838e-01 -4.72708225e-01 -4.71009880e-01 -1.02241087e+00 -2.07018554e-01 1.64069477e-02 -3.20834607e-01 5.16278982e-01 7.37818003e-01 6.45321548e-01 8.19893360e-01 3.62434387e-01 -1.21586394e+00 -2.81556845e-01 -9.93737459e-01 -6.05540037e-01 3.98345232e-01 7.22481430e-01 -3.59373599e-01 -2.62451530e-01 1.65633798e-01]
[11.169465065002441, -2.1930179595947266]
95c9f7d4-a17c-4e10-bf0c-c7d92f32fc33
retrieve-caption-generate-visual-grounding
2109.03892
null
https://arxiv.org/abs/2109.03892v3
https://arxiv.org/pdf/2109.03892v3.pdf
Retrieve, Caption, Generate: Visual Grounding for Enhancing Commonsense in Text Generation Models
We investigate the use of multimodal information contained in images as an effective method for enhancing the commonsense of Transformer models for text generation. We perform experiments using BART and T5 on concept-to-text generation, specifically the task of generative commonsense reasoning, or CommonGen. We call our approach VisCTG: Visually Grounded Concept-to-Text Generation. VisCTG involves captioning images representing appropriate everyday scenarios, and using these captions to enrich and steer the generation process. Comprehensive evaluation and analysis demonstrate that VisCTG noticeably improves model performance while successfully addressing several issues of the baseline generations, including poor commonsense, fluency, and specificity.
['Varun Gangal', 'Eduard Hovy', 'Teruko Mitamura', 'Malihe Alikhani', 'Zhuofu Tao', 'Kevin Lu', 'Steven Y. Feng']
2021-09-08
null
https://openreview.net/forum?id=MvFv92fUFof
https://openreview.net/pdf?id=MvFv92fUFof
akbc-workshop-cskb-2021-10
['concept-to-text-generation']
['natural-language-processing']
[ 6.48062766e-01 5.88567555e-01 2.60674894e-01 -1.44381836e-01 -7.33399153e-01 -5.93928337e-01 1.28800154e+00 -2.11230084e-01 1.47101702e-02 9.10104156e-01 9.14008439e-01 -3.03609729e-01 4.75922137e-01 -7.26397097e-01 -7.17191517e-01 -2.11945370e-01 6.05676234e-01 8.35973918e-01 -3.64769101e-01 -6.11007869e-01 3.34740996e-01 8.52745548e-02 -1.22842038e+00 9.13278461e-01 1.20829034e+00 5.05873680e-01 2.60303319e-01 7.29576468e-01 -4.16201413e-01 1.26371717e+00 -1.02141798e+00 -9.87754226e-01 -2.70533174e-01 -1.12094080e+00 -9.74508822e-01 1.85625374e-01 4.67143387e-01 -5.22612572e-01 -1.96149617e-01 9.63381767e-01 6.31488085e-01 2.81232953e-01 9.03153539e-01 -1.49596477e+00 -1.62011564e+00 1.28256416e+00 -1.55200586e-01 5.03608733e-02 1.01668036e+00 5.90769231e-01 7.71580815e-01 -1.08678973e+00 1.22878742e+00 1.94763327e+00 3.66594613e-01 9.62360799e-01 -1.41807115e+00 -6.38949335e-01 7.45604793e-03 5.01532331e-02 -9.83956099e-01 -7.53031433e-01 8.39060783e-01 -4.37986940e-01 8.19467008e-01 3.10093850e-01 1.01041162e+00 1.89559376e+00 -1.77273139e-01 1.00714576e+00 1.27847683e+00 -5.46974659e-01 1.32733747e-01 4.28892262e-02 -4.26421434e-01 6.28558636e-01 1.12514108e-01 2.65833557e-01 -8.40618849e-01 4.21715714e-02 1.00990617e+00 -5.85148215e-01 -3.34819019e-01 1.01085305e-01 -1.78366470e+00 9.78747189e-01 5.76056659e-01 -3.03073656e-02 -4.17410046e-01 6.30921841e-01 3.08851749e-01 -1.16614997e-01 4.50166047e-01 8.76313984e-01 4.06624585e-01 -2.26596072e-01 -7.20761359e-01 6.11936808e-01 4.31352556e-01 1.29455888e+00 1.72835395e-01 5.28460503e-01 -1.06879318e+00 7.94932604e-01 2.58066565e-01 7.70542383e-01 3.79158825e-01 -1.00596988e+00 7.67360032e-01 4.87888217e-01 1.89775065e-01 -8.79087806e-01 1.07362792e-01 -4.72315438e-02 -5.32573938e-01 -7.18192458e-02 2.14940533e-01 -3.54293078e-01 -1.20177460e+00 1.89342058e+00 -6.55173734e-02 -2.92783380e-01 2.80779541e-01 7.49468446e-01 1.13707304e+00 7.34582126e-01 7.48562157e-01 1.54869452e-01 1.17875028e+00 -7.65733957e-01 -1.05850244e+00 -5.29581726e-01 4.04451609e-01 -7.42506027e-01 1.45395362e+00 -1.76164404e-01 -1.47524452e+00 -3.16998005e-01 -6.96596980e-01 -2.89251059e-01 -5.32515228e-01 1.58920884e-01 6.36376858e-01 2.59559780e-01 -1.38079000e+00 3.12183857e-01 -1.42851368e-01 -4.11580473e-01 6.82137311e-01 -4.91193414e-01 -1.49184182e-01 -1.78726628e-01 -1.32402039e+00 1.30505860e+00 6.54308259e-01 -8.11225250e-02 -1.11123359e+00 -6.22867048e-01 -1.30656600e+00 -1.98807836e-01 -2.68847793e-02 -1.33842742e+00 1.47254229e+00 -9.93782759e-01 -1.25564170e+00 1.03397024e+00 3.42730917e-02 -3.52276713e-01 6.89666331e-01 -1.52221143e-01 -3.30549061e-01 3.24889451e-01 3.92625511e-01 1.59075427e+00 8.22676599e-01 -1.66592240e+00 -3.87337320e-02 8.28429684e-02 8.70927572e-02 6.11155391e-01 1.32900730e-01 -7.42760748e-02 2.63119247e-02 -1.04111874e+00 -2.23698303e-01 -6.96913838e-01 3.68575891e-03 -8.09206516e-02 -8.54662895e-01 -1.08428709e-01 6.26138806e-01 -9.80178714e-01 7.88430989e-01 -1.86776853e+00 3.53060007e-01 -1.89843461e-01 1.30748048e-01 -1.67424336e-01 -4.35608894e-01 6.28759146e-01 -1.45073637e-01 4.92576361e-01 -2.44052276e-01 -4.02983040e-01 3.57113272e-01 -1.16339393e-01 -6.20418727e-01 -3.36612940e-01 4.18434262e-01 1.67140245e+00 -1.22678030e+00 -8.89276803e-01 3.31143171e-01 6.11834228e-01 -2.89021760e-01 9.92973521e-02 -8.00558627e-01 2.71524489e-01 -1.18772186e-01 6.15336537e-01 2.63724864e-01 -2.62080967e-01 -6.67104572e-02 -1.96969479e-01 3.34593832e-01 2.43650377e-01 -3.74961466e-01 1.84642649e+00 -2.86950380e-01 1.00260699e+00 -5.86090147e-01 -1.18410531e-02 6.27838612e-01 2.46673673e-01 -3.80799174e-01 -8.06816459e-01 1.21911526e-01 -1.71478152e-01 -2.09827304e-01 -2.92204738e-01 8.79705131e-01 -6.15374684e-01 -1.44672394e-01 5.54184496e-01 -1.13097914e-01 -9.30157959e-01 3.04292947e-01 7.86733687e-01 5.23299336e-01 4.44366008e-01 1.97057322e-01 -3.50027680e-02 -2.55349368e-01 4.17300373e-01 -4.08228219e-01 6.81128025e-01 3.22818607e-02 6.53712988e-01 3.08657467e-01 7.35836774e-02 -1.07287872e+00 -1.48599720e+00 3.33137810e-01 8.98084521e-01 9.49434340e-02 -3.00851256e-01 -9.40838099e-01 -5.23268044e-01 -2.46594921e-01 1.83489156e+00 -8.90413702e-01 -4.77625757e-01 -1.50182262e-01 -3.44250560e-01 7.51446784e-01 7.90777504e-01 6.72090650e-01 -1.69161546e+00 -7.70260990e-01 1.72677904e-01 -8.93123388e-01 -1.23160601e+00 -4.88791138e-01 -5.20852983e-01 -6.53635979e-01 -6.42428696e-01 -1.00773692e+00 -5.81081867e-01 7.20331252e-01 6.36704937e-02 1.69490635e+00 2.24260278e-02 -2.87962139e-01 6.51510537e-01 -4.73999143e-01 -5.73142171e-01 -9.38352108e-01 -6.19199336e-01 -6.28431559e-01 -5.93479931e-01 -2.46780142e-01 -2.46060953e-01 -4.03282583e-01 -2.03747213e-01 -9.94173467e-01 1.03979003e+00 4.15520549e-01 7.13633537e-01 1.37908429e-01 -6.65267944e-01 5.52964211e-01 -4.94210094e-01 1.32066190e+00 -1.98990017e-01 1.33972406e-01 3.56534481e-01 -2.45238453e-01 -8.64867773e-03 6.92015290e-02 -5.02422571e-01 -1.55965722e+00 -4.01632905e-01 2.71829396e-01 -4.28219974e-01 -2.21712813e-02 3.01088065e-01 -6.65385202e-02 2.57448494e-01 8.81306291e-01 5.27265310e-01 -1.54878303e-01 1.62764952e-01 1.24768984e+00 1.70016050e-01 8.82141590e-01 -8.58951867e-01 7.48313606e-01 2.52568215e-01 -2.49107480e-01 -4.78468120e-01 -6.60630643e-01 4.91607189e-01 -1.32418916e-01 -4.49574113e-01 1.13206315e+00 -1.01755404e+00 -2.53598690e-01 2.82954484e-01 -1.63478506e+00 -6.62126780e-01 -6.94375098e-01 -8.25349614e-02 -8.48750472e-01 3.70348878e-02 -5.99740267e-01 -7.29558766e-01 -5.57029247e-01 -8.42530668e-01 1.24430287e+00 1.13458864e-01 -6.22900307e-01 -1.13513231e+00 -3.85003276e-02 4.98160154e-01 4.76921588e-01 8.32874775e-01 1.23084307e+00 -3.27902347e-01 -4.82017875e-01 2.20057115e-01 -5.41219413e-01 -2.16998756e-02 -1.48714498e-01 1.72343720e-02 -9.51316535e-01 3.29367131e-01 -6.76952183e-01 -8.13079655e-01 8.39705229e-01 2.75480479e-01 1.01545155e+00 -5.77183127e-01 -3.23168367e-01 2.96002656e-01 1.06617618e+00 2.70846397e-01 1.07355821e+00 1.02593683e-01 7.61868238e-01 5.93080580e-01 6.04705930e-01 4.86635685e-01 7.34590590e-01 4.46453333e-01 3.13208431e-01 -2.43653446e-01 -7.53653228e-01 -8.87960613e-01 4.11029696e-01 5.69088124e-02 -2.02699587e-01 -6.38153672e-01 -9.09817219e-01 7.21255124e-01 -1.59956825e+00 -1.43594933e+00 1.94398519e-02 1.44169104e+00 1.20831871e+00 -2.74973214e-01 -1.19798794e-01 -2.41778031e-01 8.78018916e-01 1.39693558e-01 -2.16703370e-01 -5.43219149e-01 -4.04787302e-01 7.47974887e-02 -3.18676949e-01 5.73996842e-01 -6.48518443e-01 1.43007028e+00 7.46087742e+00 6.62303329e-01 -5.56645453e-01 -3.89709994e-02 7.06791878e-01 -9.77552608e-02 -9.75998878e-01 -1.10252500e-01 -2.55379826e-01 2.75152773e-01 5.75474322e-01 -4.12800074e-01 6.10626221e-01 4.69523758e-01 1.35715470e-01 -5.49043976e-02 -1.22305489e+00 1.07925737e+00 5.32633662e-01 -1.67017531e+00 9.92461205e-01 -2.00906411e-01 9.48650479e-01 -6.40690982e-01 3.18537325e-01 3.81926805e-01 7.38444567e-01 -1.31909835e+00 1.32992947e+00 6.42455578e-01 1.26668406e+00 -5.33778191e-01 2.66327560e-01 -1.89308479e-01 -7.67885268e-01 3.47121149e-01 4.96698953e-02 4.34834808e-02 4.42065448e-01 3.31762791e-01 -1.26709938e+00 1.69092223e-01 1.92279860e-01 4.64917660e-01 -8.70046318e-01 5.72391272e-01 -6.89356804e-01 3.07419688e-01 2.24934816e-01 -2.40644872e-01 1.31449282e-01 1.41674951e-01 6.94712877e-01 1.48252189e+00 4.56975430e-01 1.24107696e-01 -1.02667764e-01 1.69704342e+00 -2.58494973e-01 -2.11811021e-01 -9.23221111e-01 -6.91662252e-01 5.78540325e-01 8.34379017e-01 -7.32928276e-01 -7.78780878e-01 3.95845920e-01 1.50590932e+00 1.83611587e-01 6.51439071e-01 -1.02849388e+00 -2.63950586e-01 2.36645073e-01 -1.65957920e-02 -4.93410751e-02 -1.03429407e-02 -4.68052655e-01 -1.02794433e+00 -3.33537638e-01 -1.00910938e+00 1.12376474e-01 -1.97802162e+00 -1.18675375e+00 5.50966084e-01 4.98453200e-01 -6.53380156e-01 -6.09349012e-01 -3.43253583e-01 -7.19968021e-01 8.15582633e-01 -9.28880990e-01 -1.66649771e+00 -6.07560933e-01 7.44199574e-01 8.19220185e-01 8.30049142e-02 8.11060607e-01 -5.85325420e-01 -1.38065413e-01 3.73761028e-01 -5.95171273e-01 -1.30916744e-01 6.72929764e-01 -1.44634008e+00 8.39301109e-01 7.56166637e-01 7.40431473e-02 5.10838032e-01 8.91683757e-01 -1.17685544e+00 -8.46150041e-01 -1.00010490e+00 7.68703282e-01 -7.72575915e-01 4.60653245e-01 -4.32357311e-01 -3.34627092e-01 8.88046920e-01 8.11476529e-01 -7.11922944e-01 5.23178697e-01 -3.68896365e-01 -4.25480932e-01 7.57355988e-01 -1.34768784e+00 1.30857599e+00 1.47039962e+00 -5.57375371e-01 -1.00303721e+00 5.35739064e-01 1.14664459e+00 -5.48139036e-01 -3.99026304e-01 -1.24966115e-01 1.51425377e-01 -6.60744190e-01 1.11704373e+00 -7.06688881e-01 1.26115191e+00 -7.12350383e-02 -1.94368362e-01 -1.87810814e+00 -3.31929356e-01 -8.65917206e-01 -1.34614483e-01 1.27647245e+00 4.65054899e-01 -2.33676866e-01 1.45496026e-01 6.87408328e-01 -2.50179350e-01 -8.04546252e-02 -5.85998118e-01 -4.97768968e-01 9.68381688e-02 -3.64731044e-01 7.63672829e-01 1.09065652e+00 2.79784948e-01 6.40107870e-01 -1.08928837e-01 -4.61948961e-01 6.30928278e-01 -8.41992870e-02 5.94886780e-01 -7.30363607e-01 8.33761543e-02 -5.97130299e-01 -4.50991914e-02 -5.40559590e-01 2.30660960e-01 -9.68868911e-01 2.93052673e-01 -2.24352837e+00 5.19845307e-01 1.40486896e-01 4.95149612e-01 5.65631747e-01 -4.10956413e-01 3.67969275e-01 7.21810400e-01 6.90147653e-03 -4.82596010e-01 8.18734467e-01 2.01399469e+00 -3.07396561e-01 4.41700667e-02 -9.00435269e-01 -1.22354317e+00 5.04377246e-01 7.84524322e-01 1.09115586e-01 -7.06952691e-01 -5.62259674e-01 2.94266909e-01 9.16936845e-02 1.08052564e+00 -7.11706400e-01 -3.55389208e-01 -4.52889115e-01 9.71777380e-01 -6.04642212e-01 6.47873938e-01 -1.87034398e-01 2.13619143e-01 2.48623297e-01 -7.29864419e-01 3.54538679e-01 5.80637634e-01 3.77135545e-01 1.55845895e-01 1.70891974e-02 5.47426879e-01 -3.75901699e-01 -6.92203045e-01 -3.44947070e-01 -2.85535276e-01 5.51127791e-01 9.28009868e-01 -3.47082824e-01 -9.32149112e-01 -9.15949821e-01 -7.04654872e-01 5.70270084e-02 6.28886461e-01 4.22049135e-01 1.08542514e+00 -1.78434467e+00 -1.00886083e+00 -4.50517476e-01 3.13613981e-01 -3.28224480e-01 1.19160466e-01 3.46751243e-01 -3.91389430e-01 3.51812541e-01 -3.92464727e-01 -2.82899439e-01 -8.51284742e-01 4.72431272e-01 3.08744192e-01 2.27858704e-02 -7.07362652e-01 1.02000391e+00 3.53877187e-01 -1.25541389e-02 -2.61705041e-01 -3.44707221e-01 -1.86221171e-02 -5.21275289e-02 5.04462540e-01 3.53293031e-01 -6.36427522e-01 -5.65812111e-01 -8.18652362e-02 4.47813757e-02 1.00644827e-01 -9.14343417e-01 7.33963251e-01 -9.67940018e-02 -8.93240236e-03 2.62059152e-01 4.73934650e-01 -2.87679076e-01 -1.17322040e+00 3.32509130e-01 -4.15573299e-01 -3.32160622e-01 -1.45669267e-01 -1.83706295e+00 -5.55649340e-01 7.89425731e-01 1.46639049e-01 -1.54492289e-01 8.25766027e-01 2.17199072e-01 6.59498394e-01 2.27236629e-01 1.52389005e-01 -8.97777081e-01 5.93672335e-01 4.16594267e-01 1.90701497e+00 -1.05864275e+00 -1.84565753e-01 -2.99099654e-01 -1.33706343e+00 7.82926023e-01 8.39492381e-01 3.08980316e-01 -3.24131489e-01 -2.55230051e-02 1.09201139e-02 -3.43091726e-01 -7.97134042e-01 -2.50968426e-01 4.79481697e-01 1.18462205e+00 5.51476896e-01 4.03030097e-01 -1.79020427e-02 4.16400433e-01 -6.89107001e-01 2.56527327e-02 5.06444633e-01 7.62017787e-01 -2.23262474e-01 -3.81693512e-01 -6.36900604e-01 2.43641302e-01 2.94800043e-01 -6.53832614e-01 -1.22870708e+00 9.21905637e-01 2.08023898e-02 1.08063972e+00 -2.74954643e-03 -1.98579162e-01 2.39387840e-01 4.33735549e-01 9.45429802e-01 -6.86199486e-01 -5.95607162e-01 -2.24450931e-01 6.43669248e-01 -5.45843065e-01 -2.54293859e-01 -5.12617290e-01 -1.40244734e+00 -2.97757149e-01 -3.36536951e-03 -3.35018575e-01 5.74195862e-01 9.15692627e-01 3.03267509e-01 7.16443777e-01 -1.88710734e-01 -9.51625645e-01 -2.66340345e-01 -1.24526989e+00 -2.50637159e-03 8.08576882e-01 8.86291638e-02 -6.80657923e-01 -2.27818519e-01 5.05149603e-01]
[11.178927421569824, 0.9095774292945862]
6d9c0d8b-1574-4dff-add7-5c7b5e88a1fe
harnessing-pre-trained-neural-networks-with
null
null
https://aclanthology.org/D19-1365
https://aclanthology.org/D19-1365.pdf
Harnessing Pre-Trained Neural Networks with Rules for Formality Style Transfer
Formality text style transfer plays an important role in various NLP applications, such as non-native speaker assistants and child education. Early studies normalize informal sentences with rules, before statistical and neural models become a prevailing method in the field. While a rule-based system is still a common preprocessing step for formality style transfer in the neural era, it could introduce noise if we use the rules in a naive way such as data preprocessing. To mitigate this problem, we study how to harness rules into a state-of-the-art neural network that is typically pretrained on massive corpora. We propose three fine-tuning methods in this paper and achieve a new state-of-the-art on benchmark datasets
['WenHan Chao', 'Yunli Wang', 'Lili Mou', 'Zhoujun Li', 'Yu Wu']
2019-11-01
null
null
null
ijcnlp-2019-11
['formality-style-transfer']
['natural-language-processing']
[ 5.14820397e-01 4.41296279e-01 -2.49108046e-01 -7.56821394e-01 -1.86668023e-01 -5.54007828e-01 6.04566574e-01 3.96010578e-02 -7.71970570e-01 9.58207428e-01 3.10105920e-01 -6.47484362e-01 3.72832865e-02 -9.32376027e-01 -8.54999304e-01 -1.72963142e-01 4.31092590e-01 4.98029381e-01 3.93533707e-01 -6.53811693e-01 1.97134707e-02 2.53221810e-01 -1.30847681e+00 5.71006060e-01 1.18185806e+00 6.92063570e-01 -2.75786400e-01 3.81321222e-01 -6.91655278e-01 7.99075365e-01 -6.02720559e-01 -9.14971054e-01 2.32261345e-01 -3.66007864e-01 -8.94624949e-01 -5.75993299e-01 5.07303119e-01 -1.51523322e-01 -3.66697103e-01 1.37385857e+00 2.88894087e-01 1.96629301e-01 6.22194052e-01 -8.25543940e-01 -9.69655335e-01 1.37388623e+00 -4.25718367e-01 1.56562015e-01 9.07270238e-02 -2.11569499e-02 7.38151789e-01 -6.38992071e-01 5.60917139e-01 1.44287336e+00 6.27400100e-01 1.04117846e+00 -1.42911398e+00 -8.53275597e-01 2.36375749e-01 1.07925281e-01 -1.16689754e+00 -4.03235644e-01 9.53416944e-01 -3.01422417e-01 7.97433555e-01 2.34417781e-01 5.30161500e-01 1.38272858e+00 -1.40601033e-02 8.79137099e-01 1.19946182e+00 -8.51087987e-01 3.83030921e-02 2.36181259e-01 4.28511083e-01 4.25343871e-01 3.70590895e-01 7.67507255e-02 -6.18341386e-01 2.61658549e-01 7.12049067e-01 -1.76067770e-01 -1.54168874e-01 4.77599055e-02 -9.89562750e-01 1.03286040e+00 2.66232193e-01 7.43899047e-01 -1.16829090e-01 -1.17474131e-01 5.29145956e-01 5.56845367e-01 5.37236571e-01 6.22885883e-01 -8.32389355e-01 -2.97394276e-01 -9.84482884e-01 3.12908053e-01 8.42265725e-01 8.92805636e-01 5.97334623e-01 9.92540419e-02 -2.10334718e-01 1.06545627e+00 8.78663547e-03 1.54237926e-01 8.11974704e-01 -5.79475403e-01 6.00302339e-01 7.80205131e-01 -4.60292459e-01 -5.43358922e-01 -1.36898011e-01 -4.69810218e-01 -1.18546784e+00 -9.43308603e-03 6.39868677e-01 -2.88893551e-01 -9.44590151e-01 1.92246413e+00 2.42040798e-01 -1.75491467e-01 4.69655804e-02 4.87803042e-01 8.10072482e-01 7.80169010e-01 9.50282440e-02 -3.17382485e-01 1.15043616e+00 -9.64958847e-01 -7.53980160e-01 -4.76180077e-01 4.88759041e-01 -5.71412385e-01 1.49513388e+00 7.22696722e-01 -9.74976003e-01 -5.06858170e-01 -1.00274038e+00 -3.37240905e-01 -6.67594016e-01 -3.83812875e-01 6.14121735e-01 8.27067137e-01 -9.24919963e-01 8.72801125e-01 -5.99100828e-01 -6.07238352e-01 5.21623433e-01 3.06468338e-01 -3.75097632e-01 1.14204325e-01 -1.38802230e+00 1.05322909e+00 7.45982528e-01 -1.10395111e-01 -2.69948334e-01 -1.00232232e+00 -7.65761614e-01 5.15090786e-02 4.42725897e-01 -6.05445027e-01 1.49000990e+00 -1.12781560e+00 -2.11552930e+00 1.07226229e+00 5.12301400e-02 -5.34485579e-01 5.08258581e-01 -5.75052023e-01 -4.06857133e-01 -3.86673331e-01 -2.46293917e-01 5.97410321e-01 7.42952645e-01 -8.58947992e-01 -5.87216675e-01 -3.06259781e-01 2.32726306e-01 -6.24710321e-02 -7.76280940e-01 2.30457097e-01 -1.99502721e-01 -9.23333585e-01 -5.63508123e-02 -6.94978654e-01 -5.16211390e-02 -3.60212564e-01 -2.18571246e-01 -7.42056727e-01 4.64105487e-01 -5.66632390e-01 1.40558028e+00 -1.88312483e+00 6.89228326e-02 8.04424956e-02 5.60398363e-02 7.04537690e-01 4.82500121e-02 1.70599803e-01 -4.27127555e-02 3.51781964e-01 -1.56146705e-01 -1.80554256e-01 2.13194102e-01 1.85238287e-01 -4.00882989e-01 -1.03132412e-01 1.88908413e-01 8.94054651e-01 -7.98921466e-01 -4.96388972e-01 -5.15462011e-02 3.32135201e-01 -8.88982356e-01 4.57732558e-01 -3.39695245e-01 2.77427554e-01 -2.60053784e-01 1.65738508e-01 4.22432065e-01 2.46352538e-01 1.90302238e-01 5.95359020e-02 -7.67135769e-02 7.47904539e-01 -8.81277919e-01 1.93811166e+00 -4.18902040e-01 4.44128305e-01 -2.36257240e-01 -1.24451971e+00 8.43505442e-01 2.80764669e-01 3.55389458e-03 -5.11169910e-01 3.86220247e-01 2.32432902e-01 4.45167422e-01 -2.65761733e-01 5.37552118e-01 -3.94086629e-01 -9.85532627e-03 2.88130224e-01 3.86649072e-01 -1.95213348e-01 5.01174986e-01 -8.34640041e-02 1.05054808e+00 4.77168202e-01 4.87252772e-01 -6.15705490e-01 5.54292798e-01 -1.97833836e-01 7.74797380e-01 6.65002108e-01 8.13776627e-03 4.68606532e-01 3.90297264e-01 -7.49276161e-01 -8.05052876e-01 -8.93355489e-01 -1.30595356e-01 1.53121543e+00 -6.13774538e-01 -3.29132080e-01 -1.10408759e+00 -9.32372630e-01 -2.76709437e-01 9.16939855e-01 -6.55534446e-01 -2.60149419e-01 -8.06106806e-01 -4.17232156e-01 6.58290207e-01 4.10467625e-01 7.29397178e-01 -1.36148679e+00 -3.39433551e-01 3.82344246e-01 6.02399139e-03 -1.05082023e+00 -4.79216367e-01 4.16632652e-01 -8.81689668e-01 -6.17800832e-01 -5.25324583e-01 -6.42155349e-01 5.22283792e-01 -3.61063719e-01 1.28141391e+00 1.33218691e-01 3.80779147e-01 -3.66065711e-01 -4.07381028e-01 -8.77971470e-01 -6.18745565e-01 7.22453296e-01 3.97589922e-01 -2.13596562e-04 7.78268099e-01 -9.48728979e-01 -2.82823086e-01 -1.63319752e-01 -1.03149939e+00 1.52849376e-01 5.72260559e-01 7.52344012e-01 1.82503685e-02 -2.36551970e-01 5.84773481e-01 -1.57905543e+00 8.36541414e-01 8.61565620e-02 -6.32638395e-01 2.93085217e-01 -5.23823977e-01 3.85464817e-01 9.77923095e-01 -6.40848815e-01 -1.17215180e+00 -1.21182799e-01 -3.13866884e-01 -2.09138364e-01 -5.75036824e-01 6.53478503e-01 -4.15225625e-01 1.18682593e-01 9.57090795e-01 1.35164946e-01 -3.12514782e-01 -5.60536385e-01 5.33369720e-01 8.46593440e-01 5.70633531e-01 -1.10101426e+00 1.06511259e+00 -2.88444787e-01 -3.12768579e-01 -9.26308513e-01 -1.19982052e+00 1.61637389e-03 -7.32425511e-01 1.28929317e-01 7.62513876e-01 -4.29522038e-01 -4.70641017e-01 3.03650588e-01 -1.59489453e+00 -5.28753281e-01 -4.30549115e-01 2.75465071e-01 -1.51158661e-01 5.05171455e-02 -5.75720668e-01 -4.66540813e-01 -5.59715271e-01 -8.88891101e-01 3.34926099e-01 4.68278557e-01 -5.97846270e-01 -9.86740828e-01 2.24748552e-01 2.82675415e-01 7.27535367e-01 -2.92957220e-02 1.08123767e+00 -9.55189824e-01 -4.10208181e-02 1.04712456e-01 2.20001657e-02 6.53217912e-01 6.22754693e-02 -7.17246300e-03 -1.19348288e+00 9.84144732e-02 -1.93257377e-01 -4.22897816e-01 6.96695983e-01 -7.07207993e-02 1.23466468e+00 -5.65434933e-01 -7.09246192e-03 7.25270152e-01 1.10396457e+00 2.12051734e-01 7.75422871e-01 2.39755228e-01 7.68770039e-01 8.09935451e-01 1.18878810e-03 -2.16917142e-01 3.31156671e-01 4.07718956e-01 -1.97130427e-01 8.49725530e-02 -5.78016452e-02 -6.18271112e-01 3.76381248e-01 1.33224332e+00 -2.43835986e-01 9.35233384e-02 -1.03342748e+00 4.07463908e-01 -1.80134535e+00 -7.87019968e-01 2.45053470e-01 1.95182097e+00 1.78514397e+00 6.01678610e-01 1.73677430e-02 3.94584239e-01 5.43769896e-01 -3.53131481e-02 -1.09511696e-01 -8.81922007e-01 6.14774749e-02 6.79702997e-01 1.26132354e-01 5.04042447e-01 -9.79994059e-01 1.42848289e+00 5.67538118e+00 9.10601497e-01 -1.39957309e+00 8.72835070e-02 5.76577306e-01 1.60373583e-01 -3.82785797e-01 -1.44771561e-01 -1.05403161e+00 3.45724404e-01 9.55023050e-01 -2.22646836e-02 5.79197526e-01 7.49361634e-01 3.07894424e-02 1.99022427e-01 -1.65134740e+00 9.39707100e-01 5.80073148e-02 -1.24937737e+00 2.85010070e-01 -1.24050461e-01 8.11327100e-01 -5.69587760e-02 -3.04373145e-01 8.62661362e-01 5.18121302e-01 -1.10360444e+00 7.64116943e-01 3.31106335e-01 7.63937414e-01 -5.70882678e-01 5.87191701e-01 6.08163536e-01 -6.03180885e-01 3.02810490e-01 -4.24613684e-01 -5.13373852e-01 -1.55922070e-01 7.56603420e-01 -6.03647470e-01 3.25965434e-01 7.46759176e-01 3.56509686e-01 -6.16147697e-01 5.90387881e-01 -5.60517013e-01 9.94117200e-01 -5.23131251e-01 -3.70483041e-01 2.14596838e-02 -1.84127733e-01 2.22804174e-01 1.35769534e+00 6.08193129e-02 2.63315320e-01 -2.53931843e-02 8.04426789e-01 -5.44951499e-01 2.33911827e-01 -9.37996626e-01 -1.98309749e-01 2.47583792e-01 1.08203650e+00 -5.73953390e-01 -2.96185613e-01 -5.23547292e-01 5.82875967e-01 8.25654924e-01 3.88977140e-01 -5.05407274e-01 -7.04354405e-01 5.52734256e-01 2.35116482e-01 5.19790314e-02 -1.31239131e-01 -7.17978895e-01 -1.30088878e+00 5.26398420e-02 -1.20964921e+00 1.75241664e-01 -3.98806214e-01 -1.34886503e+00 5.04806757e-01 1.30495787e-01 -7.09619105e-01 -2.49634162e-01 -8.68106961e-01 -7.01101065e-01 5.47244489e-01 -1.42505026e+00 -1.02365494e+00 1.60608247e-01 5.54477692e-01 5.86187124e-01 -3.56263429e-01 8.81691694e-01 2.97820270e-01 -5.72719574e-01 9.83309448e-01 -2.25663871e-01 5.63957334e-01 7.98175514e-01 -1.33606446e+00 3.19720000e-01 1.15405524e+00 3.21971774e-01 1.09414828e+00 7.60289073e-01 -4.40937698e-01 -1.10603869e+00 -1.02579784e+00 1.12595880e+00 -4.08677220e-01 7.26131022e-01 -6.44070923e-01 -1.05010748e+00 7.56223440e-01 7.24159122e-01 8.11313912e-02 4.98976827e-01 5.22641063e-01 -3.52207065e-01 -5.26846468e-01 -8.94683182e-01 9.20558512e-01 1.25616121e+00 -3.43880862e-01 -1.31993341e+00 -1.99941501e-01 1.01119995e+00 -4.44688678e-01 -6.54834867e-01 6.01842582e-01 5.71263015e-01 -7.62300253e-01 4.61214662e-01 -7.97371805e-01 6.74289107e-01 -6.88468218e-02 1.28851444e-01 -1.70576000e+00 -3.79958421e-01 -1.07592118e+00 3.69910449e-02 1.85354483e+00 7.55850911e-01 -6.59482002e-01 7.41214514e-01 5.44458151e-01 -8.61294121e-02 -6.64784431e-01 -8.53336573e-01 -6.19660795e-01 7.09681511e-01 -5.68228960e-01 7.37298906e-01 1.08296967e+00 1.82989612e-01 9.40429032e-01 2.11894438e-02 -3.09069037e-01 3.54311794e-01 -2.51868129e-01 9.00577009e-01 -1.54524565e+00 -1.68002814e-01 -7.10359275e-01 -1.27519473e-01 -9.22499776e-01 4.66058195e-01 -1.08174372e+00 1.19029947e-01 -1.29473937e+00 -8.29925165e-02 -1.55356169e-01 -3.43945861e-01 6.90004230e-01 -1.87288225e-01 1.41277865e-01 1.23907603e-01 -5.36837041e-01 -3.69940370e-01 5.49020350e-01 1.03604865e+00 -3.51263195e-01 -2.39021227e-01 -2.30523556e-01 -9.90875900e-01 1.09319937e+00 9.28327560e-01 -6.19773030e-01 -4.97399211e-01 -7.37871587e-01 4.93668765e-01 -5.23241043e-01 -1.84394151e-01 -1.05438852e+00 2.41506681e-01 -3.63971829e-01 2.81785250e-01 -1.06632665e-01 -2.50591636e-01 -8.39011431e-01 -4.23979193e-01 1.86970681e-01 -4.47518677e-01 -7.46892840e-02 3.05942535e-01 -7.92309567e-02 -2.94220448e-01 -2.60439217e-01 9.53671455e-01 -1.27481386e-01 -1.90343916e-01 2.04112202e-01 -1.99548975e-01 5.16084254e-01 4.97404724e-01 -3.55181545e-02 -3.82215619e-01 -6.25077859e-02 -1.69242486e-01 1.14862375e-01 9.45490897e-02 6.87417448e-01 3.13954562e-01 -1.25520265e+00 -7.29446650e-01 4.30190414e-01 5.48761338e-02 4.10190254e-01 -3.25821042e-01 4.92445737e-01 -5.32277465e-01 4.08906817e-01 -1.66958764e-01 -3.92394662e-01 -8.53188515e-01 2.99849570e-01 1.72815993e-01 -5.76289952e-01 -4.09852147e-01 8.38167548e-01 3.95192593e-01 -9.45093155e-01 4.00670320e-01 -8.21878076e-01 -3.64538401e-01 -2.07372025e-01 5.17787874e-01 -2.53653340e-02 1.97197273e-01 -2.00858757e-01 -2.16251403e-01 5.19189298e-01 -3.35477144e-01 -2.31697768e-01 1.45564175e+00 2.75462359e-01 -1.52028322e-01 7.62064934e-01 8.02466750e-01 -4.96096872e-02 -7.88843453e-01 -4.75061357e-01 2.32066333e-01 5.37516773e-02 2.56425869e-02 -8.09270442e-01 -7.16123760e-01 1.11080396e+00 1.89974472e-01 3.20661366e-01 8.76910448e-01 -3.19362551e-01 7.43889153e-01 8.13579381e-01 2.42613211e-01 -1.37292790e+00 -3.82746637e-01 1.17488015e+00 9.99807358e-01 -1.18846524e+00 -2.05836222e-01 -3.76547664e-01 -3.64688933e-01 1.11139405e+00 8.19341660e-01 -1.13304295e-01 8.52448344e-01 3.67234349e-01 2.18729228e-01 2.88371712e-01 -7.41873503e-01 -3.98738869e-03 3.67779255e-01 5.04189610e-01 1.03684998e+00 -1.46871107e-02 -6.97885752e-01 1.11815429e+00 -7.88929284e-01 4.08098102e-01 3.04733992e-01 8.30828130e-01 -3.19540113e-01 -1.43543136e+00 -1.75455779e-01 4.92419660e-01 -8.39233398e-01 -3.60393584e-01 -4.48147923e-01 7.05872297e-01 4.04720515e-01 7.15757132e-01 -2.39392281e-01 -2.71406591e-01 6.07173264e-01 6.22342348e-01 7.34371781e-01 -9.39120293e-01 -9.50654984e-01 -2.65733182e-01 2.04260424e-01 -3.19073319e-01 -4.10578310e-01 -3.31695706e-01 -1.17454004e+00 -3.64254206e-01 -8.08685496e-02 2.44524792e-01 4.01982725e-01 1.33548379e+00 -1.00363478e-01 5.67485452e-01 1.76381350e-01 -6.12911880e-01 -6.20058477e-01 -1.54935122e+00 -2.39284575e-01 5.05322397e-01 1.70006499e-01 -3.50229234e-01 -1.36964560e-01 2.46750057e-01]
[10.928632736206055, 9.069101333618164]
a6c79db8-2c76-425b-a19a-83da91b3a26d
findings-of-the-iwslt-2022-evaluation
null
null
https://aclanthology.org/2022.iwslt-1.10
https://aclanthology.org/2022.iwslt-1.10.pdf
Findings of the IWSLT 2022 Evaluation Campaign
The evaluation campaign of the 19th International Conference on Spoken Language Translation featured eight shared tasks: (i) Simultaneous speech translation, (ii) Offline speech translation, (iii) Speech to speech translation, (iv) Low-resource speech translation, (v) Multilingual speech translation, (vi) Dialect speech translation, (vii) Formality control for speech translation, (viii) Isometric speech translation. A total of 27 teams participated in at least one of the shared tasks. This paper details, for each shared task, the purpose of the task, the data that were released, the evaluation metrics that were applied, the submissions that were received and the results that were achieved.
['Shinji Watanabe', 'Changhan Wang', 'Alexander Waibel', 'Yogesh Virkar', 'Marco Turchi', 'Katsuhito Sudoh', 'Sebastian Stüker', 'Matthias Sperber', 'Jiatong Shi', 'Elizabeth Salesky', 'Juan Pino', 'John Ortega', 'Xing Niu', 'Jan Niehues', 'Matteo Negri', 'Satoshi Nakamura', 'Maria Nǎdejde', 'Kenton Murray', 'Paul McNamee', 'Prashant Mathur', 'Xutai Ma', 'Surafel Lakew', 'Vĕra Kloudová', 'Dávid Javorský', 'Benjamin Hsu', 'Barry Haddow', 'Roman Grundkiewicz', 'Hongyu Gong', 'Souhir Gahbiche', 'Christian Federmann', 'Marcello Federico', 'Yannick Estève', 'Clara Emmanuel', 'Maha Elbayad', 'Kevin Duh', 'Georgiana Dinu', 'Anna Currey', 'Roldano Cattoni', 'Ondřej Bojar', 'Marcely Zanon Boito', 'Luisa Bentivogli', 'Loïc Barrault', 'Antonios Anastasopoulos']
null
null
null
null
iwslt-acl-2022-5
['speech-to-speech-translation']
['speech']
[ 2.48660743e-01 1.68736994e-01 -1.06802642e-01 -5.18712997e-01 -1.77489710e+00 -8.07178140e-01 1.30780029e+00 7.12061897e-02 -4.09909576e-01 1.00410938e+00 6.22526467e-01 -8.61102045e-01 1.58186138e-01 1.44358014e-03 -6.58069015e-01 -4.40408438e-01 1.94950715e-01 1.03928339e+00 2.35281549e-02 -5.62363803e-01 1.13299742e-01 2.68517524e-01 -9.93487656e-01 4.79364157e-01 9.17739093e-01 6.32825613e-01 3.08521628e-01 8.37196648e-01 -3.74775082e-01 4.63062972e-01 -7.61411190e-01 -4.87753570e-01 -1.46792367e-01 -8.64631057e-01 -1.26635015e+00 1.09023765e-01 1.72676459e-01 5.42590730e-02 2.62485445e-02 6.08570397e-01 8.31386924e-01 3.31623733e-01 3.34014446e-01 -1.16934967e+00 -5.11844158e-01 1.08845568e+00 1.70332804e-01 3.13451797e-01 8.48766446e-01 1.47232354e-01 7.08686471e-01 -1.53297603e+00 1.00954330e+00 1.30291164e+00 2.66768068e-01 4.54643637e-01 -1.10691583e+00 -3.29987645e-01 -7.94367641e-02 -4.32477474e-01 -1.42398012e+00 -1.30893660e+00 -9.26921442e-02 -3.24484110e-01 1.77320242e+00 4.92515296e-01 4.42224920e-01 1.35514510e+00 5.52070960e-02 5.41523576e-01 1.46392345e+00 -7.88536370e-01 2.80812263e-01 6.79177701e-01 -1.17803939e-01 6.86001554e-02 -5.76637208e-01 1.01655804e-01 -8.18867266e-01 -2.75949955e-01 -4.99174092e-03 -9.93756890e-01 1.21061252e-02 3.15024257e-01 -1.87567747e+00 5.16601145e-01 -3.60907704e-01 7.11079299e-01 -2.96384662e-01 -4.56815243e-01 8.12776625e-01 9.01120484e-01 8.82705927e-01 3.24775577e-01 -7.67289281e-01 -5.20959735e-01 -8.95592451e-01 1.91921726e-01 1.24039853e+00 1.17448306e+00 5.54686189e-01 2.92875588e-01 -3.34052682e-01 1.06478834e+00 1.23851195e-01 9.89797592e-01 6.10762179e-01 -5.23639739e-01 1.16040695e+00 -5.31659462e-02 2.64867783e-01 -4.07418281e-01 -3.19842160e-01 -1.24166802e-01 -2.52982885e-01 -4.77587551e-01 -1.12903528e-02 -5.08322716e-01 -8.88765931e-01 1.47897720e+00 1.32241137e-02 -4.19312358e-01 5.09930551e-01 9.32504296e-01 1.08752823e+00 1.04037440e+00 -2.17202321e-01 -6.25048816e-01 1.13309860e+00 -1.14439702e+00 -9.99937892e-01 -2.87998825e-01 6.13209963e-01 -1.52916479e+00 1.14530981e+00 -4.67240624e-02 -1.36101472e+00 -3.78231198e-01 -8.16668749e-01 1.81344047e-01 -3.34696889e-01 2.17368811e-01 3.37809026e-01 6.55304074e-01 -1.75442088e+00 -3.70115251e-03 -8.09768379e-01 -8.14036250e-01 -5.26795089e-01 4.26892400e-01 -4.59022135e-01 3.33546966e-01 -1.51330459e+00 1.21855140e+00 -1.47852853e-01 -2.20440865e-01 -7.70361304e-01 -2.20061794e-01 -8.12070489e-01 -2.00846836e-01 1.03407703e-01 -5.51210165e-01 1.47635937e+00 -1.32121587e+00 -2.07595682e+00 1.34615982e+00 -3.42398494e-01 -1.94268957e-01 6.83085203e-01 9.46066454e-02 -5.78413188e-01 -2.86215097e-01 5.07006288e-01 4.70286280e-01 4.66110766e-01 -1.11563969e+00 -5.63546121e-01 -1.47603065e-01 -7.53483593e-01 5.62647760e-01 2.09461287e-01 1.07128131e+00 -3.53013545e-01 -7.42205679e-01 3.86483520e-02 -1.11885059e+00 1.46711603e-01 -8.53767216e-01 -4.71755445e-01 -2.91405231e-01 3.89790863e-01 -1.07568789e+00 9.50011969e-01 -2.14738917e+00 6.38527513e-01 -1.33946642e-01 -5.26896775e-01 2.63712674e-01 -3.46538007e-01 1.35612822e+00 1.15998030e-01 2.27319926e-01 -4.32611555e-02 -1.04250097e+00 6.33612871e-02 1.52237594e-01 -5.70090950e-01 1.49992049e-01 -6.48395047e-02 8.93524826e-01 -6.21863663e-01 -2.33055219e-01 -1.79847013e-02 3.38262737e-01 3.16818118e-01 3.51201773e-01 1.07081868e-01 5.23695946e-01 6.94087893e-02 6.29118681e-01 2.50357658e-01 2.25666910e-01 -1.01860669e-02 5.04049242e-01 -7.39657998e-01 1.23831093e+00 -7.54759073e-01 1.74783182e+00 -6.23648763e-01 6.51569963e-01 6.41263247e-01 -3.79334420e-01 1.09863281e+00 9.40085649e-01 1.93837613e-01 -6.70732558e-01 -1.71090394e-01 9.25312102e-01 -2.27010742e-01 -2.34985456e-01 6.23952150e-01 -8.72773379e-02 -4.77067530e-02 9.64940727e-01 3.38225454e-01 -3.22619945e-01 2.15537190e-01 2.64963746e-01 6.49487138e-01 -7.22150654e-02 2.48482719e-01 -4.79566634e-01 4.03493732e-01 1.49042174e-01 3.47439855e-01 5.01854539e-01 -3.29089433e-01 6.21837378e-01 1.77431956e-01 -2.37946153e-01 -1.12232625e+00 -9.70940471e-01 2.50171959e-01 1.38007939e+00 -1.75495520e-01 -4.47093815e-01 -8.82778168e-01 -2.71431714e-01 -4.55031335e-01 8.04982185e-01 -1.58814445e-01 8.92695561e-02 -5.28801978e-01 -1.10184483e-01 9.94532049e-01 1.61079884e-01 3.16531181e-01 -1.32133293e+00 2.05366746e-01 7.98876509e-02 -8.48593175e-01 -1.36305535e+00 -1.12994254e+00 1.52046561e-01 -4.36332494e-01 -3.02919924e-01 -8.98775399e-01 -1.18290305e+00 7.00125024e-02 2.84027040e-01 9.01525855e-01 -3.63226682e-01 3.06577384e-01 4.33819920e-01 -6.06556237e-01 -3.86728108e-01 -1.03122139e+00 4.74534541e-01 4.54888046e-01 -3.55503596e-02 7.59370849e-02 -8.47604498e-02 1.19138286e-01 5.85598230e-01 -3.63515437e-01 1.11747019e-01 6.45559728e-01 8.07504952e-01 3.27444464e-01 -9.12495077e-01 6.24123454e-01 -3.95469517e-01 9.59413648e-01 -1.31539032e-01 -3.00152212e-01 4.70266044e-01 -5.25029123e-01 -4.00721073e-01 1.78604230e-01 -1.75826654e-01 -9.91984725e-01 4.11442295e-02 -3.44165146e-01 1.51596099e-01 -4.78905290e-02 7.70252824e-01 -1.07045949e-01 1.71431139e-01 8.10724914e-01 7.29200423e-01 4.36462849e-01 -1.82842270e-01 1.43281192e-01 1.31534433e+00 1.36857837e-01 -4.70972359e-01 5.31354666e-01 -2.45606631e-01 -5.53835034e-01 -1.14757419e+00 -9.95651409e-02 -3.89695168e-01 -6.82820559e-01 -7.08825514e-02 1.03267360e+00 -1.15730143e+00 1.32922262e-01 6.59679353e-01 -1.47811699e+00 -7.63672888e-01 -2.29169372e-02 7.18047857e-01 -6.62456453e-01 -2.86546331e-02 -7.71868348e-01 -8.64232719e-01 -6.99985981e-01 -1.55875504e+00 1.47224343e+00 -2.71362782e-01 -6.61281586e-01 -1.07053626e+00 3.88161361e-01 7.32705295e-01 8.64167392e-01 -2.66265869e-01 5.12945771e-01 -1.05896151e+00 -1.57041088e-01 -1.01186499e-01 6.57998770e-02 4.02423233e-01 2.16162100e-01 2.49804288e-01 -5.87454557e-01 -4.85728383e-01 -3.72330248e-01 -6.16340876e-01 1.90648764e-01 2.11116895e-01 -4.87959474e-01 -4.57101047e-01 -9.76557955e-02 3.20863336e-01 4.90287662e-01 2.86534131e-01 4.97883022e-01 3.00548136e-01 3.89710546e-01 8.16554606e-01 6.14279807e-01 1.36775002e-01 5.60856283e-01 1.25114143e+00 -2.59874463e-01 -1.92299545e-01 -1.50668010e-01 -1.48092225e-01 1.13278186e+00 1.65170956e+00 9.68037322e-02 -4.08918738e-01 -1.15200484e+00 5.49305558e-01 -1.77312517e+00 -5.90463400e-01 -3.27008724e-01 2.11045003e+00 8.36630940e-01 -3.33095193e-01 5.91460288e-01 -3.30706656e-01 6.78632677e-01 1.91578269e-01 1.15576044e-01 -1.14513993e+00 -4.67751473e-01 1.51799455e-01 2.21251950e-01 9.09277976e-01 -8.38922262e-01 1.52729321e+00 7.34972095e+00 4.32487756e-01 -1.21561563e+00 5.32345951e-01 2.17727885e-01 1.38817489e-01 -3.41551900e-01 2.63006955e-01 -8.06956291e-01 2.74030268e-01 1.37437844e+00 -5.34911215e-01 8.32012832e-01 4.32978153e-01 2.75705189e-01 3.59001517e-01 -7.91475117e-01 4.71831739e-01 2.19762430e-01 -1.08601820e+00 2.41647512e-01 -3.69575657e-02 8.09615612e-01 7.43235171e-01 -1.90736726e-01 2.68121511e-01 1.52423799e-01 -9.44549561e-01 9.20328140e-01 7.58313388e-02 1.13166451e+00 -4.84951586e-01 6.40271902e-01 4.20462102e-01 -7.71903634e-01 4.49157327e-01 5.56718782e-02 1.12347808e-02 4.25878525e-01 1.55755773e-01 -9.06827748e-01 9.37772810e-01 3.52139562e-01 4.15089279e-01 -1.12071447e-01 3.39429557e-01 -2.82254636e-01 9.90235567e-01 -2.85700560e-01 -2.70655483e-01 3.02601844e-01 -2.01758310e-01 9.80427623e-01 1.53741157e+00 5.96657574e-01 -3.89405251e-01 4.46405441e-01 5.12244105e-01 -4.83239964e-02 5.66726685e-01 -5.31063974e-01 -4.37716424e-01 5.78839064e-01 9.15186465e-01 -7.17681527e-01 -3.09231490e-01 -2.71976829e-01 1.30777276e+00 -2.94268668e-01 5.86598754e-01 -2.79625744e-01 -5.55278718e-01 7.72498608e-01 4.93972860e-02 -7.92895108e-02 -3.17818463e-01 -2.67174840e-01 -9.92973149e-01 2.24331975e-01 -1.28889108e+00 9.89748072e-03 -8.39519203e-01 -9.03358757e-01 1.21062040e+00 -2.42507666e-01 -1.03496945e+00 -7.11275160e-01 5.08130044e-02 -4.47977424e-01 1.61347950e+00 -6.39398396e-01 -1.33756614e+00 6.25245795e-02 4.18404758e-01 8.64455581e-01 -8.50700200e-01 1.28283215e+00 4.84587222e-01 -2.49267146e-01 9.11752582e-01 5.34939840e-02 -1.34472996e-01 1.16880178e+00 -9.52332199e-01 9.16432023e-01 6.77379310e-01 2.08653569e-01 7.04937458e-01 6.70249820e-01 -7.38874376e-01 -1.59190381e+00 -9.22144651e-01 1.88456500e+00 -4.94380772e-01 8.23596597e-01 -6.88801408e-01 -5.75484753e-01 9.28661287e-01 7.11915195e-01 -5.24616003e-01 6.46206021e-01 1.02129087e-01 2.78240908e-03 -7.87528679e-02 -7.53615141e-01 3.57353926e-01 7.99518585e-01 -8.07955027e-01 -2.72940189e-01 8.60594392e-01 1.11815000e+00 -9.18613851e-01 -5.68924725e-01 -2.46434093e-01 4.05777156e-01 -1.84750155e-01 4.26365286e-01 -7.10585535e-01 4.76128101e-01 -6.34872541e-02 -3.96860480e-01 -1.66266775e+00 -1.16074800e-01 -1.25826561e+00 4.90284055e-01 1.47626615e+00 1.12497711e+00 -1.06222510e+00 2.19693184e-02 4.06017751e-01 -8.44027579e-01 -4.89909381e-01 -1.30673325e+00 -1.05765247e+00 2.88788557e-01 -4.26490940e-02 3.55787247e-01 1.04092598e+00 2.35761791e-01 8.97266865e-01 -3.42430830e-01 -1.34716660e-01 4.06138003e-02 -3.18531513e-01 8.99734259e-01 -5.39723277e-01 -4.17404979e-01 -5.12788475e-01 -1.29186213e-01 -7.98862219e-01 1.56621501e-01 -1.20389831e+00 1.48651630e-01 -1.89114571e+00 -6.71771839e-02 -2.59535342e-01 4.35150653e-01 7.00702310e-01 5.80843957e-03 -8.95250663e-02 1.65439680e-01 4.44098234e-01 -3.21654618e-01 2.47778952e-01 8.31615806e-01 -1.12530133e-02 -4.77886617e-01 1.41929224e-01 -6.58845842e-01 -3.26122165e-01 5.88426471e-01 -2.09556803e-01 -1.69178292e-01 -6.24354124e-01 3.00054163e-01 4.61524308e-01 -1.81935102e-01 -2.76020557e-01 4.36349869e-01 -1.33278489e-01 -3.46911490e-01 -3.09806615e-01 2.64702052e-01 -7.35954791e-02 3.35786432e-01 3.46451610e-01 -5.34222126e-01 3.83778542e-01 2.64841378e-01 -3.20773244e-01 -7.05800176e-01 1.91627607e-01 4.09722716e-01 2.25488648e-01 -2.57443905e-01 -8.90363082e-02 -8.41830313e-01 6.32643998e-02 9.30697799e-01 8.66645500e-02 -2.61519521e-01 -7.37055182e-01 -9.79704082e-01 2.80933082e-01 2.87210852e-01 8.25808883e-01 4.41260546e-01 -1.31982577e+00 -1.46097612e+00 2.99469084e-02 1.13768600e-01 -4.90102082e-01 -2.18910292e-01 1.31081665e+00 -2.25896150e-01 7.38799095e-01 1.40217796e-01 -6.93014264e-01 -1.53275084e+00 4.18351889e-02 1.93648621e-01 -1.35538965e-01 -3.78014505e-01 7.40328133e-01 -4.31968838e-01 -9.71283376e-01 1.01126276e-01 2.71954387e-01 3.52220446e-01 -6.04245346e-03 5.23110688e-01 4.02417958e-01 5.82325041e-01 -1.18005955e+00 -6.77438855e-01 2.14123875e-01 -8.07542950e-02 -1.01611900e+00 9.52802718e-01 -4.66641903e-01 -4.80498612e-01 9.68492627e-01 1.13714075e+00 -7.72447139e-02 -5.03861725e-01 -2.85219073e-01 -3.13208029e-02 -1.37914434e-01 -2.02834308e-01 -1.40181661e+00 -4.04843062e-01 6.36424959e-01 1.22703023e-01 1.16453968e-01 7.90234208e-01 5.33374166e-03 8.48712504e-01 1.94559306e-01 6.86223090e-01 -1.41845071e+00 -5.04414380e-01 1.07311475e+00 1.36751795e+00 -1.12771928e+00 -4.11790609e-01 -3.43036205e-01 -1.05987942e+00 9.27415133e-01 -4.36250232e-02 5.97992420e-01 3.40287507e-01 4.37203318e-01 5.89930832e-01 8.17798898e-02 -1.06461215e+00 3.46007310e-02 2.14935005e-01 5.71694553e-01 6.83851659e-01 2.97200680e-01 -6.33914769e-01 4.29877967e-01 -4.64885622e-01 -3.76770407e-01 2.11602822e-01 8.50190282e-01 -1.83950350e-01 -1.16066921e+00 -2.64599413e-01 4.72768955e-02 -3.30958307e-01 -2.43197307e-01 -1.56037438e+00 5.77708364e-01 -4.90910023e-01 1.37831414e+00 -2.84583509e-01 -6.68207705e-01 7.68357635e-01 2.36702636e-01 1.88063920e-01 -8.36398482e-01 -1.27211356e+00 1.92442015e-01 7.44521737e-01 -4.28629279e-01 -5.04821539e-02 -1.07495391e+00 -1.15467536e+00 -5.42143404e-01 -2.22741917e-01 5.21132708e-01 1.33581853e+00 1.11923623e+00 5.45354128e-01 3.26060742e-01 8.22271705e-01 -8.09427679e-01 -8.04408133e-01 -1.50221455e+00 -1.78930670e-01 -2.54135877e-02 1.93227708e-01 1.62355024e-02 -7.94222176e-01 5.66157736e-02]
[14.341968536376953, 7.268989086151123]
f481c349-52b8-4b7b-8c89-423e749509d0
learning-transferrable-knowledge-for-semantic
1512.07928
null
http://arxiv.org/abs/1512.07928v1
http://arxiv.org/pdf/1512.07928v1.pdf
Learning Transferrable Knowledge for Semantic Segmentation with Deep Convolutional Neural Network
We propose a novel weakly-supervised semantic segmentation algorithm based on Deep Convolutional Neural Network (DCNN). Contrary to existing weakly-supervised approaches, our algorithm exploits auxiliary segmentation annotations available for different categories to guide segmentations on images with only image-level class labels. To make the segmentation knowledge transferrable across categories, we design a decoupled encoder-decoder architecture with attention model. In this architecture, the model generates spatial highlights of each category presented in an image using an attention model, and subsequently generates foreground segmentation for each highlighted region using decoder. Combining attention model, we show that the decoder trained with segmentation annotations in different categories can boost the performance of weakly-supervised semantic segmentation. The proposed algorithm demonstrates substantially improved performance compared to the state-of-the-art weakly-supervised techniques in challenging PASCAL VOC 2012 dataset when our model is trained with the annotations in 60 exclusive categories in Microsoft COCO dataset.
['Junhyuk Oh', 'Honglak Lee', 'Bohyung Han', 'Seunghoon Hong']
2015-12-24
learning-transferrable-knowledge-for-semantic-1
http://openaccess.thecvf.com/content_cvpr_2016/html/Hong_Learning_Transferrable_Knowledge_CVPR_2016_paper.html
http://openaccess.thecvf.com/content_cvpr_2016/papers/Hong_Learning_Transferrable_Knowledge_CVPR_2016_paper.pdf
cvpr-2016-6
['foreground-segmentation']
['computer-vision']
[ 6.62060261e-01 5.93195856e-01 -2.37967983e-01 -6.07409298e-01 -7.58744299e-01 -9.32863474e-01 4.12432790e-01 -2.53006846e-01 -4.27775919e-01 4.34165835e-01 -1.85992729e-04 -2.24604398e-01 5.59769750e-01 -6.80142760e-01 -1.20184708e+00 -6.76366866e-01 4.83486056e-01 5.74789405e-01 7.70200312e-01 2.45825648e-01 2.97241539e-01 1.05939142e-01 -1.49500823e+00 6.05415225e-01 8.06601644e-01 1.03403127e+00 5.75063348e-01 8.10707867e-01 -4.85757172e-01 1.09972608e+00 -6.78207219e-01 -2.45133892e-01 1.24542236e-01 -3.88087988e-01 -1.27566183e+00 4.50222254e-01 6.08296990e-01 -1.91320136e-01 -4.50091772e-02 1.13822484e+00 1.77709572e-02 -1.73124880e-01 7.63551235e-01 -9.31043148e-01 -8.44811618e-01 7.32053161e-01 -6.90873504e-01 7.74508016e-03 -7.88202062e-02 4.29535769e-02 9.05355513e-01 -8.83109272e-01 6.62062764e-01 9.64674592e-01 5.90243578e-01 8.26986074e-01 -1.01558566e+00 -3.00851911e-01 4.79627192e-01 -1.88560952e-02 -1.06111014e+00 1.30663916e-01 8.47874999e-01 -4.55933034e-01 5.34149051e-01 -6.85439399e-03 4.39041585e-01 8.58594656e-01 -1.77777350e-01 1.29214513e+00 1.03433383e+00 -3.01711649e-01 2.02822596e-01 2.38522530e-01 5.69281042e-01 7.25120604e-01 1.00822978e-01 -2.55056679e-01 -2.64268726e-01 3.89605343e-01 7.43444860e-01 -4.75125946e-02 -2.53307484e-02 -5.43540657e-01 -1.33817363e+00 5.84042847e-01 6.47908092e-01 1.80390775e-01 -3.77104655e-02 4.16460693e-01 3.25958639e-01 -3.64891142e-01 5.05618215e-01 1.73799738e-01 -7.15396881e-01 9.87837613e-02 -1.14362884e+00 -2.35863268e-01 7.51156330e-01 1.49158227e+00 8.54913771e-01 8.21776465e-02 -3.65506619e-01 7.27967739e-01 3.23034942e-01 2.82827199e-01 2.35466465e-01 -1.00078952e+00 3.65555882e-01 6.95513368e-01 -1.88006297e-01 -3.35412651e-01 -2.14130238e-01 -5.89801073e-01 -5.51037431e-01 8.77340659e-02 6.11009479e-01 -2.62344569e-01 -1.49254727e+00 1.38307953e+00 2.42596209e-01 8.22086930e-02 2.77460456e-01 9.47064757e-01 1.13722217e+00 6.70781910e-01 3.74538034e-01 3.23937833e-01 1.23211586e+00 -1.73273766e+00 -6.57294214e-01 -4.50593978e-01 4.09876436e-01 -8.06946218e-01 1.08729923e+00 3.33595663e-01 -1.10072815e+00 -9.14112747e-01 -8.74174058e-01 -4.85405296e-01 -6.32702053e-01 4.65780765e-01 6.90253139e-01 5.02447486e-01 -1.13017452e+00 2.30631739e-01 -7.76358902e-01 -4.41683114e-01 1.11880374e+00 2.72485286e-01 1.30737051e-01 4.54545803e-02 -5.16333163e-01 3.82135302e-01 6.45785213e-01 6.59146756e-02 -1.55741346e+00 -6.61303878e-01 -7.75435269e-01 5.16541936e-02 3.49879444e-01 -5.11316061e-01 1.28088558e+00 -1.65304911e+00 -1.40758610e+00 1.33462942e+00 -2.78712392e-01 -6.33006155e-01 5.76232255e-01 -4.43558306e-01 1.64921328e-01 5.16299367e-01 3.82915288e-01 1.44704604e+00 9.36413586e-01 -1.75948751e+00 -9.78353560e-01 -4.11981568e-02 1.47393107e-01 1.77490905e-01 1.97698898e-03 -1.32445455e-01 -1.07651174e+00 -7.36874402e-01 6.18713573e-02 -8.31137359e-01 -3.52716327e-01 4.45508352e-03 -8.55258465e-01 -2.89429158e-01 1.36550045e+00 -5.99615097e-01 6.78488612e-01 -2.13857937e+00 1.95382908e-01 -3.08204025e-01 1.14701493e-02 1.80067226e-01 -2.35433728e-02 -3.21826965e-01 1.15844913e-01 4.75197546e-02 -9.16521311e-01 -5.65166831e-01 -1.11955486e-01 5.68523765e-01 -2.62183964e-01 3.84151280e-01 4.32015091e-01 1.13814890e+00 -9.66696560e-01 -8.56301665e-01 3.50547969e-01 3.17857236e-01 -6.79398894e-01 6.03329599e-01 -6.85362041e-01 5.81106842e-01 -2.32891068e-01 1.06101882e+00 7.42267966e-01 -4.57378685e-01 -9.42185894e-02 -1.05153114e-01 2.88107665e-03 -2.06626326e-01 -5.24406910e-01 2.15738058e+00 -2.77131170e-01 6.65759683e-01 1.84892818e-01 -1.39565814e+00 6.77993536e-01 7.39302710e-02 1.71407554e-02 -3.78680766e-01 2.40967706e-01 1.12705767e-01 -1.88254774e-01 -3.00758898e-01 4.12942320e-01 2.65450239e-01 -2.88092881e-01 2.18322292e-01 5.62289596e-01 -4.96125877e-01 2.66748577e-01 3.33126634e-01 6.19421184e-01 8.29390347e-01 -2.55661547e-01 -5.63039362e-01 6.84002638e-01 5.59235632e-01 4.31301355e-01 8.75127733e-01 -4.72490072e-01 1.07103467e+00 5.08534908e-01 -3.22642088e-01 -1.06071246e+00 -1.24135196e+00 -3.66686940e-01 1.50395572e+00 5.55267334e-01 1.99233890e-01 -1.40169835e+00 -1.15054309e+00 -2.31248111e-01 5.43601871e-01 -8.20250571e-01 1.71122327e-01 -4.33399379e-01 -5.36703885e-01 4.88806397e-01 8.65784049e-01 1.03712559e+00 -1.23686790e+00 -5.85280299e-01 -1.47466481e-01 -7.81914033e-03 -1.36993921e+00 -4.58759129e-01 7.33185828e-01 -8.71939123e-01 -1.29068720e+00 -9.99911606e-01 -1.58468580e+00 1.07560146e+00 1.27000690e-01 1.20637548e+00 -1.64561838e-01 -2.30616421e-01 5.84227085e-01 -2.88152307e-01 -1.69514567e-01 -1.85426593e-01 2.57142931e-01 -6.96747184e-01 6.59235790e-02 4.95971560e-01 -6.02322025e-03 -6.74740732e-01 1.94869041e-01 -9.69430149e-01 3.37864459e-01 6.16908193e-01 8.12044799e-01 1.00477862e+00 -3.07748199e-01 3.56008351e-01 -1.53069019e+00 -2.14349955e-01 -6.31540656e-01 -5.58942974e-01 1.10690430e-01 -9.99891981e-02 -8.56164470e-02 7.15442479e-01 -2.23574147e-01 -1.50591481e+00 6.66253746e-01 -1.16029285e-01 -2.15116411e-01 -7.43568897e-01 -3.06299869e-02 -2.86882013e-01 -1.80828385e-02 3.37833792e-01 3.18344772e-01 -3.54619354e-01 -5.02032101e-01 8.86643350e-01 7.90410519e-01 1.02052057e+00 -6.67540789e-01 5.76056302e-01 7.44444251e-01 -3.82935762e-01 -6.00108087e-01 -1.31144238e+00 -7.05208004e-01 -1.23313498e+00 -1.62465900e-01 1.60777819e+00 -1.07147396e+00 -1.08468287e-01 5.82143188e-01 -1.34810483e+00 -9.14988816e-01 -3.75941455e-01 -1.20048754e-01 -7.84588218e-01 1.18447162e-01 -6.79016531e-01 -4.22096908e-01 -1.00259803e-01 -1.22938526e+00 1.33470118e+00 4.05521154e-01 -7.52547234e-02 -1.13695526e+00 -3.88443321e-01 7.68364310e-01 1.99056163e-01 3.28875124e-01 6.45005405e-01 -8.13109696e-01 -9.20600474e-01 5.97952567e-02 -7.28771627e-01 5.88053465e-01 -7.96745792e-02 -1.37875065e-01 -1.38924909e+00 1.09734699e-01 -3.62221599e-01 -5.07307351e-01 1.21386933e+00 4.86161947e-01 1.83870900e+00 -1.29626757e-02 -4.81880277e-01 9.14861619e-01 1.48386836e+00 1.85210913e-01 5.14905870e-01 3.46503824e-01 1.35984480e+00 6.54155672e-01 4.79166269e-01 -2.30298087e-01 3.24523598e-01 2.14273125e-01 4.74875838e-01 -7.02952862e-01 -5.60586870e-01 -1.21511139e-01 1.16646580e-01 4.25750613e-01 1.50083005e-01 -1.39241531e-01 -8.54990840e-01 1.05907071e+00 -1.83038783e+00 -4.91750509e-01 -3.19371760e-01 1.53658831e+00 9.36759472e-01 3.57341379e-01 -1.28616795e-01 -2.00814351e-01 8.35596085e-01 1.91524595e-01 -6.72967732e-01 -6.47216558e-01 -1.83359772e-01 3.48773986e-01 8.21619213e-01 4.04044867e-01 -1.52438176e+00 1.48867965e+00 6.69558191e+00 8.06523681e-01 -8.39894593e-01 4.43867445e-01 1.10027289e+00 1.96981683e-01 -1.33135110e-01 -6.53581172e-02 -7.47566164e-01 4.33964908e-01 6.73917115e-01 4.01738405e-01 -1.16016924e-01 1.38180792e+00 -1.19554311e-01 -2.45284557e-01 -1.20309412e+00 5.92908919e-01 2.46367574e-01 -1.34971416e+00 2.58070558e-01 -2.79470474e-01 1.35111189e+00 -2.81416588e-02 2.32284218e-01 3.84361029e-01 4.80903238e-01 -1.08164835e+00 9.53947484e-01 2.69247293e-01 6.63780987e-01 -7.12338269e-01 7.79150248e-01 6.59832954e-02 -1.12240732e+00 1.23589732e-01 -3.33644062e-01 7.20665529e-02 7.33465701e-02 4.19822693e-01 -5.93884289e-01 9.96340886e-02 8.47608507e-01 9.59641933e-01 -7.48112798e-01 6.96061671e-01 -5.85680962e-01 1.01897061e+00 1.11552708e-01 2.07668841e-01 7.69759297e-01 -1.77729592e-01 1.19595863e-01 1.61247766e+00 -3.06011528e-01 -3.50362575e-03 3.23962241e-01 1.23955750e+00 -3.68961573e-01 -3.17399204e-01 -4.12691832e-01 5.23331054e-02 1.03464358e-01 1.32789958e+00 -1.31990314e+00 -7.68721879e-01 -5.76387882e-01 1.50011361e+00 2.16495737e-01 7.01506257e-01 -8.99706364e-01 -4.40513700e-01 1.13797270e-01 -1.18307889e-01 5.28546154e-01 -2.51136422e-02 -9.33016598e-01 -7.49374449e-01 -3.02776039e-01 -3.93218011e-01 4.41393316e-01 -9.76180017e-01 -1.01835907e+00 5.20145893e-01 -1.81648582e-01 -9.03310359e-01 2.75895506e-01 -7.62094736e-01 -5.78803480e-01 5.58340251e-01 -1.80440450e+00 -1.61377549e+00 -3.49076301e-01 5.00041068e-01 1.21284604e+00 2.39779893e-02 4.28319275e-01 -1.47675043e-02 -5.54151058e-01 2.75847673e-01 -1.54377427e-02 5.12112319e-01 5.49179077e-01 -1.91628850e+00 2.32837155e-01 1.15302312e+00 1.61561102e-01 3.46237034e-01 2.77161628e-01 -5.74052572e-01 -8.82788301e-01 -1.61357343e+00 3.86138171e-01 -5.76886892e-01 3.53588760e-01 -4.55039114e-01 -9.02997911e-01 8.76696408e-01 7.11290956e-01 3.05742234e-01 6.29087031e-01 -2.65749186e-01 -2.29312643e-01 2.44566545e-01 -1.06331336e+00 4.56985384e-01 1.01132357e+00 -5.34405351e-01 -8.34565878e-01 5.36281645e-01 1.14646029e+00 -6.30025864e-01 -3.66751045e-01 3.14367384e-01 9.87512544e-02 -8.24947119e-01 9.43227530e-01 -3.70632380e-01 8.38340104e-01 -5.61144531e-01 -8.75982717e-02 -8.60583186e-01 3.74988429e-02 -3.22407424e-01 3.43844779e-02 1.25944638e+00 5.07033944e-01 -2.34859794e-01 1.07160091e+00 3.47017199e-01 -7.14352667e-01 -6.01707697e-01 -5.97268164e-01 -2.65234977e-01 2.81142920e-01 -3.79257411e-01 1.46895990e-01 7.50713825e-01 -4.22025353e-01 1.30923048e-01 6.37961477e-02 3.42190921e-01 9.17456448e-01 3.42414379e-01 6.57952249e-01 -8.81016612e-01 -1.33159325e-01 -3.57098699e-01 -1.85461253e-01 -1.43129766e+00 4.43889529e-01 -9.36980128e-01 5.70992291e-01 -1.74712539e+00 4.54747915e-01 -3.40352178e-01 -2.30665833e-01 6.74415827e-01 -1.93912461e-01 8.35525453e-01 1.22354059e-02 5.96051775e-02 -1.16031027e+00 2.73625702e-01 1.46905720e+00 -4.79426831e-01 5.28700240e-02 -2.98274398e-01 -6.74306810e-01 1.08163822e+00 5.79999268e-01 -3.05502594e-01 -2.96079844e-01 -6.53994203e-01 -4.21008885e-01 -4.25925881e-01 4.44862038e-01 -1.03057957e+00 4.26440418e-01 1.47196189e-01 6.80715919e-01 -1.00852072e+00 -9.13531557e-02 -7.79004455e-01 -4.77165192e-01 3.32674474e-01 -4.76963609e-01 -7.28793919e-01 3.16591352e-01 7.66020417e-01 -4.59414363e-01 -2.39369676e-01 9.28421021e-01 -3.37011784e-01 -1.31067646e+00 1.48425668e-01 -3.91854078e-01 2.27475628e-01 1.27133310e+00 -3.44822496e-01 -2.88346022e-01 1.86263174e-01 -1.04035127e+00 3.68247360e-01 5.70258915e-01 3.41418356e-01 4.64027226e-01 -9.33639526e-01 -2.49989197e-01 2.36109346e-02 2.82683402e-01 7.36549854e-01 6.99978843e-02 4.56175029e-01 -8.55081320e-01 5.25977671e-01 -8.45032856e-02 -1.14088392e+00 -9.45321620e-01 5.78521252e-01 2.57023573e-01 1.51770338e-01 -4.20847684e-01 1.29088688e+00 7.53419220e-01 -4.94312853e-01 5.74428737e-01 -4.10117984e-01 -2.13927165e-01 -8.10040310e-02 2.45236874e-01 -7.23025501e-02 -2.44373411e-01 -7.06390321e-01 -3.35621685e-01 8.94651234e-01 -8.99295583e-02 1.86993480e-01 1.08869731e+00 -2.26055622e-01 -7.98933581e-02 5.02442420e-01 1.17557645e+00 -3.39180768e-01 -1.93994570e+00 -4.56665605e-02 7.54049644e-02 -1.37116686e-01 5.43988273e-02 -9.89544690e-01 -1.31197858e+00 1.07130075e+00 6.52329445e-01 -1.85308203e-01 1.08309841e+00 2.21384332e-01 8.28477085e-01 5.45183308e-02 3.24958712e-01 -1.24670708e+00 3.03509831e-01 4.71621335e-01 4.52097476e-01 -1.49477720e+00 -4.23468202e-01 -6.69214666e-01 -9.57171082e-01 1.09557140e+00 8.83776963e-01 -4.39820319e-01 5.48469782e-01 2.53655136e-01 3.44197661e-01 -1.29343197e-01 -1.88634887e-01 -6.52560472e-01 3.28610599e-01 9.07009900e-01 5.10463059e-01 1.04951682e-02 -1.54289126e-01 7.29540944e-01 2.77485132e-01 -1.99141383e-01 5.45285523e-01 8.13139379e-01 -6.04773760e-01 -8.07341933e-01 -3.42523068e-01 1.56623080e-01 -4.61262494e-01 -2.70114273e-01 -5.03272235e-01 6.26485825e-01 4.95490789e-01 9.86477733e-01 4.27868634e-01 1.87969208e-01 -3.46483216e-02 1.32211059e-01 1.91348910e-01 -1.06004775e+00 -6.29985869e-01 1.58045828e-01 -8.46506581e-02 -6.13557756e-01 -7.75245905e-01 -3.94610941e-01 -1.55939710e+00 3.37511450e-01 -1.47685185e-02 8.45712051e-02 4.99895006e-01 1.04374492e+00 9.49811712e-02 1.06608355e+00 2.65672594e-01 -1.00353432e+00 2.26513818e-01 -7.86510825e-01 -4.72303361e-01 6.19364560e-01 2.89583236e-01 -4.91379648e-01 -3.13236684e-01 9.02627289e-01]
[9.588547706604004, 0.5639662742614746]
07cf7a6b-cdd8-4ebc-b7ec-66aaca856c81
unraveling-the-arc-puzzle-mimicking-human
2306.08204
null
https://arxiv.org/abs/2306.08204v1
https://arxiv.org/pdf/2306.08204v1.pdf
Unraveling the ARC Puzzle: Mimicking Human Solutions with Object-Centric Decision Transformer
In the pursuit of artificial general intelligence (AGI), we tackle Abstraction and Reasoning Corpus (ARC) tasks using a novel two-pronged approach. We employ the Decision Transformer in an imitation learning paradigm to model human problem-solving, and introduce an object detection algorithm, the Push and Pull clustering method. This dual strategy enhances AI's ARC problem-solving skills and provides insights for AGI progression. Yet, our work reveals the need for advanced data collection tools, robust training datasets, and refined model structures. This study highlights potential improvements for Decision Transformers and propels future AGI research.
['Sundong Kim', 'Sejin Kim', 'Sabina Ualibekova', 'Mintaek Lim', 'Sanha Hwang', 'Jaegyun Im', 'JaeHyun Park']
2023-06-14
null
null
null
null
['clustering', 'imitation-learning']
['methodology', 'methodology']
[ 1.31235585e-01 4.97302324e-01 -1.22858398e-01 1.94292501e-01 -1.99655816e-01 -4.96747136e-01 7.97177315e-01 8.98638070e-02 -3.08944643e-01 4.76171285e-01 -7.31278124e-05 -2.81997651e-01 -8.48337948e-01 -5.49120426e-01 -3.50721806e-01 -4.26577091e-01 -5.47905490e-02 1.20555294e+00 -1.31033227e-01 -4.40063536e-01 6.72500670e-01 6.53039157e-01 -1.34106278e+00 -1.68379173e-02 1.24680138e+00 7.19116449e-01 2.71731853e-01 6.10117316e-01 1.40225515e-01 1.24399543e+00 -6.02113962e-01 -3.53265613e-01 3.10213178e-01 -3.28481078e-01 -1.19800210e+00 -1.19952433e-01 -1.52270362e-01 -9.59136114e-02 -1.71040297e-01 6.25234902e-01 8.98728296e-02 3.04532111e-01 7.89386630e-01 -1.67111897e+00 -8.28983665e-01 6.30589843e-01 7.72226229e-03 2.49212280e-01 4.30842102e-01 5.05733967e-01 8.62510443e-01 -5.34482121e-01 8.35524261e-01 1.33945537e+00 7.39365816e-01 9.09805059e-01 -1.26478732e+00 -4.78965580e-01 7.74838105e-02 5.77304900e-01 -1.12040710e+00 -7.73651153e-02 9.77203727e-01 -5.94185889e-01 1.12721932e+00 -1.13192005e-02 1.17757404e+00 1.27416956e+00 8.27567726e-02 1.30174494e+00 1.04571497e+00 -5.75092077e-01 5.50825536e-01 2.69582361e-01 1.69282481e-01 6.38845623e-01 2.64385432e-01 4.01833564e-01 -7.05712616e-01 7.22149834e-02 1.15647411e+00 -1.11958690e-01 1.88724115e-01 -6.48083925e-01 -1.35420239e+00 9.04709876e-01 3.26132029e-01 6.94382548e-01 -5.15429556e-01 2.41287574e-01 1.39652207e-01 4.67297673e-01 -2.35266015e-01 1.28077900e+00 -1.99725106e-01 -4.52679902e-01 -5.83983481e-01 7.65182853e-01 1.00573170e+00 9.79135275e-01 3.10033768e-01 -5.15463054e-02 1.14459790e-01 3.76923889e-01 3.13197166e-01 7.72108883e-02 5.23847163e-01 -1.56266105e+00 4.04467940e-01 1.12644529e+00 5.29671237e-02 -1.07624280e+00 -6.09393001e-01 -3.86453986e-01 -3.60020071e-01 6.02952778e-01 4.76648241e-01 1.59844503e-01 -5.00976026e-01 1.49416602e+00 2.72972912e-01 -7.16410205e-02 3.23433012e-01 8.36410940e-01 3.07207584e-01 3.56327862e-01 7.50134587e-02 2.14071244e-01 9.53673601e-01 -1.13229585e+00 -3.26198280e-01 -1.75989181e-01 8.72895300e-01 1.18878312e-01 1.06344879e+00 7.78144002e-01 -1.57941103e+00 -4.39760745e-01 -9.38546956e-01 -2.18165845e-01 -5.12756824e-01 -1.37431845e-01 1.03432012e+00 2.67732590e-01 -1.04454982e+00 7.37534344e-01 -9.07298803e-01 -5.73820174e-01 6.31535530e-01 3.39135706e-01 -1.42958328e-01 1.69978514e-02 -8.40679467e-01 1.35854065e+00 5.41301906e-01 -3.88383470e-03 -1.02610207e+00 -6.26418352e-01 -7.53879189e-01 -6.24583028e-02 4.50969070e-01 -8.19716513e-01 1.19500661e+00 -8.61373901e-01 -1.62163341e+00 8.55094135e-01 3.02428752e-01 -7.41783261e-01 5.17524302e-01 -5.85212223e-02 8.42446089e-02 5.00493467e-01 1.77576412e-02 7.85257995e-01 7.55240023e-01 -1.17414653e+00 -6.20349705e-01 -4.65846837e-01 1.63569406e-01 4.63464558e-01 -9.78455599e-03 7.49080256e-02 4.82743233e-02 -6.57582998e-01 1.56106632e-02 -9.36950862e-01 -1.71523690e-01 -1.39851924e-02 1.32673681e-01 -7.51286268e-01 5.15147030e-01 -7.07341075e-01 1.10569561e+00 -2.19280839e+00 9.28463519e-01 2.77024329e-01 5.72133183e-01 2.60744750e-01 4.27326793e-03 4.65673238e-01 3.19831222e-01 3.56237926e-02 -2.49262512e-01 -1.30017594e-01 4.42557395e-01 4.66240615e-01 -1.45238578e-01 -5.05465306e-02 2.35928789e-01 1.25483334e+00 -1.04900396e+00 -6.31452858e-01 2.08650351e-01 -8.23760703e-02 -8.95576000e-01 3.69121492e-01 -4.02104288e-01 5.35182714e-01 -5.22715688e-01 7.79385984e-01 -5.01169525e-02 -2.62722254e-01 7.32165128e-02 4.17357951e-01 -1.62887424e-01 2.56726325e-01 -9.33489084e-01 1.89402843e+00 -2.05514848e-01 5.04190087e-01 2.34003440e-01 -1.50554323e+00 8.84880960e-01 2.25916162e-01 3.79382044e-01 -7.43473351e-01 7.94135481e-02 1.87172085e-01 4.03288782e-01 -6.31744504e-01 2.53425777e-01 1.87820569e-02 -1.77336738e-01 5.27839720e-01 6.89065009e-02 -7.18803465e-01 2.39194617e-01 3.77618074e-01 9.78728771e-01 4.98268783e-01 2.54459262e-01 -5.76844990e-01 3.35890412e-01 6.50064826e-01 4.99595046e-01 9.71674204e-01 -4.44034517e-01 1.22028086e-02 3.66064906e-01 -5.23070216e-01 -9.33448434e-01 -9.83050823e-01 1.11930244e-01 1.01706600e+00 1.21169634e-01 -2.63351291e-01 -8.92108202e-01 -3.62307608e-01 1.26113683e-01 1.16176283e+00 -6.98228419e-01 -4.84874099e-01 -6.99050426e-01 -1.82217360e-01 6.77624106e-01 7.77756751e-01 6.05379224e-01 -1.63753116e+00 -1.23348296e+00 4.73623246e-01 -9.30454582e-02 -5.57816446e-01 1.85085997e-01 1.59954220e-01 -7.47996569e-01 -9.65348184e-01 -3.24939102e-01 -9.89555418e-01 3.89875650e-01 -6.80372305e-03 1.03944266e+00 2.65082389e-01 -3.76314342e-01 9.24204230e-01 -3.27404261e-01 -7.84601271e-01 -5.23606062e-01 4.06093784e-02 2.01371640e-01 -7.19916165e-01 4.81034070e-01 -6.53119087e-01 -4.91446584e-01 3.19605887e-01 -4.09177005e-01 3.61122191e-01 4.11634266e-01 8.28486681e-01 2.41622683e-02 2.02555031e-01 4.49859977e-01 -4.92309965e-02 9.65795934e-01 -3.58331472e-01 -4.47191298e-01 3.95055294e-01 -9.59026039e-01 1.79722495e-02 4.35653508e-01 -5.94905257e-01 -1.26983988e+00 -5.84055901e-01 4.84285772e-01 -3.64750206e-01 -3.12188894e-01 3.63913089e-01 9.73408669e-02 -1.18885525e-01 7.00009882e-01 3.29093277e-01 2.18965247e-01 -3.59532267e-01 3.01372588e-01 4.34330791e-01 6.86720908e-01 -1.17060328e+00 6.06405199e-01 1.68107182e-01 -1.33881554e-01 -7.05444098e-01 -4.81368870e-01 3.75619670e-03 -7.41717517e-01 -1.77161828e-01 8.22900474e-01 -5.63864172e-01 -1.13702893e+00 4.67744172e-01 -9.68668103e-01 -8.91972125e-01 -4.54332829e-01 3.23659062e-01 -1.14892578e+00 1.55641675e-01 -6.68485522e-01 -8.60521436e-01 -3.00650507e-01 -1.01818597e+00 7.17635334e-01 3.17387074e-01 -7.66293108e-01 -1.00513923e+00 1.02869704e-01 7.87075639e-01 4.80343103e-01 1.97323523e-02 1.06753564e+00 -9.62001085e-01 -6.26779377e-01 1.14219159e-01 -1.49371758e-01 1.51828732e-02 -3.87631029e-01 -2.04422668e-01 -2.90098399e-01 -1.64807454e-01 1.95575148e-01 -6.04541540e-01 1.88289747e-01 -6.56707957e-02 1.02416003e+00 -4.89969641e-01 -3.61533254e-01 4.25793052e-01 9.21358347e-01 7.05224276e-01 5.22535563e-01 8.55924904e-01 5.55266857e-01 8.63006353e-01 6.73530579e-01 4.15738583e-01 8.24356079e-01 4.73518848e-01 1.46232456e-01 3.42644662e-01 1.79075092e-01 -3.00321609e-01 7.66068418e-03 4.81993139e-01 -5.43852866e-01 1.29128322e-01 -1.48627985e+00 6.75366402e-01 -2.06895065e+00 -1.13962328e+00 3.11362177e-01 1.54625463e+00 7.88325310e-01 1.34754956e-01 2.93756396e-01 3.83536845e-01 7.36532733e-02 -6.07891917e-01 -6.33956850e-01 -6.37698531e-01 1.21472202e-01 2.04969913e-01 -1.41668618e-01 3.44636649e-01 -4.66093540e-01 1.23232186e+00 7.59624004e+00 4.14118022e-01 -5.94772637e-01 -2.91811764e-01 1.57940120e-01 5.31585254e-02 -7.78013915e-02 -3.15943398e-02 -4.95082855e-01 2.92340130e-01 8.88934612e-01 -3.57315153e-01 1.17580235e+00 8.58926892e-01 -4.57176790e-02 -1.27336159e-01 -1.39490819e+00 8.17678630e-01 1.19601451e-01 -1.29028380e+00 6.39710352e-02 1.61591783e-01 3.47745925e-01 -4.69851136e-01 5.24579221e-03 6.22824252e-01 5.97331524e-01 -1.06966639e+00 8.89932275e-01 6.74504161e-01 -1.64532277e-04 -5.96693933e-01 2.72570997e-01 4.93757099e-01 -8.66390944e-01 -7.73280680e-01 -6.22912683e-02 -5.15112877e-01 -3.53994757e-01 -5.08825123e-01 -8.34787607e-01 3.53701472e-01 7.09169328e-01 7.12290883e-01 -5.32873333e-01 6.05613768e-01 -2.10139900e-01 5.25456190e-01 -3.33592296e-01 -1.90873563e-01 2.97180563e-01 -5.40028870e-01 6.38844311e-01 8.55043471e-01 -1.09441102e-01 4.95728254e-01 1.25503108e-01 1.37550235e+00 5.83383143e-01 -4.30629611e-01 -6.65040195e-01 -4.25179929e-01 6.51445687e-01 7.99610138e-01 -5.13275146e-01 -3.85969430e-01 -2.38189355e-01 9.12604153e-01 5.54772615e-01 2.37663805e-01 -5.73872030e-01 -3.86699766e-01 4.90499139e-01 -2.22361296e-01 1.33092374e-01 -5.59960186e-01 -7.37722635e-01 -1.08313179e+00 -2.02988446e-01 -1.38856387e+00 4.75428998e-01 -9.63613808e-01 -9.38957214e-01 4.55947407e-03 5.09274721e-01 -6.73309565e-01 -6.99953794e-01 -8.23902905e-01 -8.30292523e-01 2.84583300e-01 -1.03214455e+00 -1.36863351e+00 -3.36842656e-01 7.81944871e-01 5.40286720e-01 -2.87516624e-01 6.72533393e-01 -2.63549179e-01 -7.92741656e-01 5.16340733e-01 -1.27543420e-01 6.10328205e-02 -2.42780283e-01 -1.26643097e+00 4.07414854e-01 3.55152518e-01 -8.05527791e-02 8.64498317e-01 7.40236163e-01 -4.22461689e-01 -1.68041837e+00 -4.14797366e-01 5.64560533e-01 -7.82989979e-01 9.11466658e-01 -2.41904154e-01 -9.76917446e-01 1.02509069e+00 9.55900848e-02 -7.52306640e-01 3.25747192e-01 -2.38630194e-02 8.01481530e-02 2.88601339e-01 -1.35761619e+00 7.96991110e-01 1.38668299e+00 -4.11950797e-01 -1.54612505e+00 1.29081234e-01 6.35942817e-01 -8.40739906e-02 -1.13758266e+00 8.68386775e-02 4.31539685e-01 -7.41138160e-01 1.09614134e+00 -6.90069258e-01 3.98257226e-01 3.18726040e-02 1.37677729e-01 -1.31090820e+00 -7.01822877e-01 -8.07318330e-01 -4.01402444e-01 9.51229036e-01 2.09775463e-01 -7.79240131e-01 7.05610216e-01 1.18411553e+00 -1.34956509e-01 -5.72245181e-01 -9.10847187e-01 -9.20385301e-01 4.44525301e-01 -4.28975910e-01 4.74871159e-01 9.99666929e-01 8.97946000e-01 7.74062276e-02 1.52650282e-01 -2.55109668e-01 7.93340921e-01 1.32239684e-01 8.25268328e-01 -1.71522844e+00 -1.93806708e-01 -8.62824082e-01 -2.88816035e-01 -5.66592276e-01 2.29288489e-01 -9.61080909e-01 -3.38057488e-01 -1.61807942e+00 -9.93034020e-02 -3.27352971e-01 2.83033457e-02 4.75674361e-01 1.98943004e-01 -3.59652817e-01 3.49100709e-01 4.08608109e-01 -6.63943470e-01 5.63035727e-01 1.01869655e+00 -6.98411316e-02 -3.98460329e-01 -4.49293345e-01 -8.56180549e-01 9.67096508e-01 1.18785131e+00 -1.46044761e-01 -3.95894855e-01 -3.90861750e-01 2.58071989e-01 -1.70333207e-01 6.40964925e-01 -1.38956416e+00 5.83258390e-01 -5.47036231e-01 2.63675123e-01 -3.27642143e-01 3.22802931e-01 -9.40098345e-01 -9.70039517e-02 8.35049391e-01 -4.90966082e-01 1.23451188e-01 3.06471288e-01 3.39996278e-01 2.40280002e-01 -1.92370489e-01 5.91396272e-01 -5.45165539e-01 -8.35193455e-01 -2.10995987e-01 -7.18645394e-01 1.23859420e-01 1.29959786e+00 -6.33621037e-01 -7.71290213e-02 3.29512060e-02 -7.08268464e-01 7.29138136e-01 5.89579821e-01 2.41099343e-01 6.22625709e-01 -1.03154802e+00 -5.79370439e-01 2.46428475e-01 1.72792554e-01 2.09176466e-01 -2.58449405e-01 7.82204926e-01 -4.48901266e-01 5.38849115e-01 -6.28209114e-01 -2.88401127e-01 -8.09512258e-01 9.34083879e-01 4.30687368e-01 -1.76225677e-01 -1.07026350e+00 7.15374529e-01 -1.80189945e-02 -4.29037631e-01 4.81086344e-01 -2.75873899e-01 -2.52235025e-01 -2.46046260e-01 5.10177195e-01 8.41997862e-01 -6.48922741e-01 -1.23813756e-01 -2.11809039e-01 3.50847214e-01 -2.28983179e-01 -3.32590938e-01 1.36844242e+00 -8.51937905e-02 -2.68172119e-02 2.91793525e-01 4.50648159e-01 -7.47269928e-01 -1.10690618e+00 -1.28666282e-01 4.35408562e-01 -2.39831343e-01 -1.82443842e-01 -1.00979388e+00 -1.96844190e-01 7.67840981e-01 1.38839692e-01 2.44846538e-01 7.78746963e-01 2.69233257e-01 4.04125839e-01 8.28076541e-01 5.53435504e-01 -1.50780129e+00 3.70050311e-01 3.45234364e-01 1.24118972e+00 -1.00892091e+00 -1.51230484e-01 2.00096816e-01 -8.94281030e-01 9.52436626e-01 9.49412882e-01 -2.35443681e-01 1.12280183e-01 2.67490417e-01 -2.74861455e-01 -5.59021771e-01 -8.81639719e-01 -7.25312307e-02 8.07544887e-02 1.01317430e+00 -5.32250851e-02 -9.32375118e-02 -2.65230924e-01 5.71010888e-01 -4.52095777e-01 2.04900742e-01 3.06237549e-01 1.30585289e+00 -4.55235749e-01 -5.74146152e-01 -4.67838317e-01 1.92701399e-01 1.33813232e-01 2.53552675e-01 -7.26646245e-01 1.11638236e+00 -4.60873917e-02 8.32334936e-01 6.73746467e-02 -1.58537790e-01 2.76419222e-01 2.77808130e-01 7.66416311e-01 -2.36083210e-01 -6.89696431e-01 -5.82939267e-01 -1.96446091e-01 -7.41141140e-01 -3.02140146e-01 -8.01403821e-01 -1.42230999e+00 -2.90373385e-01 1.27200047e-02 1.89790383e-01 6.36976182e-01 9.54132080e-01 5.21918058e-01 3.61411572e-01 1.86431795e-01 -1.05789948e+00 -6.04414165e-01 -9.08200681e-01 -2.15888888e-01 3.61101151e-01 7.42583051e-02 -8.06497753e-01 -4.47573125e-01 -2.70116687e-01]
[4.214466094970703, 1.548784613609314]
448629ed-b1bd-4d5e-befc-dc9ce22e46b5
survey-of-deep-learning-for-autonomous
2210.08487
null
https://arxiv.org/abs/2210.08487v3
https://arxiv.org/pdf/2210.08487v3.pdf
Survey of Deep Learning for Autonomous Surface Vehicles in the Marine Environment
Within the next several years, there will be a high level of autonomous technology that will be available for widespread use, which will reduce labor costs, increase safety, save energy, enable difficult unmanned tasks in harsh environments, and eliminate human error. Compared to software development for other autonomous vehicles, maritime software development, especially on aging but still functional fleets, is described as being in a very early and emerging phase. This introduces very large challenges and opportunities for researchers and engineers to develop maritime autonomous systems. Recent progress in sensor and communication technology has introduced the use of autonomous surface vehicles (ASVs) in applications such as coastline surveillance, oceanographic observation, multi-vehicle cooperation, and search and rescue missions. Advanced artificial intelligence technology, especially deep learning (DL) methods that conduct nonlinear mapping with self-learning representations, has brought the concept of full autonomy one step closer to reality. This paper surveys the existing work regarding the implementation of DL methods in ASV-related fields. First, the scope of this work is described after reviewing surveys on ASV developments and technologies, which draws attention to the research gap between DL and maritime operations. Then, DL-based navigation, guidance, control (NGC) systems and cooperative operations, are presented. Finally, this survey is completed by highlighting the current challenges and future research directions.
['Carlo Ratti', 'Jie Yang', 'Fábio Duarte', 'Wei Wang', 'Jiaxin Yin', 'Yuanyuan Qiao']
2022-10-16
null
null
null
null
['self-learning']
['natural-language-processing']
[-3.70591879e-02 1.23431414e-01 3.45558091e-03 -3.52614582e-01 -2.72198543e-02 -4.91455644e-01 7.33420014e-01 -1.55316144e-01 -7.34816194e-01 7.69155383e-01 -2.68342912e-01 -3.77480149e-01 -3.90505582e-01 -8.80547881e-01 -3.33536536e-01 -8.21727693e-01 -5.69552541e-01 2.61681050e-01 1.20196521e-01 -9.60403740e-01 1.65579379e-01 8.26256096e-01 -1.73414731e+00 -3.32058966e-01 9.61805701e-01 1.14300561e+00 3.58888060e-01 6.62976980e-01 2.53576487e-01 4.89879549e-01 -4.26515490e-01 2.30248004e-01 6.05731726e-01 -4.85870093e-02 -3.62835288e-01 -2.21886694e-01 3.08104783e-01 -1.61662132e-01 -3.12922508e-01 9.59877253e-01 4.33048725e-01 4.81182039e-01 4.54739958e-01 -1.50766873e+00 -5.02438620e-02 6.94658682e-02 -4.68942411e-02 2.38515958e-02 -3.37609276e-02 1.37056410e-01 4.84804839e-01 -7.52215087e-01 3.18880707e-01 9.29273069e-01 9.29473341e-01 6.18353307e-01 -5.59093595e-01 -7.93764114e-01 1.11195818e-01 2.24247992e-01 -1.25958967e+00 -6.47352934e-01 3.43653649e-01 -5.35267830e-01 1.17011261e+00 1.18180481e-03 9.54402685e-01 6.05341733e-01 9.85038280e-01 4.88560408e-01 5.77158689e-01 -1.95733219e-01 4.22443509e-01 -3.72599997e-02 -1.49412984e-02 6.66256726e-01 8.17472100e-01 5.46845257e-01 -3.33971828e-01 1.75702665e-02 3.69049311e-01 -2.42438111e-02 -2.74750322e-01 -5.16329288e-01 -9.27765191e-01 1.10455430e+00 2.64603615e-01 2.34930784e-01 -5.72869420e-01 1.81979239e-01 3.38088661e-01 6.42169714e-01 7.33532235e-02 6.66645467e-01 -4.72130984e-01 -1.74249381e-01 -6.17210031e-01 4.87451643e-01 1.03763235e+00 8.32178652e-01 9.05783534e-01 9.36108232e-01 7.14042425e-01 4.91514415e-01 6.62535250e-01 1.00980461e+00 3.54468763e-01 -1.26793933e+00 1.05778188e-01 4.91887063e-01 2.03996912e-01 -1.08032131e+00 -9.96935606e-01 -3.39948833e-01 -8.16959202e-01 9.63992059e-01 -3.78620982e-01 -7.98497379e-01 -7.55046010e-01 1.22738266e+00 1.13045163e-01 -1.27049640e-01 8.80134940e-01 7.55686820e-01 8.24013591e-01 8.68471861e-01 -3.58174741e-01 -8.71415809e-02 9.02411640e-01 -7.08586097e-01 -9.96461451e-01 -8.28290761e-01 5.86268306e-01 -2.07771942e-01 2.12046176e-01 3.21309447e-01 -5.58956742e-01 -5.65822363e-01 -1.54051542e+00 5.12464643e-01 -7.34398782e-01 -3.87733400e-01 7.07539260e-01 6.98842108e-01 -1.30556500e+00 3.32700998e-01 -1.17569566e+00 -5.51043391e-01 1.79280251e-01 6.52673900e-01 -4.29165870e-01 7.10923364e-03 -1.61684501e+00 1.41039968e+00 1.48761794e-01 3.70446175e-01 -1.32253397e+00 -2.97520936e-01 -1.51611233e+00 -5.06142914e-01 1.34640783e-01 -3.62634093e-01 1.12225533e+00 -5.53376496e-01 -1.44682372e+00 3.32406700e-01 1.75154492e-01 -8.83476317e-01 1.63269527e-02 -4.25073773e-01 -6.62461936e-01 -1.79540992e-01 2.24607915e-01 6.38535380e-01 4.05126929e-01 -1.26401782e+00 -1.25261509e+00 -4.11111295e-01 1.01586372e-01 4.46902692e-01 -2.10233107e-01 -4.38381851e-01 1.32126004e-01 -5.14476141e-03 1.18006855e-01 -1.23828375e+00 -5.47017694e-01 3.38816121e-02 4.18236822e-01 -1.58957824e-01 1.19008279e+00 -3.04542631e-01 7.46967494e-01 -1.96127331e+00 1.02534190e-01 1.00765295e-01 -5.84078021e-02 3.97913963e-01 2.81882025e-02 8.58609617e-01 5.78759253e-01 -3.87283355e-01 -3.38245958e-01 -9.19759572e-02 -2.01946899e-01 6.58417523e-01 -8.99420604e-02 8.42083931e-01 -2.52482444e-01 4.83438075e-01 -8.53955507e-01 1.74104691e-01 3.28349262e-01 3.09876561e-01 -7.50472844e-02 -5.74166216e-02 3.50350916e-01 4.22653973e-01 -5.37250161e-01 8.36824059e-01 5.83320856e-01 4.31560010e-01 -8.83253589e-02 1.41579285e-01 -9.26294684e-01 -2.35120550e-01 -1.10811400e+00 1.51716685e+00 -5.67847133e-01 1.23626530e+00 9.38406348e-01 -1.27477038e+00 1.10833657e+00 2.29200646e-01 6.00194216e-01 -8.88421834e-01 1.99116364e-01 5.09619415e-01 -3.00868805e-02 -8.82149756e-01 6.60683930e-01 -1.79870442e-01 -2.72177547e-01 -1.35000587e-01 -2.85456568e-01 -4.44338471e-01 -1.34163618e-01 -1.67767763e-01 8.78843129e-01 -4.29950729e-02 2.34145775e-01 -5.68402112e-01 7.78501630e-01 5.78350782e-01 7.87837684e-01 5.09368539e-01 -4.58825588e-01 -3.30952674e-01 -2.09531575e-01 -8.48343849e-01 -5.26454151e-01 -7.07524359e-01 -1.85194030e-01 8.45841408e-01 8.26220751e-01 2.24928558e-01 -3.45316470e-01 -2.10338190e-01 4.39194620e-01 5.68727136e-01 -3.78293753e-01 -2.34394088e-01 -6.51302755e-01 -4.03681725e-01 7.95606375e-01 4.11659867e-01 7.17986286e-01 -9.28297400e-01 -8.89679134e-01 3.50070208e-01 2.39578754e-01 -1.12444723e+00 4.07221854e-01 2.41370961e-01 -8.16564262e-01 -1.00245094e+00 -2.47090548e-01 -1.04807055e+00 2.88556248e-01 6.98564589e-01 5.19431055e-01 2.27687638e-02 -1.00885116e-01 6.28816366e-01 -4.45511371e-01 -7.44510233e-01 -3.28480959e-01 -2.31705308e-01 9.11468744e-01 -2.53855944e-01 2.61183470e-01 -2.63580233e-01 -4.41080421e-01 3.26200575e-01 -6.16042614e-01 -4.46221530e-01 7.04804718e-01 7.38824725e-01 1.14225365e-01 2.08319172e-01 8.46599281e-01 -2.93518066e-01 5.47967434e-01 -6.95005655e-01 -9.30413067e-01 -2.25003749e-01 -7.96536386e-01 -2.27611512e-01 5.83228290e-01 2.25228459e-01 -9.00535643e-01 1.55414894e-01 -2.54680783e-01 8.21682289e-02 -1.71279073e-01 6.71931148e-01 1.91292111e-02 -8.07878494e-01 7.38753200e-01 3.70580256e-01 9.06017184e-01 -3.31435315e-02 -1.63429212e-02 1.16908133e+00 4.29984838e-01 7.39173740e-02 1.00079393e+00 8.66084933e-01 2.56186903e-01 -1.54621065e+00 -3.18611413e-01 -3.75218242e-01 -4.79564756e-01 -4.87330198e-01 8.38601410e-01 -1.11394238e+00 -4.84249562e-01 5.78213394e-01 -9.51682925e-01 -5.91707885e-01 4.69346717e-02 8.23529899e-01 -2.34290689e-01 3.71549428e-01 -1.30973369e-01 -9.46455717e-01 -3.36969644e-01 -1.32844305e+00 6.93528891e-01 3.98066550e-01 1.86682045e-01 -1.26986372e+00 4.44089681e-01 2.02987000e-01 9.01778460e-01 4.10285681e-01 6.35991246e-02 -4.59720314e-01 -4.86833692e-01 -6.30051613e-01 4.51009065e-01 5.70921421e-01 1.21664941e-01 -4.76614624e-01 -5.93254745e-01 -7.56601036e-01 -1.93938948e-02 -1.16647452e-01 6.63817763e-01 4.51673508e-01 4.07739170e-02 -2.19661504e-01 -7.87601233e-01 7.95362890e-01 1.33232009e+00 8.46277833e-01 4.63704586e-01 7.71216750e-01 3.54258835e-01 9.91049469e-01 1.19055879e+00 2.87866265e-01 6.98549628e-01 3.01452011e-01 9.82293606e-01 -1.54563636e-01 4.20503646e-01 3.61905903e-01 6.58944130e-01 6.24457300e-01 -3.66151780e-01 -2.55282879e-01 -1.08980453e+00 5.43299615e-01 -1.77710092e+00 -8.75003159e-01 -2.49835983e-01 2.00296974e+00 -1.04625285e-01 -1.73922598e-01 -5.63125908e-01 -6.54536709e-02 4.93488818e-01 -5.56146551e-04 -6.11611009e-01 -4.22687560e-01 8.59603658e-02 -3.86310130e-01 1.12596798e+00 8.04813206e-01 -1.38684666e+00 9.82064903e-01 6.54897594e+00 2.54111141e-01 -1.31838262e+00 -5.82010485e-02 -4.02766168e-01 4.51201111e-01 -4.85743694e-02 -3.84720489e-02 -1.08573723e+00 7.94214755e-02 9.79874849e-01 -9.17031020e-02 2.70519286e-01 1.12140656e+00 4.82087761e-01 -1.77639157e-01 -6.24706149e-01 7.88611352e-01 2.06319377e-01 -1.40402389e+00 -3.21994394e-01 2.43667260e-01 7.30165124e-01 7.04110980e-01 -2.43381292e-01 4.82887924e-01 1.50878787e-01 -7.64793873e-01 6.24135375e-01 5.21733463e-01 5.10357380e-01 -9.44936633e-01 1.40072024e+00 6.50279999e-01 -1.37242055e+00 -6.59278989e-01 -5.93712866e-01 -6.85896635e-01 4.22919691e-01 1.56186044e-01 -5.35538614e-01 5.51896870e-01 9.38848376e-01 8.29755425e-01 -5.59825962e-03 1.06218553e+00 -9.97202471e-02 -1.93156525e-02 -1.59332529e-01 -4.98855203e-01 6.47655785e-01 -3.10608238e-01 8.58851552e-01 9.65608954e-01 4.17779237e-01 2.37501979e-01 3.91111016e-01 5.09548746e-02 6.63468599e-01 -3.27440649e-01 -1.31467116e+00 1.91186577e-01 6.65483057e-01 1.28062725e+00 -1.71541199e-01 -9.59481969e-02 -6.66455328e-01 3.91833991e-01 -2.84373254e-01 1.96380869e-01 -6.46354496e-01 -1.01784563e+00 1.26205182e+00 3.16902101e-02 -1.93943635e-01 -1.08536077e+00 -5.53003997e-02 -5.94068289e-01 -4.39219415e-01 -6.10837579e-01 1.71898007e-02 -2.19557002e-01 -7.07052112e-01 8.11101556e-01 7.20322365e-03 -1.78485417e+00 -2.84276873e-01 -7.34737992e-01 -5.20866811e-01 2.98822373e-01 -1.98562360e+00 -8.98877263e-01 -7.27351904e-01 9.45299491e-02 7.63590693e-01 -9.36977386e-01 8.48566592e-01 2.21130908e-01 -1.62354782e-01 3.39337415e-03 3.96187812e-01 -8.54270607e-02 4.09103364e-01 -7.33239770e-01 3.30143452e-01 8.33864093e-01 -5.04001379e-01 4.35566038e-01 6.85079336e-01 -7.81070352e-01 -2.17010617e+00 -1.11199653e+00 4.75733131e-01 -6.42018989e-02 6.19279921e-01 -3.30037534e-01 -5.45220435e-01 7.68998146e-01 4.77091670e-01 -2.44295970e-01 5.05595684e-01 -3.29976499e-01 6.36071682e-01 -5.17030537e-01 -9.63777125e-01 3.67158949e-01 6.81870103e-01 2.06349656e-01 -5.99841654e-01 3.05464804e-01 6.47959888e-01 -3.22597951e-01 -7.31529891e-01 7.46383071e-01 7.67159522e-01 -6.66587710e-01 6.20435655e-01 -3.13225985e-01 -5.40041387e-01 -6.47810578e-01 -1.70354217e-01 -1.42014980e+00 -4.59866643e-01 -7.52187133e-01 3.14507067e-01 4.92719591e-01 4.52603519e-01 -1.10604751e+00 7.61479378e-01 3.18665326e-01 -1.14473021e+00 -4.60986823e-01 -1.19943154e+00 -9.45102572e-01 -5.42307496e-02 -4.05796409e-01 2.36574799e-01 7.97265351e-01 4.74261120e-02 1.26024798e-01 -5.65722704e-01 7.90026665e-01 7.33408153e-01 -1.70631796e-01 9.01626348e-01 -1.67841291e+00 4.47600842e-01 -4.06386890e-02 -9.50196087e-01 -9.50198591e-01 2.56749123e-01 -5.31926692e-01 5.21522403e-01 -2.08374095e+00 -9.19701040e-01 -4.13541526e-01 1.56348452e-01 5.44465125e-01 6.01529062e-01 1.47623569e-01 -3.06817234e-01 1.26401946e-01 -3.55138361e-01 7.96026826e-01 1.05713749e+00 -2.84861147e-01 -4.59809065e-01 2.02136919e-01 -4.42089915e-01 8.94104898e-01 8.95004570e-01 -7.52209872e-02 -2.54412293e-01 -6.01519167e-01 3.80197883e-01 1.33123845e-01 -6.27492070e-02 -1.52580583e+00 1.02371359e+00 -3.63913268e-01 -1.42448219e-02 -4.66267586e-01 4.33900923e-01 -1.20206356e+00 -3.22450586e-02 1.16262412e+00 3.33147109e-01 1.22732393e-01 3.60119820e-01 6.83438480e-01 -5.22036076e-01 -2.76009619e-01 8.60756934e-01 -4.96659912e-02 -1.62443829e+00 1.59342051e-01 -1.24117899e+00 -3.69327277e-01 1.57264245e+00 -5.87666333e-01 -2.59767741e-01 -5.06359279e-01 -4.92834836e-01 9.86998856e-01 3.06041837e-01 6.08277738e-01 9.88962293e-01 -8.02945137e-01 -7.68827021e-01 5.28414726e-01 1.39921000e-02 4.59250435e-02 4.25318629e-01 7.20020950e-01 -9.90918398e-01 4.86797482e-01 -4.90809023e-01 -4.93427396e-01 -9.16221619e-01 1.57716982e-02 5.30007005e-01 6.23848975e-01 -6.04972780e-01 7.02159345e-01 -1.30955964e-01 -7.18719840e-01 1.72202066e-01 9.23489332e-02 -7.52725065e-01 -7.17312545e-02 5.39858103e-01 6.84378743e-01 2.59479089e-03 -8.57764661e-01 -6.95872188e-01 9.43650544e-01 4.08112109e-01 1.83647927e-02 1.41460335e+00 -3.41352671e-01 -4.11087759e-02 2.18970150e-01 7.66012132e-01 -8.49857777e-02 -1.27866232e+00 4.35919195e-01 -8.47145617e-02 -6.72785565e-02 4.43437338e-01 -2.59087831e-01 -9.24519479e-01 6.05503082e-01 7.78759241e-01 3.87748688e-01 8.36471200e-01 -1.94754764e-01 7.78002620e-01 1.10009968e+00 8.93496156e-01 -1.35245836e+00 -2.34373719e-01 1.08621180e+00 9.74639893e-01 -1.45670128e+00 1.03167683e-01 -3.95665802e-02 -8.53820860e-01 1.35767710e+00 6.40282571e-01 -3.67308408e-01 8.57316434e-01 6.45096242e-01 5.67217946e-01 -3.73601228e-01 -5.75224757e-01 -2.22222880e-01 -1.51011840e-01 1.01754725e+00 5.66441007e-02 -1.04941100e-01 -2.06869483e-01 2.81487465e-01 -1.16199896e-01 -4.62098211e-01 7.07587123e-01 1.40734661e+00 -1.35798740e+00 -8.74615967e-01 -4.44245160e-01 2.50659943e-01 -5.64136431e-02 2.73171961e-01 1.07360490e-01 1.16629779e+00 2.05889180e-01 1.14735925e+00 2.00005531e-01 -7.42511272e-01 6.24681890e-01 -4.72087651e-01 -2.33343542e-01 -4.83724385e-01 -1.36558980e-01 -4.59084332e-01 3.44649434e-01 -5.94046831e-01 -4.38459009e-01 -6.53697014e-01 -1.77080619e+00 4.75208908e-02 -2.27347702e-01 5.77295899e-01 1.27113914e+00 8.88770878e-01 4.43905830e-01 4.32331085e-01 8.04221511e-01 -1.26866007e+00 -4.17897671e-01 -8.87428761e-01 -6.82011485e-01 -6.50926590e-01 6.23272061e-01 -1.10839188e+00 -6.25291407e-01 -2.63530463e-01]
[7.504782199859619, -1.5401514768600464]
6cd0ca9d-8eae-4fa1-9eaf-4901e8edb7f0
distribution-shift-inversion-for-out-of-1
2306.08328
null
https://arxiv.org/abs/2306.08328v1
https://arxiv.org/pdf/2306.08328v1.pdf
Distribution Shift Inversion for Out-of-Distribution Prediction
Machine learning society has witnessed the emergence of a myriad of Out-of-Distribution (OoD) algorithms, which address the distribution shift between the training and the testing distribution by searching for a unified predictor or invariant feature representation. However, the task of directly mitigating the distribution shift in the unseen testing set is rarely investigated, due to the unavailability of the testing distribution during the training phase and thus the impossibility of training a distribution translator mapping between the training and testing distribution. In this paper, we explore how to bypass the requirement of testing distribution for distribution translator training and make the distribution translation useful for OoD prediction. We propose a portable Distribution Shift Inversion algorithm, in which, before being fed into the prediction model, the OoD testing samples are first linearly combined with additional Gaussian noise and then transferred back towards the training distribution using a diffusion model trained only on the source distribution. Theoretical analysis reveals the feasibility of our method. Experimental results, on both multiple-domain generalization datasets and single-domain generalization datasets, show that our method provides a general performance gain when plugged into a wide range of commonly used OoD algorithms.
['Xinchao Wang', 'Xingyi Yang', 'Songhua Liu', 'Runpeng Yu']
2023-06-14
distribution-shift-inversion-for-out-of
http://openaccess.thecvf.com//content/CVPR2023/html/Yu_Distribution_Shift_Inversion_for_Out-of-Distribution_Prediction_CVPR_2023_paper.html
http://openaccess.thecvf.com//content/CVPR2023/papers/Yu_Distribution_Shift_Inversion_for_Out-of-Distribution_Prediction_CVPR_2023_paper.pdf
cvpr-2023-1
['domain-generalization']
['methodology']
[ 3.80677313e-01 -9.94522776e-03 -2.25855261e-01 -3.46662462e-01 -4.24961507e-01 -6.45773768e-01 5.31237245e-01 4.26341817e-02 8.64648893e-02 6.52570605e-01 -3.94170254e-01 -5.24534881e-01 -2.86657751e-01 -8.22027564e-01 -6.50774956e-01 -1.03311133e+00 2.17244968e-01 9.04043972e-01 3.43213320e-01 4.76930439e-02 -7.31450468e-02 3.61689001e-01 -1.69147503e+00 8.69130716e-02 1.19213021e+00 1.17111611e+00 2.57983387e-01 4.23066020e-01 -1.78639144e-01 3.81621391e-01 -9.39639628e-01 -3.63535166e-01 5.65922678e-01 -7.24351764e-01 -1.44128665e-01 1.91452831e-01 3.51335287e-01 -2.38913804e-01 -3.94846857e-01 1.21054673e+00 5.67900240e-01 -5.83684146e-02 1.05964267e+00 -1.52294040e+00 -1.03665721e+00 3.06756645e-01 -4.88887668e-01 1.62856922e-01 -1.42559215e-01 -3.42936754e-01 6.40089393e-01 -7.85332024e-01 5.93703091e-01 9.24427092e-01 4.53774184e-01 6.21638656e-01 -1.40725231e+00 -8.62403810e-01 -1.84934542e-01 -2.15379242e-02 -1.28593469e+00 -1.42785534e-01 8.03088546e-01 -5.52598357e-01 6.01858974e-01 -5.74271865e-02 2.95495570e-01 1.24366260e+00 2.94709146e-01 8.20237279e-01 1.01879668e+00 -5.46946049e-01 5.20915031e-01 7.57836223e-01 -5.92308156e-02 3.16309154e-01 4.00761545e-01 1.83630884e-01 -5.18900752e-01 -3.67302060e-01 4.57475394e-01 -1.66502640e-01 -3.05630654e-01 -9.06173289e-01 -4.80370343e-01 8.27552795e-01 2.14744374e-01 1.71639755e-01 -1.39427871e-01 -4.16790962e-01 1.50368959e-01 7.24859893e-01 7.30785310e-01 1.50541961e-01 -5.32526731e-01 -1.53303459e-01 -9.48605537e-01 6.83887377e-02 1.06623340e+00 1.11432266e+00 9.35150266e-01 -9.48702022e-02 2.44859666e-01 8.73585880e-01 1.91830918e-01 7.44920313e-01 7.09385276e-01 -4.61637259e-01 7.20051289e-01 7.22751081e-01 -6.24895617e-02 -9.57599580e-01 -3.03820390e-02 -7.33642459e-01 -8.22215676e-01 6.26074299e-02 6.88138008e-01 -1.11349169e-02 -8.02164853e-01 1.68855917e+00 6.27290010e-01 1.66220635e-01 1.98079035e-01 5.24292588e-01 -3.20807323e-02 6.55588269e-01 -2.95962721e-01 -5.10352217e-02 7.11111844e-01 -5.17306924e-01 -2.94147074e-01 -1.64851069e-01 8.32896054e-01 -7.17460155e-01 1.09145117e+00 6.31594181e-01 -5.68119943e-01 -5.89666426e-01 -1.21727860e+00 2.21555397e-01 -3.45717996e-01 -5.67158014e-02 1.92099497e-01 9.55509961e-01 -7.97470331e-01 6.39186561e-01 -6.57663226e-01 -1.19938858e-01 5.08886218e-01 3.96682203e-01 -4.03424978e-01 -5.57598770e-01 -9.75424051e-01 6.06754839e-01 2.79374897e-01 -1.11341052e-01 -6.03102326e-01 -8.45728576e-01 -5.73175073e-01 7.22363889e-02 3.25591303e-02 -2.69729704e-01 8.38473618e-01 -1.25873911e+00 -1.38712204e+00 4.79795158e-01 -9.18994471e-02 -3.20396334e-01 7.08578825e-01 -1.27235845e-01 -4.28747326e-01 -2.35046715e-01 1.25853837e-01 3.11412662e-01 1.18828213e+00 -1.16689885e+00 -6.96126640e-01 -6.29062891e-01 -6.58552468e-01 8.11988339e-02 -6.79678321e-01 -6.28820956e-01 -4.62177664e-01 -6.23569846e-01 2.63998926e-01 -1.00810742e+00 3.91095012e-01 -4.85818908e-02 -4.77323711e-01 -2.03111753e-01 1.13165104e+00 -4.03982729e-01 1.30883574e+00 -2.75224233e+00 2.47752726e-01 7.89147198e-01 1.26854584e-01 9.92167220e-02 -2.50921905e-01 3.72187257e-01 -2.15912029e-01 -2.43997157e-01 -3.65190238e-01 -2.79060096e-01 1.79599434e-01 4.17569786e-01 -7.74034441e-01 6.10074520e-01 1.30182937e-01 3.71186793e-01 -7.87793815e-01 -1.35517120e-01 -1.05295181e-01 5.16997039e-01 -7.53611505e-01 3.81562978e-01 -1.79861218e-01 3.75501573e-01 -4.07245964e-01 5.91133714e-01 9.82708454e-01 -1.60046801e-01 3.96271825e-01 1.71640873e-01 3.35121363e-01 1.82086974e-01 -1.15466666e+00 1.18759370e+00 -3.42212677e-01 6.62438333e-01 -2.38384143e-01 -9.70187187e-01 1.34816861e+00 -1.53504327e-01 2.88372993e-01 -8.21207881e-01 5.83736710e-02 6.87182426e-01 2.43322611e-01 -2.48756021e-01 3.31350356e-01 -2.02314034e-01 -8.81260820e-03 6.06595814e-01 2.30132565e-01 -1.64863482e-01 1.17755253e-02 -2.02174783e-01 9.49856818e-01 1.32683858e-01 1.46170231e-02 -3.35948318e-01 1.16180271e-01 -1.85397208e-01 4.34120774e-01 7.75397658e-01 -1.05530731e-01 5.68681240e-01 8.19847584e-01 -1.46197155e-01 -1.21047449e+00 -1.26464164e+00 -4.60336417e-01 8.90909553e-01 2.00810716e-01 -1.60266578e-01 -7.10416675e-01 -9.73740518e-01 4.13356543e-01 9.90312874e-01 -6.61336124e-01 -5.70569813e-01 -3.58058631e-01 -8.79033267e-01 4.82297003e-01 3.24228138e-01 3.86635840e-01 -3.43030870e-01 -1.89530164e-01 1.10606574e-01 1.85220763e-01 -7.18658864e-01 -3.45572054e-01 5.36301255e-01 -9.07636881e-01 -8.31272662e-01 -6.56476617e-01 -8.01647961e-01 7.52296507e-01 8.18464458e-02 6.92822099e-01 -2.87256628e-01 -7.23923221e-02 2.00271904e-01 -1.30315647e-01 -3.25906664e-01 -8.26203287e-01 3.16691607e-01 1.52640224e-01 1.59965515e-01 7.09594250e-01 -8.31116796e-01 -3.05340171e-01 5.28016150e-01 -9.16696250e-01 -4.04107094e-01 6.09612226e-01 9.90733624e-01 3.93280596e-01 4.86748517e-01 5.96588969e-01 -9.83597636e-01 6.40113950e-01 -8.52025092e-01 -7.35634327e-01 1.12989895e-01 -8.06017220e-01 2.29185581e-01 7.70131707e-01 -8.96717310e-01 -8.36791575e-01 -2.31575280e-01 1.23365037e-01 -6.53904080e-01 3.53346206e-02 2.52420515e-01 -6.07962191e-01 2.01994494e-01 8.34757626e-01 4.88915086e-01 2.79199809e-01 -5.21333456e-01 1.63714901e-01 1.01768148e+00 4.04588103e-01 -6.40272081e-01 1.06269264e+00 2.58281857e-01 -6.31352067e-02 -1.01706004e+00 -5.92239261e-01 -2.03914911e-01 -6.81574047e-01 9.66493636e-02 3.58218431e-01 -5.98506570e-01 -5.34122474e-02 4.81197327e-01 -8.14145625e-01 -4.75837767e-01 -4.93900180e-01 3.82078290e-01 -3.67963910e-01 3.37396681e-01 -3.99874002e-02 -5.92256188e-01 1.71979606e-01 -1.15093970e+00 8.89862478e-01 9.05857384e-02 -1.56385615e-01 -1.15761495e+00 1.49088129e-01 -6.50738552e-02 5.30491054e-01 -2.61262357e-01 1.45129645e+00 -1.21521425e+00 -4.67670351e-01 -5.37852049e-01 -1.92945629e-01 7.97046363e-01 4.23112720e-01 -3.18949521e-01 -9.30726767e-01 -3.98596793e-01 2.36394703e-01 -3.10590893e-01 5.93334436e-01 6.56082481e-02 1.08218753e+00 5.15620336e-02 -4.63729471e-01 6.32100165e-01 1.13876760e+00 2.48123810e-01 6.05565667e-01 3.06602269e-01 4.37872440e-01 5.87482870e-01 4.65323955e-01 3.43641371e-01 1.44153461e-01 5.68328798e-01 1.51882634e-01 2.11277768e-01 -2.33494788e-01 -6.36711955e-01 2.97666490e-01 6.24730170e-01 5.79490244e-01 -5.95486522e-01 -8.27584445e-01 2.83619016e-01 -1.45219874e+00 -6.81139410e-01 3.83735001e-01 2.56283236e+00 8.66425872e-01 2.28750899e-01 -1.54229984e-01 2.44728357e-01 5.91785908e-01 -2.96656549e-01 -8.97427917e-01 -7.82855675e-02 -1.90583970e-02 2.62260109e-01 3.92859042e-01 3.75818223e-01 -9.96081829e-01 5.08546352e-01 5.84228849e+00 1.05986607e+00 -1.31715798e+00 -4.85767573e-02 5.55803895e-01 3.10389668e-01 -5.36664009e-01 -1.89314112e-01 -1.05782199e+00 5.62460363e-01 8.95503759e-01 -9.11983922e-02 3.71133268e-01 1.03474665e+00 -3.07769179e-01 -3.13876644e-02 -1.41200292e+00 8.91590953e-01 1.94039032e-01 -7.89614677e-01 1.02596752e-01 3.17908049e-01 6.25757158e-01 -8.19748342e-02 4.74496365e-01 5.08429229e-01 -5.85909858e-02 -6.92012429e-01 7.24555552e-01 1.24819361e-01 8.32747281e-01 -6.60786569e-01 5.35678983e-01 5.79488397e-01 -6.74492717e-01 -2.34601662e-01 -6.82481706e-01 3.49977732e-01 -3.64609092e-01 8.40955019e-01 -1.14687371e+00 3.93911511e-01 5.11848927e-01 5.62120497e-01 -4.29178476e-01 7.99097002e-01 7.15335191e-04 7.39422917e-01 -6.19037986e-01 4.77604046e-02 -1.54411048e-01 -3.27492356e-01 5.53102374e-01 7.09948301e-01 7.84061551e-01 -5.69604933e-01 -2.55583357e-02 8.39485645e-01 -1.21787161e-01 -8.72567818e-02 -8.00407469e-01 -2.49595329e-01 6.19592130e-01 4.66750652e-01 -5.87028623e-01 -5.05572706e-02 -4.33893740e-01 1.13764417e+00 3.66978288e-01 5.11118412e-01 -8.61028016e-01 -5.56226075e-01 6.07317805e-01 2.15674415e-01 5.97229123e-01 -1.52764335e-01 -1.74794957e-01 -1.27004170e+00 1.64088264e-01 -8.92954111e-01 2.46104807e-01 -2.72633880e-01 -1.76561761e+00 6.17864490e-01 -5.50399311e-02 -1.37609494e+00 -2.28982732e-01 -7.61331439e-01 -2.90196985e-01 9.11406934e-01 -1.50677443e+00 -8.73780549e-01 1.98751520e-02 5.31437814e-01 3.10325176e-01 -4.61500645e-01 6.87672675e-01 6.21584356e-01 -4.01377738e-01 9.87982571e-01 6.41731858e-01 -1.66304320e-01 9.08187747e-01 -1.14126730e+00 7.24460259e-02 5.36551774e-01 2.06928793e-03 5.27204037e-01 7.60168731e-01 -6.88639164e-01 -1.37516665e+00 -1.17631447e+00 9.15139496e-01 -4.76765096e-01 7.37859488e-01 -7.34862745e-01 -1.14422643e+00 6.65734470e-01 -2.59419769e-01 -1.44873396e-01 7.90059030e-01 -3.56890522e-02 -5.91162562e-01 -3.75386089e-01 -1.12971759e+00 3.55311006e-01 7.70713985e-01 -6.13947690e-01 -5.20290434e-01 9.89933684e-02 3.86489660e-01 -3.89817655e-01 -5.75361967e-01 1.99962616e-01 3.68618935e-01 -8.60469341e-01 7.29670465e-01 -3.39649111e-01 2.94330031e-01 -2.43972629e-01 -5.24018466e-01 -1.31834161e+00 7.74037391e-02 -3.90902758e-01 -1.89992979e-01 1.28454173e+00 6.00874305e-01 -1.08073509e+00 7.20284462e-01 4.29925442e-01 -1.40045909e-02 -6.69241905e-01 -1.22588408e+00 -9.22928631e-01 2.72122085e-01 -4.91123617e-01 5.70736468e-01 7.40382791e-01 -1.75748229e-01 2.58066386e-01 -1.99413151e-01 4.34749484e-01 5.80017388e-01 2.73472399e-01 9.19180453e-01 -1.16275942e+00 -7.32017517e-01 -2.32329935e-01 -5.46116292e-01 -1.56724393e+00 2.20643029e-01 -1.13509715e+00 -2.29089931e-02 -7.83149123e-01 8.30917209e-02 -7.82988191e-01 -7.90180787e-02 2.09607482e-01 1.30845726e-01 1.22879989e-01 -1.60112958e-02 3.43374878e-01 -1.20965615e-01 7.25903094e-01 1.08009982e+00 -9.15835351e-02 -3.82235289e-01 2.86137760e-01 -6.28088593e-01 4.03817713e-01 5.15944839e-01 -8.10306132e-01 -9.10487294e-01 -2.56637901e-01 6.82757422e-02 -3.19153488e-01 1.36381134e-01 -9.82802153e-01 1.73778534e-01 1.43143952e-01 5.90469599e-01 -3.55999023e-01 2.51020402e-01 -8.73080075e-01 -5.05640581e-02 3.50637943e-01 -6.83658570e-02 -1.67685717e-01 2.36723945e-01 8.63043249e-01 -1.92282617e-01 -1.69663027e-01 7.48495221e-01 7.86416769e-01 -3.50628555e-01 2.52379477e-01 -7.98022822e-02 2.07590789e-01 1.04436111e+00 -2.31951207e-01 -3.84595126e-01 -2.03833967e-01 -5.55361092e-01 -9.80928466e-02 5.27779400e-01 4.83048469e-01 4.62437987e-01 -1.14561892e+00 -3.13731700e-01 9.03503835e-01 -1.22386441e-02 -5.48380800e-02 9.56861377e-02 7.44174480e-01 -2.43243128e-01 6.95521086e-02 7.46307522e-02 -8.75722766e-01 -6.62729263e-01 5.96677184e-01 3.75776440e-01 -9.34050456e-02 -5.62859535e-01 6.55363858e-01 4.51706350e-01 -6.65513992e-01 2.62709647e-01 -2.56462872e-01 4.70081419e-01 -6.89742640e-02 4.91254836e-01 1.75974473e-01 8.25718418e-02 -2.14969903e-01 -1.71309397e-01 3.71920586e-01 -1.10125735e-01 8.94308835e-02 1.32413125e+00 -1.43781975e-01 -5.38673922e-02 5.93788445e-01 1.41635549e+00 1.70189574e-01 -1.36912072e+00 -1.83644101e-01 5.43528656e-03 -6.71589732e-01 -2.44244337e-01 -6.08094692e-01 -8.28586698e-01 9.49247658e-01 7.20989764e-01 5.43497860e-01 1.18423045e+00 1.60781011e-01 5.64639509e-01 2.94956833e-01 3.15348446e-01 -8.62847030e-01 1.56020135e-01 3.85079533e-01 5.14021397e-01 -9.56609011e-01 -3.59721988e-01 -2.88075507e-01 -4.57648814e-01 1.24508238e+00 5.74277222e-01 -1.26313612e-01 1.06847990e+00 4.37491417e-01 -4.23572771e-02 1.47909701e-01 -6.22164011e-01 3.04400265e-01 1.65080696e-01 7.90686190e-01 5.68709187e-02 -1.96865320e-01 3.33912641e-01 6.92600012e-01 -2.60742724e-01 -1.07773170e-01 2.59180993e-01 8.32221746e-01 -4.30687666e-01 -1.32999897e+00 -2.04537868e-01 4.79027063e-01 8.53875279e-02 3.39522846e-02 -2.59754121e-01 8.96443844e-01 2.06004098e-01 5.53294241e-01 3.33046407e-01 -3.36536556e-01 4.00086522e-01 4.29966509e-01 4.56448913e-01 -4.99389857e-01 1.53960034e-01 -1.08606997e-03 -3.20359409e-01 -2.28387192e-01 1.10766865e-01 -5.42878687e-01 -9.50154781e-01 -3.20653021e-01 -3.41669559e-01 2.52912998e-01 5.87496400e-01 9.61987734e-01 6.07833207e-01 2.37667114e-01 6.83714390e-01 -5.64008117e-01 -1.26063263e+00 -7.67452002e-01 -9.98179376e-01 3.23695332e-01 4.40177828e-01 -8.02014470e-01 -7.43789554e-01 -1.03803843e-01]
[10.218761444091797, 3.203514814376831]
d2d198b8-8560-4bed-92ec-800d78f2c184
reducing-ann-snn-conversion-error-through
2302.02091
null
https://arxiv.org/abs/2302.02091v1
https://arxiv.org/pdf/2302.02091v1.pdf
Reducing ANN-SNN Conversion Error through Residual Membrane Potential
Spiking Neural Networks (SNNs) have received extensive academic attention due to the unique properties of low power consumption and high-speed computing on neuromorphic chips. Among various training methods of SNNs, ANN-SNN conversion has shown the equivalent level of performance as ANNs on large-scale datasets. However, unevenness error, which refers to the deviation caused by different temporal sequences of spike arrival on activation layers, has not been effectively resolved and seriously suffers the performance of SNNs under the condition of short time-steps. In this paper, we make a detailed analysis of unevenness error and divide it into four categories. We point out that the case of the ANN output being zero while the SNN output being larger than zero accounts for the largest percentage. Based on this, we theoretically prove the sufficient and necessary conditions of this case and propose an optimization strategy based on residual membrane potential to reduce unevenness error. The experimental results show that the proposed method achieves state-of-the-art performance on CIFAR-10, CIFAR-100, and ImageNet datasets. For example, we reach top-1 accuracy of 64.32\% on ImageNet with 10-steps. To the best of our knowledge, this is the first time ANN-SNN conversion can simultaneously achieve high accuracy and ultra-low-latency on the complex dataset. Code is available at https://github.com/hzc1208/ANN2SNN\_SRP.
['Zhaofei Yu', 'Tiejun Huang', 'Jianhao Ding', 'Tong Bu', 'Zecheng Hao']
2023-02-04
null
null
null
null
['temporal-sequences']
['reasoning']
[ 2.63993025e-01 -5.09499967e-01 2.70719737e-01 -1.68241799e-01 -1.51352778e-01 -1.71025082e-01 1.83381826e-01 -1.32075056e-01 -6.98518991e-01 9.68708396e-01 -4.93137687e-01 -2.51975954e-01 -4.80990522e-02 -6.57332957e-01 -8.76209378e-01 -9.53542054e-01 2.47320414e-01 1.17509775e-02 4.93878037e-01 -6.37518913e-02 4.03366625e-01 3.86522293e-01 -1.60437143e+00 5.97014129e-02 8.60840678e-01 1.33003831e+00 4.28726345e-01 1.67390540e-01 -4.09672223e-02 3.16876382e-01 -6.55170798e-01 -1.43283635e-01 2.59364873e-01 -5.50470233e-01 -1.31586462e-01 -6.58211589e-01 6.12708963e-02 -7.44252205e-02 -3.99292290e-01 1.27455389e+00 8.28986406e-01 -3.53016853e-01 4.59980637e-01 -1.34045529e+00 -5.05180120e-01 6.36814773e-01 -4.35048699e-01 5.63501120e-01 -1.97969884e-01 1.60685882e-01 2.63613671e-01 -6.71219289e-01 3.49654913e-01 6.19817674e-01 6.86430514e-01 6.86340749e-01 -1.14895511e+00 -1.19213521e+00 -1.14632867e-01 1.86984807e-01 -1.46381199e+00 -4.15553808e-01 6.31498694e-01 -2.27543563e-01 1.10153341e+00 -8.47588107e-02 8.13902080e-01 9.96535957e-01 5.30096710e-01 4.03917551e-01 1.25186193e+00 -6.63897172e-02 3.93051863e-01 -2.42487371e-01 3.76682043e-01 1.68377191e-01 7.09112465e-01 -9.58353654e-02 -4.79115576e-01 2.34304711e-01 1.02231705e+00 1.80821255e-01 -4.28715676e-01 3.43199670e-01 -1.03093815e+00 1.53177604e-01 7.19951212e-01 5.08376002e-01 -3.40477288e-01 4.30771172e-01 2.10544914e-01 5.99216558e-02 -2.40708664e-01 2.64294863e-01 -2.21212775e-01 -2.49842554e-01 -7.41863310e-01 1.90110244e-02 5.67608953e-01 9.02782500e-01 5.10177791e-01 4.94370222e-01 9.12432894e-02 7.31631756e-01 -1.21576801e-01 7.43014514e-01 7.58459628e-01 -5.85056782e-01 3.53145272e-01 5.94274938e-01 -2.14711502e-01 -7.82623708e-01 -4.41245794e-01 -5.85868359e-01 -1.25593865e+00 2.09453404e-01 6.23510778e-01 -7.71355405e-02 -9.56633925e-01 1.79712522e+00 -4.16843802e-01 3.36632580e-01 1.36985511e-01 9.37447071e-01 8.92053246e-01 6.72109723e-01 7.10896626e-02 -2.53792882e-01 1.32865405e+00 -4.58146244e-01 -8.58763635e-01 -4.56765473e-01 1.05457999e-01 -4.41781670e-01 7.23286092e-01 2.31268823e-01 -9.62367594e-01 -4.64266032e-01 -1.27887022e+00 1.13071874e-01 -3.97857606e-01 1.91178307e-01 7.28607237e-01 3.30286384e-01 -8.42553675e-01 6.94660008e-01 -9.11086917e-01 -3.43283653e-01 6.74946308e-01 5.72436512e-01 -7.91550428e-02 1.96487963e-01 -1.07601023e+00 7.37623334e-01 2.21536294e-01 3.18319827e-01 -5.50161064e-01 -6.62837446e-01 -1.45911008e-01 1.21176615e-02 -1.40481487e-01 -3.73256952e-01 1.11029661e+00 -9.35301304e-01 -1.29144645e+00 6.20460093e-01 -3.10272604e-01 -6.66206479e-01 1.18432686e-01 1.75312340e-01 -4.96213198e-01 -4.46049944e-02 -1.64777011e-01 5.23765981e-01 2.96676844e-01 -8.21917295e-01 -2.88297683e-01 -5.91719151e-01 -3.55513930e-01 -2.18962818e-01 -5.38566232e-01 -3.39828841e-02 -2.37465113e-01 -5.42256713e-01 3.52729589e-01 -8.04739058e-01 -5.03777377e-02 6.24848679e-02 -1.71098635e-01 -1.04907691e-01 6.30524755e-01 -3.10381949e-01 1.22222352e+00 -2.21809793e+00 -2.65223294e-01 1.49742765e-02 2.32190397e-02 6.19413435e-01 3.98284495e-02 5.55639565e-02 -2.73106489e-02 7.06785694e-02 -4.44055974e-01 2.12166086e-01 -3.94307762e-01 -2.66698655e-02 -3.23553711e-01 2.77410954e-01 2.79615939e-01 1.02748466e+00 -5.35250783e-01 -3.55370827e-02 -8.47677886e-02 5.72689176e-01 -1.58886030e-01 -2.01486647e-01 1.86750785e-01 2.67537057e-01 -2.00100064e-01 7.34098911e-01 8.52700412e-01 -2.45207220e-01 5.26780970e-02 -4.07008380e-01 -4.65389520e-01 1.99061409e-01 -1.00599074e+00 1.45553529e+00 -2.73882877e-02 7.13796854e-01 -8.64446759e-02 -9.08754766e-01 1.23953545e+00 1.55147776e-01 2.19325483e-01 -1.18640637e+00 5.16716659e-01 8.42371643e-01 5.23379207e-01 -9.29359943e-02 -1.47400588e-01 -6.95132315e-02 6.07512444e-02 5.77208921e-02 1.41109452e-01 2.20072389e-01 1.76334843e-01 -8.88524055e-02 1.05697465e+00 -2.18287230e-01 1.71421155e-01 -5.81049860e-01 1.72117636e-01 -3.63905847e-01 6.32928312e-01 5.91518998e-01 -3.41772586e-01 4.46615547e-01 4.62874621e-01 -2.15960920e-01 -9.56685424e-01 -9.74923432e-01 -5.30137718e-01 2.57408798e-01 4.21302229e-01 2.76636314e-02 -8.17329407e-01 3.17587137e-01 -7.59712905e-02 4.11990047e-01 -3.16525817e-01 -2.53129750e-01 -6.04254544e-01 -1.07817769e+00 9.41992164e-01 7.61901975e-01 9.39845800e-01 -1.31070590e+00 -1.01903939e+00 3.48126143e-01 3.06419544e-02 -1.28015459e+00 5.49148722e-03 6.28205538e-01 -1.11777747e+00 -8.65635037e-01 -5.85896969e-01 -1.01257062e+00 7.61005223e-01 5.95774055e-02 6.14703774e-01 -7.17594149e-03 -4.63773102e-01 -4.37945426e-01 1.06621623e-01 -6.54939175e-01 3.14678341e-01 -4.43091020e-02 1.40443087e-01 -2.16270015e-01 5.58362901e-01 -1.07816601e+00 -7.52231240e-01 2.79201329e-01 -7.74638593e-01 1.48272678e-01 6.68203771e-01 5.65976501e-01 9.00807679e-01 -7.78943747e-02 9.63086188e-01 -4.71024871e-01 3.34056586e-01 -3.04458052e-01 -8.02531719e-01 -5.71523495e-02 -6.76536560e-01 -1.88650824e-02 1.04220808e+00 -6.55013621e-01 -6.24884903e-01 -8.59726779e-03 -2.35704139e-01 -3.10834050e-01 -1.12328961e-01 1.71914458e-01 -1.33633772e-02 -1.81553096e-01 6.05461836e-01 5.23227274e-01 -9.94326249e-02 -2.88222194e-01 -5.57343304e-01 5.89153647e-01 7.15376079e-01 -2.83236831e-01 4.80764389e-01 5.05973697e-01 1.95333958e-02 -7.01687038e-01 -2.60755420e-01 -7.09055811e-02 -4.61833067e-02 -1.69860587e-01 5.73780715e-01 -9.32554960e-01 -9.05842960e-01 1.20796883e+00 -1.16186428e+00 -3.47653419e-01 7.84061253e-02 4.89250422e-01 -1.79521680e-01 -1.08821161e-01 -8.46141636e-01 -8.35672379e-01 -8.11294854e-01 -1.00776660e+00 2.80993849e-01 8.18558335e-01 2.02346593e-01 -4.42402005e-01 -3.13260168e-01 -2.28538722e-01 7.46515453e-01 1.82169333e-01 6.24599993e-01 -5.47271311e-01 -5.79090476e-01 -5.33446632e-02 -5.05093634e-01 7.29963183e-02 -1.00967757e-01 -1.13768941e-02 -1.17414296e+00 -1.41817972e-01 9.84708294e-02 -1.55127570e-01 1.11858881e+00 6.25439167e-01 1.42587984e+00 4.38832082e-02 -4.21978623e-01 8.27752233e-01 1.90795732e+00 6.59928739e-01 1.10600948e+00 2.91972339e-01 4.19325173e-01 -6.80835033e-03 2.05276497e-02 3.67959470e-01 7.15498626e-02 4.78496313e-01 4.79122132e-01 8.64702538e-02 -1.18940748e-01 -7.03607202e-02 1.49328619e-01 9.37079549e-01 -2.69120365e-01 -2.56833404e-01 -9.44822311e-01 5.10553658e-01 -1.63537633e+00 -8.36227834e-01 -4.39794838e-01 2.28215957e+00 7.89159179e-01 4.69156742e-01 -2.31961995e-01 3.08223486e-01 9.13270235e-01 -1.60286963e-01 -9.85983729e-01 -2.06237093e-01 -5.33316731e-01 3.75142276e-01 6.95017457e-01 -1.06842781e-03 -5.99956989e-01 5.61052322e-01 5.33597612e+00 7.97615111e-01 -1.46854520e+00 -7.25448579e-02 4.15251046e-01 -2.51307160e-01 -1.03597619e-01 -2.48446867e-01 -1.13137519e+00 1.00008929e+00 1.08741820e+00 -1.41189218e-01 6.90530360e-01 4.01807606e-01 -1.44481033e-01 -1.41662270e-01 -8.17959845e-01 1.26917672e+00 -2.20232576e-01 -1.31353843e+00 -2.65020374e-02 -1.27088830e-01 4.89084095e-01 2.35098422e-01 -6.01713806e-02 1.45618185e-01 -3.42372477e-01 -1.05849552e+00 7.10661829e-01 5.13362527e-01 9.97370660e-01 -6.14880025e-01 8.84953380e-01 3.75611752e-01 -1.22625816e+00 -2.48103186e-01 -5.36593199e-01 -4.05462593e-01 -9.77158993e-02 9.15353417e-01 -1.67601168e-01 1.72632784e-01 8.86659563e-01 7.19091058e-01 -4.32674676e-01 1.29664993e+00 -7.06843333e-03 5.86194456e-01 -5.47085106e-01 -4.91765916e-01 -1.63770229e-01 -1.55696109e-01 2.66630173e-01 9.83402908e-01 6.81086183e-01 4.55868542e-01 -6.01945162e-01 1.09421992e+00 -2.98340112e-01 -1.87236428e-01 -4.89668220e-01 -1.62633479e-01 9.00642037e-01 1.24459815e+00 -9.62879896e-01 -1.08714171e-01 -2.69890755e-01 7.03770459e-01 1.96911663e-01 3.30972344e-01 -9.03521895e-01 -7.90331483e-01 5.17782569e-01 1.28939211e-01 2.93933332e-01 -8.45005438e-02 -8.01509857e-01 -1.00715542e+00 3.47898811e-01 -3.89030963e-01 -3.49315889e-02 -7.29280889e-01 -9.10131216e-01 9.27494407e-01 -4.99942303e-01 -1.20004737e+00 1.38009563e-01 -8.05916667e-01 -7.68997669e-01 7.25746155e-01 -1.62238157e+00 -4.75045621e-01 -5.21099627e-01 5.78544259e-01 1.78365842e-01 -1.52593842e-02 8.58187616e-01 5.50028622e-01 -8.42554152e-01 7.19354808e-01 2.14195207e-01 2.51490086e-01 3.93400222e-01 -6.21184587e-01 3.44875842e-01 9.09757376e-01 -3.91000450e-01 5.92511535e-01 4.55953538e-01 -4.68693525e-01 -1.37000763e+00 -8.45839024e-01 7.99027383e-01 2.09746361e-01 5.58640659e-01 -5.26756525e-01 -1.09246933e+00 3.02048385e-01 5.40252961e-02 1.53754935e-01 3.45438063e-01 -5.76546252e-01 -3.73361886e-01 -5.94449699e-01 -1.06099319e+00 6.43254280e-01 1.22616386e+00 -2.96179563e-01 -2.17944771e-01 -1.69598058e-01 3.74152124e-01 -2.67761588e-01 -6.79682374e-01 6.79947257e-01 6.37282729e-01 -1.04735518e+00 6.52785897e-01 5.47428653e-02 4.52089936e-01 -4.94479597e-01 -1.76314577e-01 -8.95322859e-01 -1.47328213e-01 -2.90266126e-01 2.18293294e-02 1.20459127e+00 3.35552126e-01 -1.11197591e+00 4.52258080e-01 3.19999069e-01 -2.08347395e-01 -1.02757251e+00 -1.10376620e+00 -8.87091219e-01 2.18784455e-02 -1.85135752e-01 4.58100110e-01 4.24581051e-01 -3.89229096e-02 3.60099114e-02 8.29018503e-02 9.60350260e-02 6.06443465e-01 1.57512471e-01 1.29710779e-01 -1.26392436e+00 1.48015261e-01 -6.25890255e-01 -6.67951226e-01 -7.91540205e-01 -1.76412746e-01 -6.59941792e-01 1.59770012e-01 -1.46752810e+00 2.49554113e-01 -3.87574762e-01 -6.86187208e-01 6.06667638e-01 2.56521925e-02 5.37848651e-01 1.50669023e-01 3.36832166e-01 -4.33431149e-01 2.98625171e-01 1.01540148e+00 1.99997172e-01 4.70496677e-02 -1.92677394e-01 -6.24657810e-01 6.00223243e-01 1.30248737e+00 -4.96183783e-01 -2.59490520e-01 -5.62677443e-01 1.00258365e-01 -8.17338452e-02 5.37151694e-01 -1.65217257e+00 6.40363455e-01 1.22609571e-01 6.08420193e-01 -3.70963126e-01 2.61998773e-01 -7.27979779e-01 4.29932743e-01 8.38788390e-01 -4.05153669e-02 -7.64610572e-03 5.25350511e-01 3.54840577e-01 -1.86168492e-01 -3.67473625e-02 9.41964030e-01 6.65238202e-02 -6.58247113e-01 2.71448940e-01 -4.79321569e-01 9.01072323e-02 9.52566087e-01 -4.63988334e-01 -7.73388982e-01 5.41993268e-02 -6.63308054e-02 3.58310379e-02 2.96654701e-01 1.86769694e-01 7.17848301e-01 -1.33856344e+00 -4.33616698e-01 5.85043490e-01 -9.85714272e-02 -1.15424402e-01 3.51742357e-01 9.98499572e-01 -2.72669375e-01 5.24395585e-01 -7.66595721e-01 -5.44742703e-01 -9.33463395e-01 1.73314184e-01 5.38913429e-01 1.02501057e-01 -2.78441131e-01 8.83377790e-01 2.45799012e-02 2.86485441e-02 2.47200564e-01 -4.05384094e-01 -8.71474892e-02 -1.92676544e-01 5.04203022e-01 2.71345675e-01 2.58022517e-01 -2.52930105e-01 -6.63097382e-01 7.34454334e-01 1.10315762e-01 1.66106805e-01 1.34526491e+00 2.30052143e-01 -2.31179744e-01 5.21096647e-01 1.01432335e+00 -4.90048021e-01 -1.28010869e+00 2.60212482e-03 -3.53700906e-01 -3.95268202e-02 -1.68123811e-01 -8.46025407e-01 -1.24447799e+00 1.04539633e+00 8.60523641e-01 -1.03723109e-01 1.50988531e+00 -2.77195364e-01 1.04533923e+00 4.15978044e-01 5.81876159e-01 -9.37879324e-01 -1.35371849e-01 5.39009273e-01 6.84605122e-01 -8.54925275e-01 -3.14216554e-01 -1.23415008e-01 -2.97317475e-01 1.08273256e+00 1.10503626e+00 -4.08799559e-01 6.43942475e-01 7.22303867e-01 -6.16863146e-02 1.23386942e-01 -6.86499596e-01 5.64242043e-02 -2.30866492e-01 3.37712377e-01 2.57842839e-01 9.55920592e-02 -5.53062499e-01 1.09059572e+00 -8.98033530e-02 4.13256615e-01 4.06907797e-01 7.70567358e-01 -5.10548353e-01 -5.91166973e-01 -1.01605766e-01 7.38987029e-01 -5.49849033e-01 -3.78658563e-01 2.77834814e-02 5.76136231e-01 9.31205451e-02 6.46176159e-01 3.25121433e-01 -5.60828269e-01 4.02508050e-01 7.22354800e-02 4.82114375e-01 -3.55404951e-02 -5.93646705e-01 -9.56959557e-03 -3.97622466e-01 -4.39310789e-01 -2.45675907e-01 -3.17011595e-01 -1.90577888e+00 -4.11777943e-01 -2.53220737e-01 -6.10553361e-02 8.95584404e-01 7.19345629e-01 5.30354679e-01 7.14302123e-01 1.97114751e-01 -6.49657786e-01 -4.12389100e-01 -9.10579741e-01 -6.89501345e-01 2.38332555e-01 1.19444259e-01 -4.48460191e-01 -5.19348562e-01 -1.74053743e-01]
[8.242831230163574, 2.5207738876342773]
33cf9278-71eb-41b2-9661-55f518880bf0
ams-net-adaptive-multiscale-sparse-neural
2207.11735
null
https://arxiv.org/abs/2207.11735v1
https://arxiv.org/pdf/2207.11735v1.pdf
AMS-Net: Adaptive Multiscale Sparse Neural Network with Interpretable Basis Expansion for Multiphase Flow Problems
In this work, we propose an adaptive sparse learning algorithm that can be applied to learn the physical processes and obtain a sparse representation of the solution given a large snapshot space. Assume that there is a rich class of precomputed basis functions that can be used to approximate the quantity of interest. We then design a neural network architecture to learn the coefficients of solutions in the spaces which are spanned by these basis functions. The information of the basis functions are incorporated in the loss function, which minimizes the differences between the downscaled reduced order solutions and reference solutions at multiple time steps. The network contains multiple submodules and the solutions at different time steps can be learned simultaneously. We propose some strategies in the learning framework to identify important degrees of freedom. To find a sparse solution representation, a soft thresholding operator is applied to enforce the sparsity of the output coefficient vectors of the neural network. To avoid over-simplification and enrich the approximation space, some degrees of freedom can be added back to the system through a greedy algorithm. In both scenarios, that is, removing and adding degrees of freedom, the corresponding network connections are pruned or reactivated guided by the magnitude of the solution coefficients obtained from the network outputs. The proposed adaptive learning process is applied to some toy case examples to demonstrate that it can achieve a good basis selection and accurate approximation. More numerical tests are performed on two-phase multiscale flow problems to show the capability and interpretability of the proposed method on complicated applications.
['Guang Lin', 'Wing Tat Leung', 'Yating Wang']
2022-07-24
null
null
null
null
['sparse-learning']
['methodology']
[ 6.67348853e-04 -1.27745584e-01 -2.70489007e-02 2.32218765e-02 -1.78645030e-01 -5.67554273e-02 3.29075269e-02 -5.57504259e-02 -7.41158202e-02 9.35913205e-01 -5.36294505e-02 1.50023699e-01 -5.89865148e-01 -7.16982007e-01 -5.91801584e-01 -1.11920774e+00 -1.92284763e-01 9.83244628e-02 -1.75732777e-01 -1.74030125e-01 1.42028853e-01 6.37144744e-01 -1.36999738e+00 2.01492384e-01 8.29008579e-01 1.01400316e+00 7.05952719e-02 3.46367687e-01 -5.02901189e-02 7.81900406e-01 -2.76750654e-01 4.90647346e-01 6.06715679e-01 -5.18430471e-01 -3.29958677e-01 2.72948653e-01 7.76699409e-02 -1.63984969e-01 -5.99012911e-01 1.01811290e+00 4.68682557e-01 7.61973262e-01 5.63904762e-01 -6.88956618e-01 -1.98849529e-01 3.31950903e-01 -4.26256984e-01 2.70860463e-01 -1.87815055e-02 2.11522520e-01 6.84995115e-01 -1.14040649e+00 5.60884237e-01 1.04174662e+00 7.12043881e-01 1.62565500e-01 -1.20542431e+00 -4.84713972e-01 8.68633837e-02 2.76286840e-01 -1.47228968e+00 -4.99260277e-01 1.43051291e+00 -5.00817180e-01 5.21287978e-01 3.10781240e-01 6.34427488e-01 4.16330934e-01 3.04250896e-01 2.69101560e-01 8.15873563e-01 -4.77673680e-01 4.11250323e-01 1.93660021e-01 3.55575860e-01 7.62509048e-01 3.04415345e-01 1.03908166e-01 -5.11365056e-01 -3.70094031e-01 9.73331153e-01 3.03492010e-01 -7.46347725e-01 -3.99505705e-01 -8.64075541e-01 9.09467161e-01 6.90944195e-01 3.96623284e-01 -8.30426276e-01 -2.35031232e-01 1.90708801e-01 2.04891294e-01 3.14104885e-01 5.98121881e-01 -1.63867295e-01 3.40153039e-01 -8.67241561e-01 3.31087224e-02 6.80262804e-01 3.03027689e-01 1.12827778e+00 5.51406026e-01 -1.56558186e-01 8.60687971e-01 5.63867986e-02 3.32522899e-01 4.96172726e-01 -1.23378754e+00 2.01609433e-01 4.35334384e-01 2.68089533e-01 -1.35638833e+00 -1.38194293e-01 -7.17513204e-01 -1.37120867e+00 3.36757302e-01 2.67752022e-01 -4.25757706e-01 -6.07088685e-01 1.41043019e+00 6.40613616e-01 6.43993139e-01 1.23861451e-02 1.33679283e+00 4.28045064e-01 1.15799904e+00 -3.30464274e-01 -7.37352908e-01 7.56510019e-01 -6.81407452e-01 -7.31415093e-01 3.00824326e-02 3.14170092e-01 -6.31982028e-01 6.89332366e-01 4.10670578e-01 -1.10283422e+00 -7.54236758e-01 -9.57401156e-01 4.54435825e-01 1.04617879e-01 2.61741608e-01 4.24949914e-01 -1.66541547e-01 -5.25935471e-01 1.11511600e+00 -9.24550593e-01 1.92442879e-01 -2.26855278e-02 7.53087327e-02 -6.51634932e-02 6.45882357e-03 -1.16208971e+00 5.67046642e-01 3.06283712e-01 5.34386635e-01 -7.26759195e-01 -8.74722540e-01 -4.93415713e-01 3.68750006e-01 2.25883663e-01 -7.36196816e-01 4.64671135e-01 -1.20893049e+00 -1.41751993e+00 5.84558435e-02 -2.43552446e-01 -1.78014234e-01 1.52665898e-01 4.13039438e-02 -2.32481346e-01 4.96228516e-01 9.69161466e-02 -2.09852457e-01 1.26918137e+00 -1.11343634e+00 -3.38085890e-01 1.03518724e-01 -1.13134876e-01 3.06547880e-01 -5.89790106e-01 -5.61612546e-01 -7.03571886e-02 -7.77463734e-01 4.72918153e-01 -7.64554024e-01 -5.82440794e-01 1.79849304e-02 -2.34769315e-01 2.46914208e-01 8.43712509e-01 -7.91583240e-01 1.34343231e+00 -2.26182818e+00 6.49277866e-01 5.90305805e-01 2.41733000e-01 1.50535911e-01 7.82557130e-02 4.53054726e-01 -1.32264897e-01 -4.07666564e-01 -5.23807645e-01 -3.62604707e-02 -5.43750167e-01 1.85959637e-01 -3.95205975e-01 5.05903780e-01 -8.25301781e-02 1.40070513e-01 -5.25562227e-01 -2.10176677e-01 3.28447759e-01 6.75576031e-01 -4.69514281e-01 4.74099934e-01 1.49003804e-01 6.22306466e-01 -7.95738459e-01 2.38346443e-01 6.05984092e-01 -4.14210379e-01 1.39775947e-01 -6.19593918e-01 -2.74595767e-01 -1.34571493e-01 -1.79941499e+00 1.15700066e+00 -6.46390021e-01 2.79516548e-01 6.88536048e-01 -1.59818697e+00 9.94131267e-01 3.26950073e-01 8.15832436e-01 -2.78650969e-01 5.25422730e-02 1.68185219e-01 -2.13549659e-02 -6.46998942e-01 -3.93119454e-02 -2.46050283e-01 3.63048553e-01 2.43336588e-01 -1.48110777e-01 1.67314082e-01 1.85521573e-01 -4.40775268e-02 7.84297347e-01 -3.40420038e-01 4.25238460e-01 -4.64773983e-01 1.10821974e+00 -5.78329153e-02 9.11399662e-01 6.50869310e-01 1.73067510e-01 3.82499844e-01 1.09904565e-01 -9.32216763e-01 -1.02585626e+00 -5.61751962e-01 -1.72600299e-01 6.13614976e-01 1.16223060e-01 -8.47945735e-02 -4.45021510e-01 -2.51493722e-01 2.47238800e-02 3.48039567e-01 -4.97782171e-01 -4.24624324e-01 -9.20519292e-01 -7.29352653e-01 -1.71710208e-01 1.79360509e-01 4.67717528e-01 -1.06901765e+00 -6.17419124e-01 3.43558043e-01 -3.14289667e-02 -7.63451874e-01 -1.82008296e-01 2.27226049e-01 -1.12708294e+00 -1.04434705e+00 -6.05777919e-01 -8.10548007e-01 9.38359618e-01 2.39424169e-01 5.79144061e-01 2.05416664e-01 -1.52433619e-01 2.29280829e-01 5.40270051e-03 3.63212317e-01 -1.84704453e-01 -1.32551491e-01 3.25585514e-01 6.34643137e-01 -3.84032190e-01 -9.99027371e-01 -4.99648064e-01 -5.35356142e-02 -6.64251089e-01 -2.37540901e-02 5.34720302e-01 1.06240404e+00 7.97024369e-01 3.27852339e-01 5.60608745e-01 -5.94465375e-01 6.66051924e-01 -5.99582255e-01 -7.24913120e-01 -3.24493907e-02 -3.88726711e-01 3.11931908e-01 1.20096469e+00 -7.21118271e-01 -1.12864709e+00 1.09442867e-01 2.10417330e-01 -1.01708698e+00 7.05960318e-02 7.87049592e-01 8.70554373e-02 -4.81925488e-01 5.01427829e-01 3.69682431e-01 7.03461766e-02 -6.27631485e-01 9.88226831e-02 2.73684084e-01 3.06258887e-01 -7.69648314e-01 7.94592679e-01 3.30983311e-01 5.42105317e-01 -1.16266978e+00 -8.27715039e-01 -4.48181361e-01 -3.63740206e-01 -3.72737676e-01 3.99690062e-01 -6.48606837e-01 -6.71296000e-01 2.27842152e-01 -8.98469567e-01 -2.37874463e-02 -5.58964193e-01 7.18507171e-01 -2.53076583e-01 4.20297295e-01 -7.78278887e-01 -8.04091752e-01 -5.42355418e-01 -8.67087841e-01 3.83258700e-01 3.74222249e-01 -3.73288281e-02 -9.70296562e-01 1.77002162e-01 -1.82253033e-01 5.09793699e-01 2.53554195e-01 9.84330535e-01 -2.96082526e-01 -4.59351510e-01 -1.27526075e-01 2.22835809e-01 3.46543878e-01 2.02687606e-01 4.93823402e-02 -5.94883263e-01 -4.46844488e-01 6.43181801e-01 3.71100567e-02 7.89489567e-01 7.67641068e-01 1.03819036e+00 -8.32438707e-01 -4.44329321e-01 9.25774038e-01 1.50726712e+00 2.20903471e-01 1.82841927e-01 6.90899938e-02 8.62332761e-01 4.70185906e-01 1.43374383e-01 6.93444431e-01 -3.44435006e-01 3.62862140e-01 2.10569382e-01 -2.51953769e-02 5.85934147e-02 4.04190831e-02 1.71137974e-01 1.22245741e+00 -2.22415119e-01 2.21799567e-01 -5.15811920e-01 4.10297543e-01 -1.91344535e+00 -1.02733386e+00 3.07756737e-02 2.19318819e+00 6.24698937e-01 5.01259118e-02 -2.16597095e-01 2.05484286e-01 8.93180966e-01 2.83122092e-01 -7.70781338e-01 -2.38618180e-01 6.07908741e-02 2.03562379e-01 -2.82028727e-02 7.85605907e-01 -9.08518136e-01 3.37689877e-01 5.94715405e+00 6.61579669e-01 -1.60181963e+00 -1.40175387e-01 6.04766428e-01 -1.06773973e-01 -1.15317525e-04 5.80477081e-02 -5.38410723e-01 5.45919180e-01 6.08131766e-01 -1.86489731e-01 7.92618454e-01 1.00352883e+00 5.59413314e-01 1.92594707e-01 -6.72667444e-01 9.68046129e-01 -3.29283088e-01 -1.67601013e+00 3.29513550e-02 -3.40389997e-01 8.41536939e-01 -3.46860290e-01 -6.41722158e-02 1.20749377e-01 -2.21929759e-01 -6.56035364e-01 3.15787107e-01 1.05071139e+00 3.77642095e-01 -6.65849864e-01 5.18513143e-01 6.01795852e-01 -1.18641949e+00 -3.30650598e-01 -5.50888419e-01 -3.64535660e-01 2.51609057e-01 1.01674414e+00 -3.26737553e-01 4.08062726e-01 6.06017649e-01 9.87044096e-01 1.74092930e-02 1.01630545e+00 2.98237950e-02 7.25678205e-01 -5.60843110e-01 1.31315827e-01 2.44885355e-01 -8.39592278e-01 9.67757821e-01 6.77501082e-01 6.11257970e-01 5.64976573e-01 4.35640574e-01 1.01999581e+00 1.45255536e-01 1.72408208e-01 -5.35889745e-01 2.81116784e-01 4.09642637e-01 1.58443582e+00 -5.67030370e-01 -4.11369354e-01 -2.96471298e-01 4.58808243e-01 2.63068914e-01 9.61747408e-01 -4.75285530e-01 -2.55589962e-01 6.22512043e-01 2.76939422e-01 4.61836785e-01 -2.39149868e-01 -1.61564887e-01 -1.26028633e+00 2.26153620e-02 -7.27225721e-01 4.90245938e-01 -6.17252529e-01 -1.32271397e+00 5.05614281e-01 1.22180404e-02 -1.30489922e+00 -2.98757643e-01 -3.13653767e-01 -1.06735492e+00 9.97533917e-01 -1.21375549e+00 -5.37836730e-01 -3.78515750e-01 6.90091968e-01 3.14060569e-01 -3.01964074e-01 5.19780815e-01 2.79402763e-01 -7.82163799e-01 1.34565189e-01 4.38842922e-01 -4.95261513e-03 2.05353498e-01 -6.30857646e-01 -6.25686407e-01 8.93471062e-01 -1.71823412e-01 7.76895523e-01 8.45167935e-01 -6.13094330e-01 -1.32476437e+00 -9.01992977e-01 3.32009941e-01 4.93141979e-01 3.55741948e-01 1.23071484e-01 -1.37655652e+00 4.95377213e-01 -1.89684499e-02 4.03807372e-01 2.12702096e-01 -1.85981408e-01 1.77062675e-01 -5.27233899e-01 -1.10045302e+00 3.48591715e-01 3.43854368e-01 -2.06229493e-01 -4.45876420e-01 3.44688594e-01 3.48621666e-01 -2.51205176e-01 -8.05753171e-01 5.24057329e-01 2.58524150e-01 -7.84436285e-01 1.15164161e+00 -8.24460804e-01 3.22599083e-01 -4.92983431e-01 4.80105914e-02 -1.36984241e+00 -7.36014843e-01 -8.29162002e-01 -6.35426879e-01 8.64152193e-01 -4.72478271e-02 -7.24687994e-01 7.09777117e-01 4.34454650e-01 -1.13435894e-01 -1.08874810e+00 -1.00697446e+00 -3.75362933e-01 -3.77617963e-02 1.68454826e-01 2.65505701e-01 1.18806696e+00 -3.67706493e-02 4.15095687e-01 -4.69195843e-01 5.37520468e-01 8.54126513e-01 5.51222861e-01 2.62721241e-01 -1.00600302e+00 -4.69144046e-01 -2.61362135e-01 1.66382995e-02 -8.44282448e-01 2.48746008e-01 -7.00234413e-01 -1.84809655e-01 -1.24585533e+00 -1.62962377e-01 -5.19171357e-01 -3.88215184e-01 2.40006387e-01 -2.00521782e-01 -2.24676311e-01 7.36828148e-02 6.05641782e-01 -4.98058051e-02 7.78769732e-01 1.21026385e+00 -8.75801742e-02 -4.91276652e-01 6.54225945e-02 -1.73558161e-01 8.64934623e-01 6.82026863e-01 -4.45383191e-01 -2.98862696e-01 -1.83375984e-01 -1.73327520e-01 5.28491259e-01 2.44547173e-01 -1.23133707e+00 4.10887659e-01 -4.29917485e-01 4.57257211e-01 -1.36163250e-01 4.68242288e-01 -9.23103631e-01 3.54793876e-01 5.12115300e-01 -3.85681778e-01 -2.41753012e-01 9.73078907e-02 3.86734486e-01 -3.49552065e-01 -4.68558788e-01 9.98136699e-01 -2.64694095e-01 -4.52684969e-01 2.83169091e-01 -4.35565382e-01 -1.55261815e-01 7.11488128e-01 6.58610230e-03 3.08718652e-01 -4.72594231e-01 -9.94143486e-01 1.34981796e-01 3.70676108e-02 -2.32141614e-01 6.00880921e-01 -1.52783144e+00 -4.86194134e-01 6.33673489e-01 -4.46234405e-01 -5.60487956e-02 5.13944983e-01 7.49793589e-01 -4.54334408e-01 1.52286842e-01 -3.87204140e-01 -4.91085052e-01 -6.98275626e-01 5.91972768e-01 6.97056115e-01 -2.33636513e-01 -7.96201169e-01 3.86194736e-01 6.00981340e-02 -1.35316566e-01 -1.89116281e-02 -2.54309237e-01 -2.75853127e-01 -1.07539453e-01 3.87177348e-01 7.08376467e-01 -1.40082896e-01 -6.48566484e-01 -9.17316079e-02 8.77507985e-01 3.36539298e-01 1.29045665e-01 1.59054351e+00 -7.09997350e-03 -3.76671880e-01 2.84364581e-01 1.32554102e+00 7.55418837e-02 -1.45375752e+00 -3.13206077e-01 -5.65100074e-01 -2.49456897e-01 3.03525865e-01 -1.78505629e-01 -1.50672734e+00 7.66130745e-01 3.84192407e-01 2.38886818e-01 1.24730694e+00 -3.67453516e-01 7.67569423e-01 5.90497136e-01 -1.96507294e-03 -9.75401580e-01 -3.13968435e-02 4.03718382e-01 8.87390971e-01 -8.19701970e-01 1.62452936e-01 -3.06386828e-01 -8.60912129e-02 1.44413424e+00 6.41985238e-01 -6.37100279e-01 8.42901587e-01 1.70643732e-01 -1.30967453e-01 -8.35299194e-02 -4.44292873e-01 3.39009494e-01 4.29953128e-01 2.30970737e-02 1.09688692e-01 -2.68597215e-01 -4.19949114e-01 6.14970326e-01 2.36002088e-01 -1.38131008e-01 4.43951339e-01 6.61192417e-01 -5.84368348e-01 -5.68954408e-01 -6.32721305e-01 3.40950698e-01 -5.98112494e-02 2.46787950e-01 2.43729278e-01 6.01297259e-01 -1.16016483e-02 4.61562335e-01 -2.32724383e-01 -6.41968623e-02 2.71252662e-01 1.20675109e-01 9.97141972e-02 -4.01134610e-01 -2.81078190e-01 2.29616627e-01 -2.70525962e-01 -5.31561434e-01 -1.81617424e-01 -4.77732539e-01 -1.36843348e+00 -1.01362526e-01 -1.30290359e-01 6.84165835e-01 -3.10278460e-02 8.53058159e-01 4.40327287e-01 5.90783954e-01 7.87486613e-01 -1.15048361e+00 -5.69072485e-01 -7.47521520e-01 -6.09297037e-01 4.19912279e-01 5.98386228e-01 -7.31555998e-01 -8.15963805e-01 9.25115347e-02]
[6.532783508300781, 3.4596073627471924]
ac15bd23-607a-425a-9511-40716e30ff28
data-augmentation-for-seizure-prediction-with
2306.08256
null
https://arxiv.org/abs/2306.08256v1
https://arxiv.org/pdf/2306.08256v1.pdf
Data Augmentation for Seizure Prediction with Generative Diffusion Model
Objective: Seizure prediction is of great importance to improve the life of patients. The focal point is to distinguish preictal states from interictal ones. With the development of machine learning, seizure prediction methods have achieved significant progress. However, the severe imbalance problem between preictal and interictal data still poses a great challenge, restricting the performance of classifiers. Data augmentation is an intuitive way to solve this problem. Existing data augmentation methods generate samples by overlapping or recombining data. The distribution of generated samples is limited by original data, because such transformations cannot fully explore the feature space and offer new information. As the epileptic EEG representation varies among seizures, these generated samples cannot provide enough diversity to achieve high performance on a new seizure. As a consequence, we propose a novel data augmentation method with diffusion model called DiffEEG. Methods: Diffusion models are a class of generative models that consist of two processes. Specifically, in the diffusion process, the model adds noise to the input EEG sample step by step and converts the noisy sample into output random noise, exploring the distribution of data by minimizing the loss between the output and the noise added. In the denoised process, the model samples the synthetic data by removing the noise gradually, diffusing the data distribution to outward areas and narrowing the distance between different clusters. Results: We compared DiffEEG with existing methods, and integrated them into three representative classifiers. The experiments indicate that DiffEEG could further improve the performance and shows superiority to existing methods. Conclusion: This paper proposes a novel and effective method to solve the imbalanced problem and demonstrates the effectiveness and generality of our method.
['Xun Chen', 'Ruobing Qian', 'Aiping Liu', 'Le Wu', 'Yuchang Zhao', 'Kai Shu']
2023-06-14
null
null
null
null
['seizure-prediction']
['medical']
[ 2.32306737e-02 -2.48743996e-01 4.23080735e-02 -2.32985213e-01 -2.19371274e-01 -6.77849278e-02 4.39872861e-01 -1.84928596e-01 -1.23265542e-01 8.21466148e-01 3.97763491e-01 2.60550410e-01 -1.93678126e-01 -8.92738342e-01 -9.49284285e-02 -1.11340368e+00 -7.85221308e-02 3.73874247e-01 1.02667063e-01 -2.76121140e-01 1.41211748e-01 4.32762742e-01 -1.55452514e+00 2.89549142e-01 1.29400921e+00 8.27679157e-01 3.58182192e-01 -1.69174194e-01 -2.95778036e-01 2.38308996e-01 -1.05688143e+00 3.69164556e-01 2.09957421e-01 -8.83289158e-01 -3.79583657e-01 1.27852052e-01 -6.18858635e-01 3.13372677e-03 -4.82883215e-01 1.25690222e+00 8.35967898e-01 -8.64766017e-02 4.95179266e-01 -1.51171684e+00 -5.98853886e-01 6.67326033e-01 -6.28133893e-01 3.81311923e-01 1.67398393e-01 3.31049450e-02 7.84812402e-03 -6.86569750e-01 3.42550457e-01 8.41805637e-01 4.35487449e-01 8.16817343e-01 -1.23956716e+00 -1.16528153e+00 1.06452517e-01 3.46506119e-01 -1.40209508e+00 -1.11103430e-01 1.01921129e+00 -4.22260851e-01 4.77043539e-01 4.31019723e-01 1.29286563e+00 1.23188770e+00 3.23429376e-01 7.55203307e-01 1.32643831e+00 -5.43583557e-02 3.29124272e-01 2.46047787e-02 1.06337629e-01 -8.85500535e-02 2.19040245e-01 9.44125815e-04 -5.71185231e-01 -1.59042060e-01 6.71182752e-01 4.60063815e-01 -8.08843732e-01 -1.76219523e-01 -1.09166694e+00 5.92331588e-01 4.44861740e-01 8.44597936e-01 -3.96606445e-01 -4.87354547e-01 1.64019749e-01 2.96006829e-01 6.29718840e-01 5.42405546e-01 -2.03029081e-01 -1.66105956e-01 -1.18705404e+00 9.62371156e-02 3.71216714e-01 4.98536199e-01 6.39737189e-01 1.00402728e-01 -1.51313975e-01 8.80031943e-01 -9.05020609e-02 2.44213909e-01 1.34055138e+00 -3.66490148e-02 2.80236453e-01 1.08211935e+00 -4.66991395e-01 -9.36034203e-01 -5.14737427e-01 -8.42120528e-01 -1.33591545e+00 1.37812749e-01 -1.72776207e-02 -2.85348475e-01 -1.02293217e+00 1.70998025e+00 5.60926348e-02 2.38191321e-01 4.92202789e-02 7.65674531e-01 7.88258553e-01 6.54792845e-01 -1.35155082e-01 -6.28760040e-01 1.03654802e+00 -4.88533705e-01 -1.06295073e+00 -3.12051415e-01 4.57770258e-01 -5.12373030e-01 8.91845882e-01 7.02607989e-01 -8.29311311e-01 -4.07134116e-01 -1.01871991e+00 5.84120929e-01 -9.71278846e-02 1.32756472e-01 5.10702372e-01 3.40082228e-01 -7.95727432e-01 4.07359332e-01 -9.43934500e-01 -3.33302864e-03 6.21634245e-01 3.80989134e-01 -4.22506303e-01 9.81614180e-03 -1.28919721e+00 6.48915470e-01 5.82918584e-01 8.84021968e-02 -5.83094060e-01 -6.58619404e-01 -4.38876569e-01 1.00114755e-02 4.69039455e-02 -3.01517725e-01 6.17863059e-01 -1.02452540e+00 -1.13993680e+00 2.25462973e-01 -1.33088514e-01 -3.61352235e-01 5.31192839e-01 3.47753689e-02 -6.32452607e-01 -4.08775598e-01 9.83673483e-02 4.59889561e-01 7.63071358e-01 -1.04250133e+00 -4.78208333e-01 -5.87875783e-01 -6.20516181e-01 1.97001159e-01 -6.05032563e-01 -2.45768875e-01 -2.43704513e-01 -1.00243664e+00 5.60657799e-01 -6.79966748e-01 -2.39976317e-01 -5.85186481e-01 -3.27985376e-01 -2.39802688e-01 1.18594182e+00 -5.18851340e-01 1.55910683e+00 -2.43700624e+00 5.22636697e-02 3.64919752e-01 4.65900213e-01 2.07917437e-01 4.00723331e-02 2.83244103e-01 -4.14820641e-01 1.34053675e-03 -6.35322273e-01 -3.24950963e-02 -3.93150270e-01 7.31958002e-02 -2.84735829e-01 1.46463886e-01 7.44271502e-02 7.22488105e-01 -7.40427256e-01 -1.19314589e-01 -2.62098052e-02 5.45283556e-01 -4.97978717e-01 2.88181752e-01 1.73763022e-01 8.00977290e-01 -4.39091831e-01 2.50139028e-01 8.86489272e-01 1.54417798e-01 -1.31910935e-01 -7.92389587e-02 6.31181002e-02 3.53225111e-03 -1.31029952e+00 1.53202879e+00 6.97732642e-02 5.52160203e-01 -4.02695745e-01 -1.27538121e+00 1.49984264e+00 1.58636376e-01 6.01398885e-01 -7.09456682e-01 1.75126746e-01 2.71572083e-01 4.33680713e-01 -7.18357146e-01 5.87228090e-02 -2.10688829e-01 1.00089557e-01 4.03397322e-01 -2.30987847e-01 -1.76936746e-01 1.08982570e-01 -4.22810242e-02 9.59544778e-01 -1.96428165e-01 2.07562253e-01 -2.82728881e-01 2.69953042e-01 -1.86888903e-01 9.60176528e-01 2.26583466e-01 1.36509255e-01 6.01230860e-01 6.20119691e-01 -3.65805387e-01 -6.95434690e-01 -8.71281564e-01 -3.34778875e-01 2.31957749e-01 1.22161798e-01 -2.66550869e-01 -1.01916087e+00 -6.38914704e-01 -2.96991020e-01 6.28498375e-01 -7.58583486e-01 -6.87650502e-01 -3.98711652e-01 -1.48072457e+00 2.86969185e-01 5.37374735e-01 8.97949755e-01 -1.18432117e+00 -4.67177063e-01 2.72439003e-01 -3.82214248e-01 -4.35622185e-01 -3.33489269e-01 2.19873145e-01 -9.33978438e-01 -9.47643936e-01 -6.93164229e-01 -9.04921889e-01 8.99588168e-01 1.83759518e-02 5.52407861e-01 9.07376856e-02 -2.56781369e-01 -2.86621094e-01 -5.33045650e-01 -4.55416083e-01 -2.13286087e-01 -5.91913201e-02 2.01426178e-01 1.44407630e-01 6.39421582e-01 -9.16720986e-01 -7.59487867e-01 1.83563069e-01 -1.08689523e+00 1.52236193e-01 6.91912115e-01 9.02655661e-01 5.81173182e-01 4.33010727e-01 1.00926602e+00 -5.17282546e-01 1.15093672e+00 -6.82245493e-01 -1.95496127e-01 -1.37143657e-01 -6.09921515e-01 4.98322807e-02 6.79331720e-01 -7.37819612e-01 -8.00790250e-01 7.18935300e-03 -1.18476748e-01 -3.97359431e-01 -1.13356642e-01 3.66325438e-01 -6.45420611e-01 2.32802108e-01 5.21212816e-01 7.78018892e-01 1.99060127e-01 -4.43216413e-01 -2.14998826e-01 8.50263774e-01 4.29628581e-01 -7.11548775e-02 4.42955881e-01 2.43319511e-01 -2.13569432e-01 -6.51535034e-01 -1.18074238e-01 -4.84878346e-02 -4.21978056e-01 -5.17999753e-02 6.50318623e-01 -5.29619217e-01 -1.95950836e-01 6.40937209e-01 -8.20471287e-01 -1.44810989e-01 -5.79959929e-01 8.21586192e-01 -1.82457387e-01 1.96767241e-01 -1.97041929e-01 -5.67707419e-01 -3.02576542e-01 -1.43119466e+00 4.90322679e-01 2.94119775e-01 -2.50193208e-01 -6.45603776e-01 1.04956523e-01 -1.29634112e-01 4.40292776e-01 4.10274088e-01 9.35948789e-01 -9.99245703e-01 -2.54800737e-01 -2.91452557e-01 1.73015222e-01 4.53595161e-01 5.98745286e-01 -3.42183590e-01 -8.28192770e-01 -2.98693419e-01 6.80822670e-01 7.60710239e-02 7.95080841e-01 3.87523979e-01 1.48426187e+00 -2.58296460e-01 -6.11692727e-01 6.36104047e-01 1.02270472e+00 8.69349718e-01 8.34394693e-01 4.02810663e-01 1.96840733e-01 4.46329802e-01 3.01056623e-01 4.33445185e-01 2.73607834e-03 2.77100891e-01 2.61536926e-01 -2.43553042e-01 -2.18720227e-01 -3.31305005e-02 7.93518797e-02 9.03682411e-01 -9.39926729e-02 4.57671611e-03 -9.05965090e-01 5.84809542e-01 -1.57757401e+00 -9.20191169e-01 -6.67315945e-02 2.25021362e+00 8.92173111e-01 1.24217108e-01 -6.85622320e-02 6.57724142e-01 7.08843589e-01 -2.03422710e-01 -6.71196222e-01 1.23845376e-01 -3.38083267e-01 3.29295874e-01 -1.39477253e-01 1.43304355e-02 -7.35677838e-01 4.90580350e-01 5.74402142e+00 9.69711363e-01 -1.40817535e+00 4.38242704e-02 8.10841084e-01 -3.37135762e-01 -3.20887268e-01 -2.71003783e-01 -4.92687345e-01 1.04860353e+00 4.61873025e-01 -3.93356830e-01 3.80695432e-01 5.05640090e-01 3.33863765e-01 -1.24036700e-01 -9.49312449e-01 1.31883323e+00 1.17445283e-01 -1.10642505e+00 1.92732871e-01 1.87257364e-01 8.29137743e-01 -1.64248377e-01 -1.00370171e-02 1.83737323e-01 -1.26338154e-01 -1.07744312e+00 2.13007599e-01 9.19826865e-01 6.42029166e-01 -9.19743299e-01 9.70855594e-01 6.08571112e-01 -9.59597349e-01 -9.72291008e-02 -2.32238844e-01 1.87220741e-02 -1.02951258e-01 9.42818105e-01 -5.84385693e-01 4.59079772e-01 7.93873668e-01 6.31012559e-01 -6.53820693e-01 1.23646879e+00 -1.94264784e-01 5.44009984e-01 -2.76231498e-01 4.12304439e-02 -9.91809443e-02 -4.29531157e-01 6.87996149e-01 8.96893620e-01 6.95644200e-01 3.01558793e-01 9.48648453e-02 9.61412907e-01 -4.32621641e-03 1.61011562e-01 -5.88720918e-01 2.41535947e-01 5.83805561e-01 9.90635335e-01 -7.51078188e-01 -3.46109092e-01 -1.18670464e-01 8.64465356e-01 -7.48602375e-02 2.93808371e-01 -5.34584820e-01 -6.26340568e-01 3.33360791e-01 1.80478558e-01 -2.62932181e-01 2.51905948e-01 -5.12969673e-01 -1.20574713e+00 2.25446612e-01 -1.04729414e+00 1.09691516e-01 -5.36613226e-01 -1.14633501e+00 1.02830899e+00 -1.07932247e-01 -1.56005335e+00 -1.79416403e-01 4.68196794e-02 -8.16661835e-01 9.89150584e-01 -9.93001223e-01 -6.12649679e-01 -6.61563039e-01 6.96511507e-01 6.79207206e-01 -4.57242608e-01 6.92199051e-01 3.57622206e-01 -5.75330973e-01 4.85164285e-01 2.38264486e-01 6.48333281e-02 6.09497726e-01 -9.45019603e-01 1.01081029e-01 8.70838881e-01 -9.33462754e-02 5.60771227e-01 5.41197956e-01 -8.43548894e-01 -7.39315808e-01 -9.79616523e-01 5.09210587e-01 9.27639753e-02 4.45583969e-01 -3.99657816e-01 -1.40924048e+00 3.69779348e-01 1.41984954e-01 -8.70092437e-02 6.75049126e-01 -2.99656749e-01 3.49472612e-01 -2.87178993e-01 -1.20576608e+00 5.18787563e-01 8.93557787e-01 -2.33196001e-02 -8.61032784e-01 3.34247835e-02 3.22583139e-01 -2.49812603e-01 -6.84599817e-01 6.17212892e-01 1.62998885e-01 -9.33751106e-01 4.60892439e-01 -2.35540181e-01 2.20456257e-01 -3.45002145e-01 2.73522973e-01 -1.97491491e+00 -2.53219992e-01 -5.71161747e-01 8.01618546e-02 1.23267877e+00 2.62483954e-01 -1.02065051e+00 7.59148955e-01 4.64966297e-01 -2.19502911e-01 -1.09982538e+00 -8.41940522e-01 -6.83323026e-01 1.60296142e-01 -3.38610649e-01 1.15394890e+00 1.09923017e+00 3.77875328e-01 8.24481715e-03 -8.47172514e-02 -6.10810779e-02 3.57664973e-01 -1.08149543e-01 5.10155022e-01 -1.34735703e+00 5.02783619e-02 -5.19870043e-01 -6.80786371e-01 -7.57449567e-01 -1.68897331e-01 -1.00786698e+00 -1.68852627e-01 -1.65448415e+00 3.32247972e-01 -5.97784877e-01 -3.11964154e-01 5.83563626e-01 -2.54108459e-01 2.38984376e-01 -5.36519289e-02 5.16896486e-01 1.60192758e-01 8.07127595e-01 1.45851398e+00 -1.06488228e-01 -6.76986098e-01 1.28764242e-01 -7.44376779e-01 7.76639640e-01 1.12220013e+00 -5.00089645e-01 -6.85810804e-01 -2.06961200e-01 -2.30298281e-01 -1.17567360e-01 8.37572515e-02 -1.31397033e+00 2.23325059e-01 2.04277337e-02 8.34267855e-01 -5.22253394e-01 1.43833071e-01 -6.92027986e-01 3.19302827e-01 6.57401025e-01 1.50751248e-02 4.29100990e-02 2.54325271e-01 2.13932171e-01 -4.52751189e-01 -7.84222409e-02 7.22863495e-01 9.65811983e-02 -4.28742170e-01 4.39379424e-01 -4.56255823e-01 9.72016081e-02 1.28374398e+00 -4.34062868e-01 -9.84738022e-02 -2.23643914e-01 -1.10622311e+00 1.64599687e-01 1.83355570e-01 5.41828275e-01 8.13966751e-01 -1.47324610e+00 -7.35099792e-01 9.37178016e-01 -1.42789945e-01 1.45863906e-01 2.69986391e-01 9.44887578e-01 -7.89276138e-02 5.87095767e-02 -4.09302890e-01 -5.51256835e-01 -1.14358354e+00 3.77072871e-01 4.38183397e-01 -8.76110047e-02 -8.09387565e-01 5.58982611e-01 4.66369152e-01 -1.03002846e-01 6.75908104e-02 -4.24001008e-01 -6.02656603e-01 3.02183002e-01 7.21086979e-01 3.01579446e-01 2.78698266e-01 -2.74285972e-01 -2.43880853e-01 3.05503726e-01 -7.70941898e-02 6.86963722e-02 1.61873877e+00 1.61707953e-01 -3.69599700e-01 3.62411678e-01 9.12426770e-01 -6.01871535e-02 -1.03102803e+00 9.41002592e-02 -2.50762045e-01 -4.07953173e-01 3.43432426e-02 -8.85896862e-01 -1.39597416e+00 8.49560440e-01 9.61061537e-01 4.05519038e-01 1.66426706e+00 -4.17045318e-02 6.75801039e-01 -4.83499952e-02 2.72845775e-01 -9.19769287e-01 3.68745625e-02 7.95246735e-02 9.91520166e-01 -7.92674184e-01 -3.09866905e-01 -1.22297414e-01 -6.08481228e-01 1.06843901e+00 7.60038674e-01 -2.50770241e-01 6.77562177e-01 5.10914564e-01 -7.12283701e-02 -1.28231108e-01 -4.21326965e-01 8.52789506e-02 2.54033297e-01 7.39176154e-01 1.67916983e-01 -1.65640609e-03 -7.84189343e-01 1.14720023e+00 -5.75201631e-01 1.10187612e-01 4.77974534e-01 7.15540767e-01 -4.51792657e-01 -1.27096844e+00 -5.31955302e-01 8.29915702e-01 -2.23917022e-01 -1.18112274e-01 -3.11286390e-01 8.21829736e-01 6.23407006e-01 9.05217469e-01 2.92819858e-01 -5.67305088e-01 4.64480460e-01 1.17742278e-01 2.04755858e-01 -5.24868190e-01 -3.45285982e-01 3.96247655e-01 -5.63993454e-01 -2.00072765e-01 -2.38684744e-01 -7.08282590e-01 -1.40741074e+00 1.51390165e-01 -4.09065962e-01 5.99162698e-01 4.89086807e-01 9.19673443e-01 4.69034851e-01 9.79932308e-01 8.26765299e-01 -5.14365792e-01 -2.63791889e-01 -1.28616977e+00 -6.98528409e-01 3.46260577e-01 2.29632944e-01 -6.76464736e-01 -5.07343769e-01 -3.49090132e-03]
[13.123577117919922, 3.4492175579071045]
676f148b-95d1-46a6-9ce6-5d7f10fdeae1
contextual-visual-similarity
1612.02534
null
http://arxiv.org/abs/1612.02534v1
http://arxiv.org/pdf/1612.02534v1.pdf
Contextual Visual Similarity
Measuring visual similarity is critical for image understanding. But what makes two images similar? Most existing work on visual similarity assumes that images are similar because they contain the same object instance or category. However, the reason why images are similar is much more complex. For example, from the perspective of category, a black dog image is similar to a white dog image. However, in terms of color, a black dog image is more similar to a black horse image than the white dog image. This example serves to illustrate that visual similarity is ambiguous but can be made precise when given an explicit contextual perspective. Based on this observation, we propose the concept of contextual visual similarity. To be concrete, we examine the concept of contextual visual similarity in the application domain of image search. Instead of providing only a single image for image similarity search (\eg, Google image search), we require three images. Given a query image, a second positive image and a third negative image, dissimilar to the first two images, we define a contextualized similarity search criteria. In particular, we learn feature weights over all the feature dimensions of each image such that the distance between the query image and the positive image is small and their distances to the negative image are large after reweighting their features. The learned feature weights encode the contextualized visual similarity specified by the user and can be used for attribute specific image search. We also show the usefulness of our contextualized similarity weighting scheme for different tasks, such as answering visual analogy questions and unsupervised attribute discovery.
['Xiaofang Wang', 'Martial Hebert', 'Kris M. Kitani']
2016-12-08
null
null
null
null
['image-similarity-search']
['computer-vision']
[ 5.30605614e-01 -1.38457820e-01 -2.42995724e-01 -6.67603433e-01 -3.87282521e-01 -7.66163647e-01 6.67936444e-01 6.97898090e-01 -4.42152470e-01 1.39606133e-01 -1.06966663e-02 -2.25706607e-01 -1.59874365e-01 -6.77823544e-01 -5.83943665e-01 -5.51917255e-01 2.50372708e-01 1.97822481e-01 3.02636743e-01 -1.17838979e-01 6.46860063e-01 5.11306226e-01 -1.87865531e+00 3.18399072e-01 5.42326808e-01 1.07853794e+00 5.89384496e-01 4.41425890e-01 -2.54493862e-01 3.40048701e-01 -5.46393812e-01 -2.80748099e-01 3.48302513e-01 -6.23061538e-01 -1.09249604e+00 4.10079747e-01 9.02358711e-01 5.89945354e-02 1.90750007e-02 1.31753242e+00 1.65928140e-01 5.31987131e-01 7.37599850e-01 -1.54199600e+00 -1.14296246e+00 1.62776157e-01 -5.43125451e-01 2.44806424e-01 7.50122547e-01 1.04126008e-02 1.30944824e+00 -9.01087046e-01 6.53775811e-01 1.22018957e+00 3.10802817e-01 4.26007450e-01 -1.39613616e+00 -3.52075636e-01 4.56927240e-01 4.34358597e-01 -1.49658346e+00 -1.59129813e-01 6.85618877e-01 -3.91912878e-01 6.35220468e-01 5.28649509e-01 7.47517288e-01 5.99643409e-01 -2.02086076e-01 5.70411801e-01 1.12942243e+00 -5.93977988e-01 3.38068306e-01 5.41654587e-01 1.25464931e-01 3.75449657e-01 1.24061659e-01 -9.38324165e-03 -2.40267590e-01 -1.74339622e-01 4.31629896e-01 3.90373051e-01 -2.23918959e-01 -7.24423289e-01 -1.17219818e+00 8.61962199e-01 8.17758501e-01 5.94075322e-01 -1.80794865e-01 1.40907422e-01 1.56380355e-01 3.62936705e-01 -1.00989483e-01 6.87938094e-01 -2.62985360e-02 3.94547194e-01 -4.88811076e-01 2.14615315e-01 4.80113417e-01 1.06553757e+00 1.17746401e+00 -3.80307078e-01 -2.82569341e-02 8.17295730e-01 2.38749132e-01 4.56083804e-01 4.95463192e-01 -1.10453832e+00 8.96397531e-02 6.04062736e-01 3.99568155e-02 -1.32581806e+00 -9.20507535e-02 7.95604512e-02 -4.82909322e-01 3.56530577e-01 4.54178005e-01 6.08170271e-01 -9.01308119e-01 1.84397006e+00 3.63897175e-01 9.22283716e-03 1.06561944e-01 1.20602655e+00 8.82572949e-01 4.16602403e-01 4.59196210e-01 -1.25244912e-02 1.71478355e+00 -7.31652021e-01 -2.98521280e-01 -7.23198473e-01 4.06459272e-01 -9.59574401e-01 1.49997532e+00 -9.95194837e-02 -8.94028127e-01 -6.91028893e-01 -1.09494722e+00 -1.75718248e-01 -7.01405823e-01 -2.85946876e-01 4.26886529e-01 4.97928143e-01 -8.40320051e-01 4.23964143e-01 -6.42464980e-02 -1.04496861e+00 4.72963229e-02 1.22609111e-02 -6.46098971e-01 -1.90329120e-01 -1.10245204e+00 9.62094903e-01 4.02414560e-01 -4.79439884e-01 -5.18404365e-01 -6.13584459e-01 -9.63966727e-01 1.58807188e-01 2.46268928e-01 -8.17665815e-01 1.09579253e+00 -1.52399182e+00 -6.88121021e-01 1.43217576e+00 -3.31376255e-01 -1.60169676e-01 1.05597451e-01 1.62355691e-01 -5.43808460e-01 3.21409047e-01 4.66069877e-01 8.22538614e-01 1.01783788e+00 -1.60657573e+00 -8.18466783e-01 -5.00440180e-01 2.88223922e-01 4.65452194e-01 -1.48753494e-01 2.52111107e-02 -6.44915223e-01 -6.30629778e-01 2.34224513e-01 -9.08040166e-01 -3.01019102e-01 2.99856037e-01 -2.61966586e-01 -1.29098386e-01 7.89451540e-01 -4.56977226e-02 9.98421669e-01 -2.53188038e+00 -3.15267920e-01 6.78654909e-01 2.13779598e-01 -1.17679447e-01 -3.39113414e-01 3.10167998e-01 -4.79516774e-01 3.04756939e-01 -2.24549294e-01 2.25117803e-01 -1.90735217e-02 1.78160369e-01 -1.57208681e-01 2.33135790e-01 -2.79807076e-02 7.81927466e-01 -1.13955522e+00 -6.37534559e-01 1.57723427e-01 1.26960367e-01 -2.94301540e-01 1.76160872e-01 2.31072024e-01 1.47764400e-01 -4.52170074e-01 5.25832057e-01 5.63873291e-01 -4.50110197e-01 1.71037868e-01 -4.42659467e-01 1.38323814e-01 -1.75435662e-01 -1.27471685e+00 1.42157376e+00 -4.83284652e-01 7.53983498e-01 -4.47437257e-01 -1.00239289e+00 8.90224278e-01 3.81336734e-02 1.98702544e-01 -1.02383339e+00 -1.50196612e-01 -3.11283134e-02 -4.02931124e-02 -5.20804405e-01 5.53621888e-01 -3.34454060e-01 -1.23769522e-01 7.91001856e-01 -3.28061402e-01 -5.61463177e-01 1.88113257e-01 3.98414403e-01 6.06153309e-01 -4.16756958e-01 6.93365335e-01 -1.88182309e-01 5.70317686e-01 1.14302086e-02 2.34980017e-01 9.05865431e-01 -3.15791130e-01 7.79519260e-01 2.33986750e-01 -4.67481434e-01 -1.14025974e+00 -1.33425450e+00 -1.79064218e-02 1.38111579e+00 9.00445104e-01 -4.77563292e-01 -5.86642981e-01 -6.84678555e-01 2.68382460e-01 5.58832705e-01 -8.40224326e-01 -1.85886130e-01 -2.76292682e-01 -1.07904360e-01 -1.64642502e-02 5.74373662e-01 2.94229895e-01 -1.00894165e+00 -8.06437552e-01 -3.65798026e-01 -2.78187171e-02 -8.32608819e-01 -9.77977753e-01 2.95295902e-02 -5.32203913e-01 -1.22502232e+00 -6.01713300e-01 -1.10328543e+00 9.95268166e-01 8.13395143e-01 1.28606236e+00 4.64954793e-01 -4.96000648e-01 8.90967488e-01 -3.72133940e-01 -1.90669626e-01 -2.80663788e-01 -5.02981842e-01 1.07874544e-02 -1.93528663e-02 5.55632770e-01 -3.30188155e-01 -9.11311746e-01 5.64839900e-01 -1.10676682e+00 -1.00437395e-01 4.71949458e-01 7.01184273e-01 8.81648540e-01 -2.48494044e-01 2.42003798e-01 -7.58349538e-01 7.80655205e-01 -4.19236124e-01 -1.07429005e-01 7.13696778e-01 -6.39261544e-01 8.03155303e-02 3.29791129e-01 -7.16778636e-01 -7.42540121e-01 2.53089041e-01 4.33468401e-01 -3.50777149e-01 -3.20821464e-01 3.48615766e-01 -6.86088130e-02 -5.51368482e-02 8.46394777e-01 1.40601143e-01 -6.41238838e-02 -5.70477024e-02 7.37722158e-01 4.94519442e-01 5.18034816e-01 -4.17903781e-01 6.80195749e-01 5.42060316e-01 -8.41016546e-02 -8.13498080e-01 -5.92034340e-01 -7.28993833e-01 -5.76368809e-01 -1.29950881e-01 9.30752397e-01 -5.48778951e-01 -8.20681155e-01 -1.87375739e-01 -9.53639984e-01 7.79365227e-02 -5.40383220e-01 3.08157921e-01 -7.41189718e-01 5.94564319e-01 1.09110042e-01 -5.22316754e-01 -3.40486392e-02 -1.05132604e+00 9.18694377e-01 2.39707544e-01 -5.90741038e-01 -9.07579005e-01 -1.65645331e-01 4.85664532e-02 2.44086489e-01 -6.72919303e-02 1.35642445e+00 -8.08684886e-01 -4.07018781e-01 -2.33888239e-01 -4.95733529e-01 1.66271687e-01 4.01876301e-01 -8.16583037e-02 -8.50969672e-01 -1.20973453e-01 -2.92621534e-02 -1.81705162e-01 4.39716518e-01 2.25220621e-01 1.15594816e+00 -1.91807628e-01 -4.15250778e-01 3.53710800e-01 1.58102667e+00 4.44218099e-01 3.78407955e-01 4.35833007e-01 5.70562243e-01 7.98437357e-01 7.62962699e-01 2.01545879e-01 2.38674283e-01 8.29106569e-01 2.46444508e-01 -2.91776061e-01 5.47273969e-03 -2.17755139e-01 -1.85845986e-01 1.03362687e-01 1.61681324e-01 9.22137871e-02 -8.49769413e-01 6.31202042e-01 -1.75933015e+00 -1.14277065e+00 1.74884629e-02 2.51484060e+00 5.55907786e-01 -2.41960436e-01 1.62258744e-01 -1.39342308e-01 9.79945242e-01 1.28389597e-01 -6.02367520e-01 -6.11311972e-01 -1.68457523e-01 -1.42578661e-01 1.80155590e-01 3.53928000e-01 -9.95207548e-01 7.89889812e-01 6.81271362e+00 8.10073256e-01 -9.46044505e-01 -1.21144675e-01 6.07313037e-01 6.98611364e-02 -6.57807589e-01 3.05491805e-01 -2.54916877e-01 3.49313378e-01 1.89102128e-01 -5.66895902e-01 2.72354931e-01 1.08533597e+00 -7.78742805e-02 -4.34395194e-01 -1.58600211e+00 1.34939873e+00 2.06143424e-01 -1.06582367e+00 5.09802520e-01 -2.93013960e-01 5.02442002e-01 -5.22266388e-01 4.67208356e-01 -8.33188221e-02 1.85134903e-01 -1.15218353e+00 6.95452809e-01 4.71401334e-01 7.11319983e-01 -6.42246187e-01 3.01215112e-01 -5.24068587e-02 -1.46614885e+00 -3.93492579e-02 -4.38479096e-01 7.97943920e-02 -8.38011354e-02 5.28286323e-02 -5.19946456e-01 1.39600173e-01 1.01760542e+00 5.28615117e-01 -8.22627544e-01 9.95965183e-01 -9.48849991e-02 -2.24426225e-01 -8.60778987e-02 5.11225536e-02 2.80955344e-01 -1.98115900e-01 2.56963402e-01 1.01100707e+00 3.14497143e-01 1.38197288e-01 2.32244208e-01 9.51250672e-01 3.98727693e-02 2.64468759e-01 -9.83394325e-01 2.77392030e-01 6.49028897e-01 1.15029263e+00 -7.20458448e-01 -5.94650984e-01 -6.46711767e-01 1.23465848e+00 8.88660029e-02 5.31939864e-01 -5.26973307e-01 -5.67688167e-01 8.10486317e-01 6.38472661e-02 1.29111558e-01 2.54235685e-01 -2.66246825e-01 -7.30566382e-01 4.62393090e-02 -7.54817545e-01 6.25681937e-01 -1.19731772e+00 -1.53261805e+00 5.23301065e-01 1.19527653e-01 -1.48021305e+00 -2.57255048e-01 -3.57170761e-01 -7.82499671e-01 9.18386519e-01 -1.15019369e+00 -1.02969420e+00 -6.37572944e-01 8.36583674e-01 3.67611259e-01 7.45420381e-02 7.70746231e-01 -1.14346296e-01 1.70613676e-01 5.13126075e-01 -8.38935971e-02 -3.52864228e-02 9.20096874e-01 -1.27517998e+00 3.92547429e-01 5.63491225e-01 4.47655559e-01 9.65094566e-01 8.12971413e-01 -4.02409315e-01 -1.00098789e+00 -7.65492320e-01 9.81274784e-01 -4.65772629e-01 6.10191107e-01 -2.97249816e-02 -9.98517394e-01 4.32284683e-01 2.70447046e-01 1.94768324e-01 8.86299670e-01 -9.15673226e-02 -7.50046670e-01 -1.01394616e-01 -1.27284873e+00 8.83354664e-01 9.84854519e-01 -9.16363478e-01 -7.60568380e-01 2.12660149e-01 5.08360445e-01 1.50940061e-01 -8.15857053e-01 1.93189263e-01 7.45508611e-01 -1.01752663e+00 1.33474946e+00 -7.55558968e-01 4.38915528e-02 -5.62542140e-01 -4.45680916e-01 -1.25093472e+00 -4.57532495e-01 4.08183783e-02 6.54828727e-01 9.23771799e-01 3.34673554e-01 -3.28605145e-01 5.97662091e-01 9.12595570e-01 3.71468931e-01 -4.45844561e-01 -6.44906759e-01 -1.03761768e+00 -1.16616018e-01 -3.18645000e-01 7.00025439e-01 1.03802335e+00 2.41166264e-01 2.56919712e-01 -1.25040747e-02 1.09951556e-01 3.04686993e-01 5.94626665e-01 6.43505692e-01 -1.12551558e+00 -7.98794851e-02 -5.58725297e-01 -7.38630056e-01 -8.94140720e-01 8.16077963e-02 -8.50455999e-01 -2.64076479e-02 -1.38221610e+00 5.93304694e-01 -4.76883531e-01 -3.84672523e-01 3.20788175e-01 -3.70460004e-01 5.49647748e-01 6.34393215e-01 3.90965432e-01 -4.91941869e-01 4.19325717e-02 1.07941329e+00 -4.13524449e-01 1.27300564e-02 -1.72819257e-01 -8.99832547e-01 6.88648582e-01 6.60718203e-01 -3.33264321e-01 -5.71978211e-01 -2.26695210e-01 1.77410722e-01 -1.73864424e-01 5.65073669e-01 -6.75775588e-01 3.05353671e-01 -4.95310813e-01 4.60707307e-01 -1.51786000e-01 2.34336704e-01 -1.18855250e+00 1.84929535e-01 4.08815861e-01 -5.54972410e-01 4.20379639e-01 -4.15810235e-02 6.56489193e-01 -3.79853487e-01 -5.31456888e-01 7.95973957e-01 -3.59435350e-01 -1.32225251e+00 1.71798050e-01 -3.53885442e-01 8.36824551e-02 1.28385556e+00 -8.68873119e-01 -2.88496166e-01 -5.54614127e-01 -8.00098777e-01 3.13794911e-02 9.83728528e-01 5.21381676e-01 9.91089642e-01 -1.59800768e+00 -2.77324438e-01 1.20306075e-01 9.95110929e-01 -6.32097304e-01 2.71979243e-01 5.79660594e-01 -1.97409630e-01 1.39798537e-01 -2.74514228e-01 -6.70586646e-01 -1.63173318e+00 1.19868743e+00 1.59748420e-01 4.76625770e-01 -3.84391099e-01 5.57197094e-01 9.55401063e-01 -7.39460364e-02 2.16170460e-01 -2.39315450e-01 -2.51043856e-01 6.14992790e-02 6.49006307e-01 -3.71961109e-02 -2.39576548e-01 -7.74402678e-01 -4.91068482e-01 1.12317693e+00 -3.91571335e-02 -1.38185069e-01 7.40544438e-01 -4.58624452e-01 -3.74740735e-02 4.95610803e-01 1.38675809e+00 -1.18893765e-01 -7.70714343e-01 -4.68063563e-01 1.67025059e-01 -8.96022916e-01 -4.60118204e-01 -6.20805800e-01 -9.79344487e-01 8.33104849e-01 9.47100699e-01 4.11459774e-01 1.20763505e+00 4.58572298e-01 1.52835920e-01 4.68122303e-01 1.53097570e-01 -1.06571150e+00 3.87625515e-01 1.69768095e-01 8.94169629e-01 -1.57232845e+00 7.83685967e-02 -3.22080910e-01 -1.13634884e+00 9.23904896e-01 6.27988040e-01 1.17477402e-03 4.86350268e-01 -3.24047118e-01 2.33843341e-01 -3.39252234e-01 -3.72007370e-01 -5.23019552e-01 6.45509183e-01 8.11796367e-01 3.83390337e-01 1.58863913e-04 -1.81624755e-01 1.09050840e-01 6.85248971e-02 -6.01826608e-01 2.23144501e-01 9.14266050e-01 -5.70044398e-01 -9.35055494e-01 -4.29981470e-01 2.71400005e-01 -1.06654122e-01 -2.73024023e-01 -6.16949916e-01 8.43099833e-01 6.97054574e-03 8.88394833e-01 4.70161080e-01 -2.70863801e-01 3.19123596e-01 -1.03576563e-01 4.20696944e-01 -7.03898489e-01 -4.37889338e-01 -3.17902356e-01 -3.08235258e-01 -4.89765406e-01 -5.67096293e-01 -3.21226507e-01 -1.17861640e+00 -2.52478689e-01 -1.51164532e-01 2.74632871e-01 6.75174892e-01 7.38137662e-01 2.07911417e-01 -1.44746602e-01 6.48898542e-01 -6.18619561e-01 8.99161994e-02 -5.77408552e-01 -6.68544710e-01 1.22253609e+00 2.48012915e-01 -4.90087926e-01 -3.60960603e-01 2.42128253e-01]
[10.625020027160645, 1.3396114110946655]
2d703c1a-5e4e-4feb-a578-57ca3097a04e
grid-partitioned-attention-efficient
2107.03742
null
https://arxiv.org/abs/2107.03742v1
https://arxiv.org/pdf/2107.03742v1.pdf
Grid Partitioned Attention: Efficient TransformerApproximation with Inductive Bias for High Resolution Detail Generation
Attention is a general reasoning mechanism than can flexibly deal with image information, but its memory requirements had made it so far impractical for high resolution image generation. We present Grid Partitioned Attention (GPA), a new approximate attention algorithm that leverages a sparse inductive bias for higher computational and memory efficiency in image domains: queries attend only to few keys, spatially close queries attend to close keys due to correlations. Our paper introduces the new attention layer, analyzes its complexity and how the trade-off between memory usage and model power can be tuned by the hyper-parameters.We will show how such attention enables novel deep learning architectures with copying modules that are especially useful for conditional image generation tasks like pose morphing. Our contributions are (i) algorithm and code1of the novel GPA layer, (ii) a novel deep attention-copying architecture, and (iii) new state-of-the art experimental results in human pose morphing generation benchmarks.
['Roland Vollgraf', 'Christian Bracher', 'Gökhan Yildirim', 'Nikolay Jetchev']
2021-07-08
null
null
null
null
['deep-attention', 'conditional-image-generation', 'deep-attention']
['computer-vision', 'computer-vision', 'natural-language-processing']
[ 4.19675976e-01 5.82588017e-01 -1.12979330e-01 -1.89398915e-01 -1.13645601e+00 -2.57685125e-01 6.06525779e-01 -3.47914129e-01 -2.24438652e-01 7.73590028e-01 4.28644359e-01 -1.20224208e-01 -7.84067810e-02 -1.10572875e+00 -1.16716337e+00 -4.12059516e-01 -1.23323880e-01 9.26845491e-01 2.43718520e-01 -3.31304133e-01 1.80133194e-01 5.15444219e-01 -1.73409724e+00 5.91066122e-01 6.53749824e-01 1.08292055e+00 4.33890820e-01 8.31733525e-01 -1.42974049e-01 7.14400291e-01 -7.56353498e-01 -5.47046840e-01 2.44161725e-01 -4.82338548e-01 -1.20991015e+00 -2.96904474e-01 1.02846634e+00 -5.61877429e-01 -2.88966298e-01 7.15020895e-01 8.48664463e-01 3.22208852e-01 6.57166600e-01 -1.11648798e+00 -1.14884830e+00 7.77762234e-01 -5.52923977e-01 2.66854852e-01 3.68095517e-01 3.28719616e-01 1.15451193e+00 -9.45280194e-01 9.77722943e-01 1.39501023e+00 7.13553488e-01 8.60819578e-01 -1.35919273e+00 -4.68724310e-01 1.62237346e-01 2.28514373e-01 -1.27407122e+00 -1.73634693e-01 5.27274668e-01 -2.08616242e-01 1.45018947e+00 2.21965432e-01 7.69632697e-01 1.32062876e+00 2.63698429e-01 8.52852106e-01 8.77918422e-01 -2.07133517e-01 8.40204358e-02 -1.70136124e-01 -4.00085926e-01 7.66539156e-01 -1.64087191e-02 -8.95042904e-03 -7.53282547e-01 -1.02999061e-01 1.21876860e+00 -3.75878662e-01 -3.21930468e-01 -4.61193979e-01 -1.29242873e+00 9.12791669e-01 1.14042151e+00 5.91036342e-02 -3.13325584e-01 9.70922232e-01 1.62364900e-01 1.13047630e-01 2.47791931e-01 9.28190351e-01 -3.87161702e-01 -5.81359863e-02 -9.18484330e-01 5.90756536e-01 5.57243228e-01 1.22639847e+00 7.80006468e-01 8.79961327e-02 -5.33385038e-01 8.09564531e-01 -1.98813334e-01 3.75886172e-01 5.61801136e-01 -9.69626307e-01 4.23220694e-01 2.56846249e-01 -1.25315564e-03 -9.33383644e-01 -3.15390408e-01 -2.13278547e-01 -8.00512373e-01 2.49643341e-01 3.53864916e-02 -6.01040535e-02 -1.15417325e+00 2.02175331e+00 1.81733537e-02 -6.69061840e-02 -2.63945580e-01 9.08866346e-01 9.68497336e-01 6.88490033e-01 2.37632990e-01 2.85004377e-01 1.50034535e+00 -1.05885410e+00 -3.96267593e-01 -2.84276217e-01 1.09823324e-01 -4.58875597e-01 1.47606087e+00 6.37401864e-02 -1.69193769e+00 -7.92213023e-01 -1.00130332e+00 -7.25428462e-01 -5.60689569e-01 -1.35629550e-01 9.75299537e-01 2.25492284e-01 -1.49945843e+00 6.73444331e-01 -7.41905630e-01 -4.68343079e-01 6.99452341e-01 6.63027883e-01 -2.06189483e-01 2.89442003e-01 -1.16629243e+00 8.19636941e-01 3.50447088e-01 -1.64800540e-01 -8.16890597e-01 -9.36582327e-01 -9.43679392e-01 2.35595867e-01 1.54508546e-01 -1.40595555e+00 1.23910189e+00 -8.99242520e-01 -1.42808318e+00 9.72380817e-01 1.01716295e-01 -6.96954250e-01 4.91552860e-01 -4.56251532e-01 2.61020303e-01 1.99239537e-01 2.34565049e-01 1.78154099e+00 8.71800125e-01 -1.09077334e+00 -3.18313867e-01 -2.23693609e-01 2.77641594e-01 4.62468952e-01 -9.98759642e-02 -4.13814545e-01 -5.70267022e-01 -1.11663115e+00 -1.21100247e-01 -1.04657578e+00 -3.72148544e-01 1.45047918e-01 -4.43927735e-01 -1.51628479e-01 7.19054759e-01 -3.60503584e-01 9.09621716e-01 -1.78872395e+00 5.23137093e-01 -9.89949405e-02 2.11638268e-02 -3.59322913e-02 -3.15823525e-01 2.85332918e-01 -2.35372648e-01 2.69987434e-01 -2.50989884e-01 -2.65596986e-01 5.69954105e-02 1.77565321e-01 -7.22795367e-01 -2.20509619e-01 5.27075171e-01 1.49380767e+00 -8.13747942e-01 -4.96003777e-01 6.52959719e-02 4.96468633e-01 -1.06886888e+00 3.59573275e-01 -5.94616294e-01 1.60991743e-01 -2.19817430e-01 7.19426751e-01 4.37883705e-01 -3.96745145e-01 -2.88064927e-01 -5.51916897e-01 3.23744595e-01 1.62304804e-01 -8.07577729e-01 2.02382231e+00 -6.00078523e-01 5.42291522e-01 -1.75270557e-01 -3.56745064e-01 5.60441375e-01 8.42984244e-02 1.87760487e-01 -5.58969498e-01 -1.83738656e-02 -1.59693265e-03 -3.40146363e-01 -4.17265952e-01 8.98377776e-01 1.58545852e-01 -3.09924543e-01 4.29150313e-01 3.59316766e-01 -5.26888907e-01 1.25956029e-01 2.70676672e-01 9.18072999e-01 4.46977675e-01 1.57426998e-01 -4.08416092e-01 5.60033135e-02 -2.52174474e-02 7.51826838e-02 1.01506007e+00 1.34785295e-01 9.87323701e-01 5.54580688e-01 -8.78910422e-01 -1.04377246e+00 -1.13621652e+00 3.92150367e-03 1.51249039e+00 1.27892449e-01 -3.49232435e-01 -9.33852077e-01 -5.03691196e-01 5.00198230e-02 4.88568127e-01 -1.02699172e+00 -1.42300636e-01 -8.40907037e-01 -6.79357290e-01 5.89241326e-01 1.10057664e+00 6.03523195e-01 -1.55719829e+00 -9.26168442e-01 2.71975901e-02 -2.59534866e-01 -8.52031052e-01 -6.25643790e-01 2.18224093e-01 -7.77088583e-01 -8.25262308e-01 -1.03699970e+00 -7.94481456e-01 6.91504300e-01 -1.87375963e-01 1.75082779e+00 1.85675416e-02 -5.65849483e-01 3.86528820e-01 -8.27244818e-02 -4.49414700e-01 -1.33740559e-01 6.55604601e-01 -2.70418584e-01 -3.64625424e-01 2.32298505e-02 -6.61525905e-01 -7.15344071e-01 -4.19878475e-02 -8.94698024e-01 3.29094529e-01 8.82475197e-01 1.11159635e+00 8.61658692e-01 -6.34542167e-01 4.01599616e-01 -1.00117159e+00 6.75362408e-01 -2.28895724e-01 -5.58367848e-01 5.16550578e-02 -2.01248571e-01 4.86251503e-01 2.60062218e-01 -4.02068704e-01 -9.91237819e-01 1.65526137e-01 -2.72615850e-01 -4.82092917e-01 1.08913518e-01 1.48819283e-01 -6.77196383e-02 -8.32049102e-02 8.78061235e-01 4.43630517e-02 -1.45930022e-01 -3.50413740e-01 7.38768578e-01 4.31540422e-02 8.47386599e-01 -8.41304123e-01 5.89603603e-01 3.72784495e-01 7.38044828e-02 -4.52867180e-01 -6.37298167e-01 2.40501016e-01 -4.45332408e-01 2.60869890e-01 1.17313695e+00 -7.79043496e-01 -7.15502918e-01 2.60861576e-01 -1.19301915e+00 -6.56186044e-01 -7.01253951e-01 -2.36930642e-02 -1.11926627e+00 -2.39535481e-01 -9.03734386e-01 -3.47158939e-01 -6.18317246e-01 -1.24959743e+00 1.39283907e+00 1.67432398e-01 -6.20759130e-01 -7.11244941e-01 4.71035615e-02 1.13765061e-01 7.08058059e-01 4.29829478e-01 1.06134653e+00 -2.11509138e-01 -1.13421988e+00 3.25561404e-01 -3.62077177e-01 -3.06034982e-01 -3.23857427e-01 -2.13572606e-01 -1.02637684e+00 -2.15596870e-01 -7.14135587e-01 -8.41757536e-01 1.17619455e+00 4.93577898e-01 1.63077343e+00 -4.70820308e-01 -3.95784706e-01 1.14779949e+00 1.29361439e+00 -3.41001242e-01 9.01993930e-01 2.07601532e-01 8.56132269e-01 4.15869832e-01 3.08529228e-01 2.57375807e-01 3.54598045e-01 8.91733289e-01 5.65234125e-01 -3.09170574e-01 -5.07705986e-01 -5.48256814e-01 5.14380261e-02 2.21387208e-01 -3.37356806e-01 -1.20196953e-01 -6.75161839e-01 6.74094915e-01 -1.71708179e+00 -1.04949462e+00 5.10347545e-01 1.95193470e+00 1.05024922e+00 7.45830387e-02 3.41703407e-02 -3.39849085e-01 4.19146657e-01 3.31474632e-01 -5.84114015e-01 -7.12129772e-01 -5.83718345e-02 7.14608908e-01 5.81401527e-01 5.63122690e-01 -1.18727839e+00 1.17199993e+00 7.37310839e+00 8.54771495e-01 -8.36428046e-01 -9.72515643e-02 8.29225719e-01 -4.09665793e-01 -4.07477528e-01 -3.57986003e-01 -8.04458141e-01 1.25539571e-01 6.62522793e-01 -2.89272703e-02 5.34263551e-01 1.05247355e+00 -5.27169585e-01 1.37429178e-01 -1.36106634e+00 1.14020431e+00 2.52324104e-01 -1.62933576e+00 7.60587215e-01 4.94727073e-03 8.68854225e-01 1.12456545e-01 4.96649653e-01 3.86590242e-01 5.24218321e-01 -1.36533368e+00 7.21171975e-01 4.97487694e-01 1.25279343e+00 -8.24676454e-01 5.75709045e-01 -2.20929921e-01 -1.18031418e+00 -1.91603303e-01 -4.38862741e-01 2.51745731e-01 1.37718886e-01 -1.51112145e-02 -7.57363498e-01 1.47519782e-01 1.05091012e+00 4.35183644e-01 -7.17892349e-01 6.32958531e-01 -2.79081613e-01 -1.24801815e-01 -1.86531588e-01 1.10692658e-01 2.55364537e-01 2.91027486e-01 4.42481339e-01 1.26692510e+00 4.65697199e-01 -6.56437054e-02 -1.59029886e-01 1.27706671e+00 -3.98541987e-01 -1.70036718e-01 -7.45388925e-01 1.65706977e-01 5.05947948e-01 1.12572241e+00 -6.56980395e-01 -4.27404135e-01 -6.88801706e-03 1.25696826e+00 6.98175550e-01 3.39344174e-01 -9.65421498e-01 -4.68375772e-01 9.08457637e-01 4.28964645e-01 6.83876574e-01 4.30445001e-02 -3.32532734e-01 -7.89998770e-01 -2.20174983e-01 -7.50723898e-01 5.31704545e-01 -1.20787239e+00 -1.10366189e+00 9.66129005e-01 2.30114803e-01 -8.88829052e-01 -6.12322211e-01 -5.26979387e-01 -4.20355707e-01 8.17784548e-01 -1.25979209e+00 -1.44931221e+00 -5.37003577e-01 6.18008077e-01 6.41288877e-01 7.68320113e-02 1.07250428e+00 -8.57382044e-02 -1.64083615e-01 9.05781090e-01 -5.57912409e-01 4.03108560e-02 6.59453690e-01 -1.57994962e+00 9.63914335e-01 3.53202939e-01 3.18203747e-01 6.11863732e-01 5.12214184e-01 -3.83773953e-01 -1.18821239e+00 -1.09746337e+00 6.84483409e-01 -6.49545789e-01 1.69330597e-01 -4.50864524e-01 -6.81068480e-01 9.59340215e-01 4.51015502e-01 2.19495088e-01 4.22068864e-01 1.03842437e-01 -4.61425096e-01 -5.92549890e-02 -1.10928071e+00 8.92119944e-01 1.38862157e+00 -4.73573714e-01 -4.48564321e-01 3.83957475e-01 9.92607415e-01 -9.14699137e-01 -8.42297494e-01 4.78019327e-01 5.32470286e-01 -1.00960100e+00 1.24318361e+00 -6.60309911e-01 6.87638938e-01 -1.29325554e-01 6.95955232e-02 -1.22376513e+00 -6.63667083e-01 -7.72520363e-01 -3.43268394e-01 7.96964169e-01 5.28980374e-01 -3.05337816e-01 9.48451579e-01 7.34701872e-01 -1.70736268e-01 -1.15093529e+00 -7.78937578e-01 -3.90265286e-01 1.41720623e-01 -1.46660730e-01 8.15954089e-01 5.53781748e-01 -2.36311823e-01 3.88141513e-01 -3.03177178e-01 -1.04383208e-01 3.19137245e-01 2.47210056e-01 9.08514380e-01 -7.81392932e-01 -6.04717314e-01 -4.95264500e-01 -5.35822093e-01 -1.33493209e+00 -1.02061801e-01 -6.61409497e-01 2.24690195e-02 -1.48142457e+00 2.56690443e-01 -2.96941906e-01 -1.19280638e-02 6.25863314e-01 -3.17381322e-01 7.80588627e-01 3.02462935e-01 1.39231578e-01 -5.60098350e-01 5.51883578e-01 1.29488194e+00 -2.44932100e-01 -2.27118120e-01 -4.95496094e-01 -8.90618086e-01 5.93586445e-01 5.03961265e-01 -2.09496826e-01 -7.25462258e-01 -9.00811791e-01 3.67581099e-01 -1.97218046e-01 5.39071679e-01 -1.26967967e+00 2.22200468e-01 3.16097997e-02 8.04424644e-01 -7.11259723e-01 6.44669294e-01 -5.18521547e-01 2.13807538e-01 2.94732153e-01 -6.13952875e-01 4.32009965e-01 3.78440976e-01 4.97738332e-01 -1.06629401e-01 2.05061719e-01 7.80090630e-01 -5.72867811e-01 -7.61128128e-01 5.31127751e-01 -9.81396288e-02 3.90407562e-01 1.04795420e+00 -2.11893737e-01 -1.97372109e-01 -5.07461846e-01 -8.73700678e-01 -3.91553454e-02 5.55465162e-01 5.22248507e-01 6.66996896e-01 -1.66095448e+00 -7.44009256e-01 2.72629857e-01 1.37178764e-01 2.98721403e-01 2.19347522e-01 3.33517492e-01 -8.58690262e-01 4.11329597e-01 -4.43483770e-01 -5.14685810e-01 -1.00190973e+00 7.18220949e-01 3.85585666e-01 -2.78167635e-01 -5.39574862e-01 1.41553891e+00 4.90198284e-01 -1.71613097e-01 2.19011575e-01 -4.79774147e-01 4.24170643e-02 -1.11502595e-01 6.41388714e-01 6.96073249e-02 -5.48772141e-02 -5.71830571e-01 -2.12255061e-01 7.25078225e-01 -2.45765805e-01 -1.82596475e-01 1.11141264e+00 7.88901597e-02 3.12077664e-02 -5.14454069e-03 9.23931241e-01 -2.90415883e-01 -1.39347947e+00 8.59021489e-03 -4.43250746e-01 -4.93450612e-01 -2.30680168e-01 -7.43892610e-01 -1.13867974e+00 8.76708090e-01 3.72808367e-01 -1.90496296e-02 1.15885913e+00 2.27753207e-01 1.06382000e+00 3.69786590e-01 4.20679480e-01 -1.02480519e+00 5.20852923e-01 3.92717987e-01 1.49962544e+00 -9.23454523e-01 -8.05028826e-02 -1.94515467e-01 -6.82358563e-01 8.94570708e-01 1.07485425e+00 -2.55772501e-01 3.97272855e-01 5.34009695e-01 -2.69640952e-01 -2.38119021e-01 -9.00981843e-01 -2.87000567e-01 3.09406847e-01 9.28472102e-01 4.07863259e-01 -1.60174698e-01 1.68662608e-01 5.17124116e-01 -6.79676414e-01 -9.07962024e-02 6.36665225e-02 7.76890397e-01 -1.51699826e-01 -8.32617640e-01 -1.89623058e-01 3.75126570e-01 -4.32403535e-01 -4.08012241e-01 -5.79552650e-01 9.15451348e-01 3.53768975e-01 2.29826391e-01 4.64738309e-01 -2.91756481e-01 3.71273309e-01 -6.48014843e-02 7.94765532e-01 -6.32503033e-01 -7.16986418e-01 -3.82860482e-01 -5.84196895e-02 -1.09040749e+00 -1.79053098e-01 -1.06171228e-01 -1.09072471e+00 -3.13816965e-01 -9.64239910e-02 -2.12426946e-01 1.56045005e-01 2.60847956e-01 9.67189729e-01 7.37856388e-01 -3.50970253e-02 -1.37405610e+00 -4.27115202e-01 -9.01852667e-01 -1.09034866e-01 5.88919461e-01 1.89082518e-01 -5.74627280e-01 6.02446720e-02 2.80208588e-01]
[11.330978393554688, -0.33600690960884094]
2ab03902-bf2d-49b9-874b-99130bbafc0d
multi-channel-target-speaker-extraction-with
2302.07928
null
https://arxiv.org/abs/2302.07928v1
https://arxiv.org/pdf/2302.07928v1.pdf
Multi-Channel Target Speaker Extraction with Refinement: The WavLab Submission to the Second Clarity Enhancement Challenge
This paper describes our submission to the Second Clarity Enhancement Challenge (CEC2), which consists of target speech enhancement for hearing-aid (HA) devices in noisy-reverberant environments with multiple interferers such as music and competing speakers. Our approach builds upon the powerful iterative neural/beamforming enhancement (iNeuBe) framework introduced in our recent work, and this paper extends it for target speaker extraction. We therefore name the proposed approach as iNeuBe-X, where the X stands for extraction. To address the challenges encountered in the CEC2 setting, we introduce four major novelties: (1) we extend the state-of-the-art TF-GridNet model, originally designed for monaural speaker separation, for multi-channel, causal speech enhancement, and large improvements are observed by replacing the TCNDenseNet used in iNeuBe with this new architecture; (2) we leverage a recent dual window size approach with future-frame prediction to ensure that iNueBe-X satisfies the 5 ms constraint on algorithmic latency required by CEC2; (3) we introduce a novel speaker-conditioning branch for TF-GridNet to achieve target speaker extraction; (4) we propose a fine-tuning step, where we compute an additional loss with respect to the target speaker signal compensated with the listener audiogram. Without using external data, on the official development set our best model reaches a hearing-aid speech perception index (HASPI) score of 0.942 and a scale-invariant signal-to-distortion ratio improvement (SI-SDRi) of 18.8 dB. These results are promising given the fact that the CEC2 data is extremely challenging (e.g., on the development set the mixture SI-SDR is -12.3 dB). A demo of our submitted system is available at WAVLab CEC2 demo.
['Nobutaka Ono', 'Manuel Pariente', 'Shinji Watanabe', 'Yoshiki Masuyama', 'Zhong-Qiu Wang', 'Samuele Cornell']
2023-02-15
null
null
null
null
['target-speaker-extraction', 'speech-enhancement', 'speaker-separation']
['audio', 'speech', 'speech']
[ 3.50481242e-01 -7.61183947e-02 5.07739127e-01 4.81280386e-02 -1.62442434e+00 -3.05399328e-01 1.60038143e-01 -2.17389241e-01 -4.74276811e-01 5.88704050e-01 7.11138666e-01 -4.48748767e-01 -2.75310189e-01 -7.37973303e-02 -3.64782721e-01 -7.85992026e-01 -3.27258527e-01 -1.49852738e-01 2.57099956e-01 -4.24059093e-01 -3.02960098e-01 4.35766369e-01 -1.72193766e+00 3.69593441e-01 8.30583572e-01 1.26265228e+00 2.87483901e-01 1.25047910e+00 5.10948539e-01 4.37615246e-01 -8.40971112e-01 -1.44173801e-01 2.66031593e-01 -6.18600011e-01 -4.21341896e-01 -1.42871246e-01 4.75269556e-01 -4.37572688e-01 -1.84379742e-01 7.69739866e-01 1.52891624e+00 1.19470336e-01 4.41360146e-01 -9.71909344e-01 2.58145481e-01 4.78820890e-01 -3.12232852e-01 2.16403559e-01 1.93443879e-01 1.11658216e-01 8.72520387e-01 -1.04665112e+00 1.94353327e-01 1.04579222e+00 9.51474786e-01 4.65231448e-01 -1.19227815e+00 -7.92672575e-01 -1.55989915e-01 2.76008755e-01 -1.31388807e+00 -1.20225859e+00 4.85897750e-01 -1.38421655e-01 1.23854244e+00 5.25171697e-01 5.18013954e-01 9.35042381e-01 -2.81974107e-01 7.15044737e-01 1.20256305e+00 -7.67795265e-01 3.03936332e-01 -4.79601063e-02 1.88913792e-02 9.37201083e-02 -5.15740514e-01 7.33099699e-01 -7.17041016e-01 -1.73238516e-01 5.09889305e-01 -8.43530834e-01 -7.65691638e-01 3.28985184e-01 -8.73383999e-01 4.20193315e-01 1.08851500e-01 4.01454955e-01 -3.79947335e-01 1.39540397e-02 3.56277853e-01 5.71873486e-01 6.48668766e-01 2.83647239e-01 -6.25605762e-01 -4.06932443e-01 -1.29733992e+00 2.45997906e-01 7.75453925e-01 6.76293969e-01 1.01799510e-01 4.49269503e-01 -3.45632821e-01 1.36836946e+00 1.59665957e-01 6.57567739e-01 3.16393673e-01 -9.17577982e-01 3.02909851e-01 -7.04396605e-01 8.99214298e-02 -2.26926476e-01 -4.88913029e-01 -1.06003690e+00 -6.70494616e-01 5.05807042e-01 2.72697061e-01 -4.67955172e-01 -1.03084433e+00 1.89604545e+00 2.64306188e-01 3.39742720e-01 7.84053281e-02 8.77734125e-01 7.12235034e-01 5.51183939e-01 -1.22563064e-01 -5.51171780e-01 1.42132723e+00 -9.33029830e-01 -8.89084578e-01 -1.38622858e-02 1.80851981e-01 -1.20446908e+00 8.06278288e-01 9.68731582e-01 -1.55276763e+00 -6.09518826e-01 -1.03162909e+00 2.30560198e-01 1.59298211e-01 7.06475377e-02 1.19190820e-01 1.19871175e+00 -1.64609468e+00 3.51530045e-01 -7.38418102e-01 -8.93030241e-02 4.12819274e-02 4.79725242e-01 -2.60541774e-02 1.50846034e-01 -1.14459133e+00 7.26305187e-01 -8.12579021e-02 1.95673872e-02 -8.07028830e-01 -9.55057740e-01 -6.19779587e-01 1.87479138e-01 2.06532523e-01 -7.19826639e-01 1.70187199e+00 -8.50469649e-01 -1.84322095e+00 5.23356795e-01 -4.23056304e-01 -6.09958529e-01 5.92481434e-01 -5.35968363e-01 -1.00831103e+00 1.51199818e-01 -3.43088001e-01 4.10326749e-01 1.05149174e+00 -1.12274873e+00 -8.60196948e-01 -1.42336026e-01 -3.24747682e-01 1.90648288e-01 -2.67126679e-01 4.67782348e-01 -4.93100375e-01 -1.06007528e+00 3.40904482e-02 -5.91669202e-01 -1.51171818e-01 -3.56130570e-01 -3.75872046e-01 3.60989422e-01 4.24778759e-01 -1.20772576e+00 1.49393070e+00 -2.43610215e+00 -1.31803766e-01 3.14313322e-01 2.13617027e-01 6.05428457e-01 -2.41848812e-01 2.75261611e-01 -4.11872864e-01 -3.38019550e-01 -3.11805129e-01 -7.03390241e-01 2.87486408e-02 -3.58513355e-01 -3.66984844e-01 2.74414301e-01 -1.03649333e-01 3.30150843e-01 -4.97496039e-01 -1.25503957e-01 2.59472370e-01 8.45671177e-01 -7.92774022e-01 3.44975919e-01 3.11970204e-01 2.70365536e-01 2.78005868e-01 4.16778088e-01 9.64240491e-01 4.30534720e-01 -1.25475094e-01 -4.07103688e-01 -3.30442846e-01 5.47235727e-01 -1.49624217e+00 1.50460660e+00 -8.13399494e-01 6.20457232e-01 8.71681631e-01 -4.88036543e-01 5.65755486e-01 8.12511981e-01 3.89870644e-01 -5.99168718e-01 -4.66066860e-02 6.09991193e-01 2.58442253e-01 -3.20086747e-01 2.53465503e-01 -4.52769160e-01 4.39660698e-01 1.81999773e-01 3.15977037e-01 -2.49203935e-01 -1.68950900e-01 -7.55301258e-03 1.21383405e+00 -3.09743434e-01 1.93879396e-01 -3.21382105e-01 4.67420638e-01 -7.02729642e-01 3.50994259e-01 7.47835636e-01 -3.90199035e-01 9.41983581e-01 1.15185872e-01 4.30291593e-01 -8.66646290e-01 -1.42310274e+00 -3.60143155e-01 1.17327821e+00 -5.44073582e-01 -3.38574469e-01 -9.62120950e-01 -3.53766084e-02 -4.10732538e-01 6.91284001e-01 1.20045140e-03 2.21464813e-01 -6.07850969e-01 -5.46631396e-01 7.85795510e-01 4.76266772e-01 3.00495863e-01 -8.59698951e-01 -2.68878371e-01 5.67054510e-01 -3.75457436e-01 -1.08282065e+00 -7.23497272e-01 3.96416605e-01 -4.54596758e-01 -5.15473187e-01 -9.86480653e-01 -7.67686248e-01 3.16834338e-02 9.86440778e-02 9.49717462e-01 -2.99233288e-01 -6.13864101e-02 5.53623915e-01 -2.28452146e-01 -5.09829044e-01 -4.50185686e-01 -2.20690757e-01 3.46770853e-01 -3.31773795e-02 -5.68253882e-02 -9.03893471e-01 -8.95991027e-01 3.27510387e-01 -7.02130795e-01 -9.24134478e-02 4.32301819e-01 1.01250899e+00 1.79966390e-01 2.24400491e-01 9.90016758e-01 -1.66876867e-01 7.42260516e-01 1.64607354e-02 -5.80240667e-01 -1.20084845e-01 -5.40731668e-01 -2.94579595e-01 4.15090859e-01 -4.43885535e-01 -1.28289390e+00 -2.19829947e-01 -1.03704906e+00 -1.12132855e-01 -2.19809279e-01 2.83878505e-01 -3.72904539e-01 -2.88531687e-02 6.74653232e-01 6.73113540e-02 8.87219235e-03 -6.62404358e-01 1.56705052e-01 1.18149877e+00 9.03364182e-01 -3.17459345e-01 5.78998566e-01 1.26319557e-01 -3.08461279e-01 -9.27446723e-01 -3.64498168e-01 -7.40073085e-01 -3.65582258e-02 -8.31069723e-02 4.83302534e-01 -1.16156948e+00 -7.25732088e-01 8.21480215e-01 -1.09375727e+00 -5.23983300e-01 -2.45481148e-01 8.30829978e-01 -6.55285776e-01 2.77839452e-01 -7.36526966e-01 -1.33255112e+00 -6.66616201e-01 -1.12800550e+00 9.65143800e-01 -2.34012201e-01 -2.13344932e-01 -6.48840249e-01 8.94966125e-02 3.14027309e-01 9.96301889e-01 -1.28581867e-01 6.04601681e-01 -4.14559871e-01 3.92144471e-02 2.20745087e-01 7.64525980e-02 8.20567906e-01 6.12457991e-02 -4.66151744e-01 -1.70753908e+00 -5.70973635e-01 4.31234032e-01 5.52539807e-03 7.45544255e-01 9.59024072e-01 6.86603963e-01 -1.93208113e-01 -2.64719613e-02 5.86444199e-01 1.04873550e+00 2.70509511e-01 7.37358987e-01 -4.12199227e-03 2.47531589e-02 5.09545028e-01 4.17711705e-01 4.41357821e-01 -1.73245389e-02 1.04246044e+00 2.31649596e-02 -6.42239690e-01 -8.48756671e-01 5.31770624e-02 6.30452454e-01 1.07535708e+00 1.70025453e-02 -2.41739333e-01 -6.76276922e-01 7.00092673e-01 -1.35018814e+00 -8.36542726e-01 -4.25700285e-03 2.49970841e+00 9.71020997e-01 2.00013548e-01 3.89993280e-01 6.20163381e-01 6.47809327e-01 -8.69511440e-02 -2.12871835e-01 -3.08793455e-01 -2.85833955e-01 1.03832543e+00 3.32097232e-01 1.12702465e+00 -8.38201165e-01 4.79205847e-01 6.17769861e+00 1.14707363e+00 -1.27887392e+00 4.88653511e-01 5.04473209e-01 -4.78546292e-01 3.11537944e-02 -5.57780921e-01 -5.79932690e-01 1.23469286e-01 1.42924690e+00 -6.83243349e-02 5.48438311e-01 3.57859850e-01 5.33205926e-01 5.82227595e-02 -8.85661423e-01 1.10901701e+00 -2.43337806e-02 -7.12641835e-01 -6.42284930e-01 8.74620751e-02 3.06505114e-01 2.08885893e-01 3.73368919e-01 3.89440507e-01 2.22274475e-02 -8.67796540e-01 8.34516823e-01 1.14130646e-01 1.20244670e+00 -7.79910922e-01 5.85781038e-01 6.07364848e-02 -1.19537497e+00 -1.44324049e-01 8.88915211e-02 2.66750515e-01 4.76301640e-01 7.81136692e-01 -1.03894913e+00 6.44314229e-01 7.53875554e-01 -9.44954082e-02 8.03927146e-03 1.46781969e+00 -2.12359503e-01 1.12389803e+00 -5.58021605e-01 4.96336758e-01 -9.72104147e-02 3.40484142e-01 9.80009854e-01 1.46287072e+00 5.82639933e-01 9.05557051e-02 -3.45832288e-01 3.63986194e-01 1.97033003e-01 6.21558819e-03 3.31404731e-02 7.39698470e-01 4.80192631e-01 1.05679560e+00 -6.68962002e-02 -6.66757151e-02 -1.54468179e-01 8.22069883e-01 -3.63952845e-01 6.97252512e-01 -7.77167320e-01 -6.38510466e-01 8.00263345e-01 1.04300305e-01 6.15933657e-01 -2.16045622e-02 -1.72544137e-01 -6.33100569e-01 6.98097199e-02 -1.24362338e+00 6.45719022e-02 -8.48367631e-01 -8.76992941e-01 1.05201268e+00 -1.76137283e-01 -1.31084192e+00 -4.34378445e-01 -6.07328296e-01 -5.11149704e-01 1.31667638e+00 -1.50153041e+00 -9.11384821e-01 1.68496430e-01 8.11381698e-01 4.71423566e-01 -4.14276347e-02 9.56093729e-01 8.56521666e-01 -1.13681361e-01 1.11717737e+00 1.58629477e-01 -3.46896440e-01 8.39321315e-01 -1.24895978e+00 3.33778262e-01 9.72595870e-01 -2.32698712e-02 5.09835482e-01 1.07652223e+00 -2.03603387e-01 -9.11396682e-01 -9.20797408e-01 9.81225252e-01 -5.21814898e-02 4.80287343e-01 -6.31806850e-01 -6.45442128e-01 2.53225476e-01 3.82111609e-01 -2.78464943e-01 8.23824644e-01 3.51545960e-01 -2.23207653e-01 -3.54814559e-01 -1.11911833e+00 5.55849850e-01 9.57197666e-01 -6.15683436e-01 -3.94591719e-01 1.15011625e-01 8.66416216e-01 -5.48611701e-01 -6.83615088e-01 4.74585444e-01 6.27745211e-01 -1.08845747e+00 9.87435937e-01 -1.46819979e-01 -1.31795183e-01 -2.98279136e-01 -3.79231274e-01 -1.60146046e+00 -1.17989510e-01 -1.28400886e+00 -9.79467109e-02 1.37194502e+00 5.71784556e-01 -7.44439721e-01 3.98580700e-01 2.34606266e-01 -6.58970714e-01 -4.97968167e-01 -1.33735776e+00 -9.24584985e-01 -1.20976828e-01 -9.79644299e-01 3.37947816e-01 3.38104099e-01 -8.20187330e-02 2.26082638e-01 -5.06810904e-01 3.51658791e-01 5.88438094e-01 -5.63445210e-01 5.01189351e-01 -8.26936603e-01 -9.19942141e-01 -4.79375660e-01 -1.24430194e-01 -1.26025403e+00 -3.44366223e-01 -4.23534274e-01 2.86944777e-01 -1.21603084e+00 -3.49954784e-01 -2.41454601e-01 -6.83144331e-01 2.28978544e-01 -1.76935211e-01 1.95342273e-01 2.43491337e-01 -2.48089463e-01 -4.37735692e-02 5.22188962e-01 7.66286433e-01 -3.51355150e-02 -5.42183816e-01 5.39856732e-01 -7.74658978e-01 4.78705674e-01 3.89799893e-01 -2.44323134e-01 -2.25047484e-01 -2.07659751e-01 -3.36297244e-01 5.15348673e-01 3.67074996e-01 -1.31231463e+00 2.57601619e-01 5.67947090e-01 -6.44731447e-02 -5.50519347e-01 7.44567394e-01 -8.75469625e-01 2.48654112e-01 3.64956260e-01 -1.43041506e-01 -3.60940963e-01 6.50326014e-01 1.70593485e-01 -4.43765312e-01 1.90350622e-01 9.56449091e-01 4.78690952e-01 -1.46261379e-01 -6.71515092e-02 -5.57035625e-01 -1.07032634e-01 2.87988454e-01 -1.76567305e-02 -1.63612530e-01 -8.12557638e-01 -1.01022065e+00 -2.37154305e-01 -2.00962916e-01 1.45254523e-01 5.21392047e-01 -1.10108149e+00 -1.14987862e+00 1.39179006e-01 -2.06025422e-01 -4.25611228e-01 5.34500599e-01 1.07465172e+00 3.16321887e-02 3.28713566e-01 6.75580949e-02 -5.23213029e-01 -1.59353721e+00 7.19232634e-02 6.48812771e-01 -2.65794158e-01 -5.19466221e-01 1.16841543e+00 3.61965448e-01 -5.22460379e-02 4.60216165e-01 -2.69735813e-01 1.30351365e-01 -1.71703726e-01 8.08593929e-01 5.46238005e-01 7.39266157e-01 -4.89506453e-01 -2.77124822e-01 2.31898785e-01 -1.59142427e-02 -8.66325319e-01 1.32087100e+00 -1.30145699e-01 1.65062308e-01 7.38069043e-02 1.24680018e+00 6.09969020e-01 -1.03745222e+00 -2.52994269e-01 -3.28485698e-01 -2.88555920e-01 5.25634944e-01 -1.30553305e+00 -9.13713813e-01 9.22080934e-01 1.35677922e+00 2.72999816e-02 1.90270936e+00 -2.34588116e-01 8.82966042e-01 -2.17686027e-01 7.95803368e-02 -9.23871338e-01 -1.23358957e-01 5.37062407e-01 1.18557572e+00 -6.94998384e-01 -3.87686729e-01 -2.86859781e-01 -2.82476872e-01 9.25073266e-01 2.01997057e-01 5.49860001e-01 7.45795488e-01 7.11187601e-01 4.33072656e-01 1.37469664e-01 -5.26907563e-01 -2.88667947e-01 3.29917043e-01 7.85719514e-01 5.82505345e-01 9.73754451e-02 -9.07565281e-02 7.93914616e-01 -6.74285233e-01 -1.13804191e-01 1.34231418e-01 5.59778690e-01 -4.76670384e-01 -1.12711298e+00 -7.22164929e-01 1.97875902e-01 -6.72712982e-01 -7.27557719e-01 1.43266231e-01 3.66844982e-01 -1.95018928e-02 1.57211030e+00 -2.25094900e-01 -3.21075410e-01 7.28691816e-01 -1.31423194e-02 3.32137287e-01 -2.16263384e-01 -1.01716876e+00 9.96435523e-01 4.17578161e-01 -4.15066957e-01 -1.72735959e-01 -6.50892198e-01 -9.04233277e-01 -1.31051451e-01 -5.33759654e-01 1.78797871e-01 6.95423782e-01 6.50324583e-01 3.72176647e-01 9.65499222e-01 6.74869478e-01 -8.39704096e-01 -5.25328338e-01 -1.29154801e+00 -8.26278150e-01 -5.47273047e-02 9.20025945e-01 -3.97783428e-01 -6.52756929e-01 -1.40538244e-02]
[15.089743614196777, 5.836637496948242]
0e8cce1f-3bb6-4454-88e9-20f52a721c56
multiple-angles-of-arrival-estimation-using
2002.00541
null
https://arxiv.org/abs/2002.00541v1
https://arxiv.org/pdf/2002.00541v1.pdf
Multiple Angles of Arrival Estimation using Neural Networks
MUltiple SIgnal Classification (MUSIC) and Estimation of signal parameters via rotational via rotational invariance (ESPRIT) has been widely used in super resolution direction of arrival estimation (DoA) in both Uniform Linear Arrays (ULA) or Uniform Circular Arrays (UCA). However, problems become challenging when the number of source signal increase, MUSIC suffer from computation complexity when finding the peaks, while ESPRIT may not robust to array geometry offset. Therefore, Neural Network become a potential solution. In this paper, we propose a neural network to estimate the azimuth and elevation angles, based on the correlated matrix extracted from received data. Also, a serial scheme is listed to estimate multiple signals cases. The result shows the neural network can achieve an accurate estimation under low SNR and deal with multiple signals.
['Jianyuan Yu']
2020-02-03
null
null
null
null
['direction-of-arrival-estimation']
['audio']
[ 2.49197334e-01 -9.34637725e-01 1.85576126e-01 -6.52516335e-02 -7.21874714e-01 -6.46338403e-01 -1.87926670e-03 -4.42052275e-01 -1.38080195e-01 6.75965369e-01 3.60407591e-01 -6.55474439e-02 -8.05328488e-01 -5.40464759e-01 -3.42198789e-01 -1.00571263e+00 -4.82231438e-01 -2.53283679e-01 -4.23793703e-01 6.09305501e-02 1.19158868e-02 8.06817830e-01 -1.28972566e+00 -6.85316771e-02 6.65200710e-01 1.24049342e+00 1.70043577e-02 7.71722138e-01 2.30033875e-01 4.72618103e-01 -9.60000813e-01 1.73761159e-01 3.55162024e-01 -3.67545187e-01 3.81411552e-01 -6.29133165e-01 2.49118239e-01 -2.95873672e-01 -3.50493968e-01 1.06336331e+00 1.22742271e+00 8.49838108e-02 7.58674800e-01 -1.08423877e+00 -3.14103186e-01 8.01595330e-01 -9.74459827e-01 4.37895298e-01 6.91715702e-02 -7.03982830e-01 2.52021044e-01 -1.24393916e+00 -1.07999660e-01 1.01550102e+00 1.24387324e+00 -2.15480059e-01 -6.11334324e-01 -8.98200035e-01 -5.53133368e-01 3.33074629e-01 -1.88897705e+00 -7.25988746e-01 9.13107514e-01 -3.85994799e-02 4.21568006e-01 4.75253344e-01 3.96847337e-01 8.35501850e-01 -1.92881469e-02 5.24035275e-01 9.45682883e-01 -4.95707661e-01 -1.93048734e-03 -5.05484462e-01 9.03707594e-02 1.03193924e-01 7.51494348e-01 1.34058282e-01 -3.74244452e-01 -2.93892562e-01 1.05779004e+00 -1.69485956e-01 -6.15003645e-01 -6.52729115e-03 -1.44051790e+00 4.71214682e-01 1.40960097e-01 6.37028813e-01 -5.98420143e-01 1.36659727e-01 9.00844336e-02 3.34785700e-01 -4.87109944e-02 6.26672864e-01 -3.60155284e-01 -3.82403806e-02 -9.52900529e-01 9.26511809e-02 5.78514695e-01 7.08819866e-01 2.36082047e-01 9.88265812e-01 4.52493131e-03 1.13469887e+00 3.82386923e-01 1.18764997e+00 5.92785895e-01 -7.26877570e-01 4.68686283e-01 -2.47269154e-01 2.88001299e-01 -1.81921935e+00 -9.20500934e-01 -1.47150564e+00 -1.52045727e+00 -8.74373168e-02 4.35016096e-01 -8.25372875e-01 -3.88124883e-01 1.49863935e+00 1.69313610e-01 7.32552946e-01 5.09351671e-01 1.16813505e+00 8.32929611e-01 9.20137584e-01 -6.66972935e-01 -8.21988106e-01 1.16933203e+00 -5.89304745e-01 -1.20068896e+00 -3.90371740e-01 6.51535988e-02 -1.19801211e+00 -1.23764113e-01 8.80213141e-01 -7.88398266e-01 -8.13592553e-01 -1.29530311e+00 6.35665238e-01 5.39910495e-02 5.31726658e-01 5.53883851e-01 9.46012855e-01 -6.26979828e-01 1.38971403e-01 -3.70118946e-01 2.62835056e-01 -1.04396366e-01 6.30841851e-02 -2.28342369e-01 -1.45252734e-01 -1.24300981e+00 4.81114089e-01 9.00015011e-02 7.57309556e-01 -2.64483035e-01 -3.47476840e-01 -6.38863921e-01 8.43950510e-02 5.22655025e-02 -4.39945608e-01 8.01790357e-01 -9.87316132e-01 -1.31968617e+00 -1.53196037e-01 -3.81431520e-01 -3.81735981e-01 -1.03172828e-02 -3.05177242e-01 -1.27810323e+00 4.81693260e-02 -2.33036913e-02 -2.03407690e-01 1.04661238e+00 -1.08711779e+00 -4.70933139e-01 -2.33859971e-01 -7.85009444e-01 3.61560583e-01 -3.00434083e-02 1.68719843e-01 -3.60934287e-02 -1.06099248e+00 8.35015893e-01 -4.42761600e-01 -2.64436394e-01 -5.38773596e-01 -1.55969948e-01 3.88881058e-01 4.93693888e-01 -8.68235588e-01 1.57057989e+00 -2.36568141e+00 -4.44336459e-02 5.95476329e-01 -2.62501299e-01 2.18574435e-01 -2.45572418e-01 3.60837221e-01 -3.77538979e-01 -3.67867291e-01 1.34773374e-01 4.38560098e-01 -3.86476457e-01 -2.05628112e-01 -3.15111548e-01 5.85032344e-01 -3.66893500e-01 5.13807237e-01 -4.37284529e-01 1.30921602e-03 -7.25608915e-02 3.69088829e-01 -1.87995002e-01 5.47636561e-02 7.36120701e-01 4.54805285e-01 -5.23358643e-01 7.56115258e-01 1.24759448e+00 -1.91449657e-01 -3.47858332e-02 -7.74889350e-01 -3.81307602e-01 -5.70253253e-01 -2.21453714e+00 1.27794528e+00 -6.53321862e-01 9.15359497e-01 5.77888608e-01 -1.11935914e+00 1.19465578e+00 5.18672645e-01 4.03190225e-01 -5.17355800e-01 3.51991355e-01 5.36542594e-01 3.99834216e-01 -7.13546872e-01 1.79365918e-01 1.87769696e-01 1.36811227e-01 2.62487322e-01 -3.88359398e-01 3.41429472e-01 -2.95052975e-01 -3.39339167e-01 8.38787258e-01 -2.13926643e-01 8.51406455e-01 -5.10803945e-02 6.71723783e-01 -4.14780915e-01 8.06323171e-01 8.50500107e-01 8.59735161e-02 7.13332295e-01 -2.85047233e-01 -4.52910125e-01 -6.79645240e-01 -7.50212073e-01 -2.08699226e-01 5.26645422e-01 2.71450609e-01 -8.05468671e-03 -1.77043110e-01 2.25740105e-01 -1.82932213e-01 4.99162048e-01 5.62947616e-02 2.05324084e-01 -9.81109440e-01 -1.11749744e+00 6.75528646e-01 2.80290127e-01 8.43392372e-01 -2.67579377e-01 -3.46029043e-01 4.67827618e-01 -3.50878149e-01 -1.19893217e+00 -1.17409803e-01 -9.98484343e-02 -4.89321530e-01 -7.49620855e-01 -1.06094623e+00 -6.36829495e-01 3.77789617e-01 9.01447117e-01 4.95563269e-01 -4.47881520e-01 -7.13886619e-02 3.46308231e-01 -3.96219015e-01 -5.58144510e-01 7.95620531e-02 -4.00162518e-01 6.21282458e-01 4.68422979e-01 -1.91943929e-01 -1.09019160e+00 -6.29840732e-01 7.50360847e-01 -4.67213333e-01 5.48353679e-02 1.00996971e+00 5.96402705e-01 2.98888385e-01 5.47398090e-01 1.00400960e+00 1.99340153e-02 8.02046001e-01 -5.67613423e-01 -6.71880364e-01 1.18258730e-01 -2.83732116e-01 -2.87768573e-01 9.17518616e-01 -5.34779131e-01 -1.13891578e+00 -2.24061579e-01 -1.35177597e-01 -3.06311607e-01 -2.25886583e-01 9.42595959e-01 -2.30033547e-01 -3.99252713e-01 8.41248333e-01 4.22620714e-01 -3.36026818e-01 -4.53397185e-01 -2.10860819e-02 1.08908784e+00 5.79984188e-01 2.98569687e-02 1.16293740e+00 2.25221038e-01 5.69526374e-01 -1.18547750e+00 -6.49387240e-01 -6.18636310e-01 -1.89383671e-01 -2.36388206e-01 4.08320487e-01 -1.27432466e+00 -5.14548123e-01 5.84685504e-01 -1.24308836e+00 4.09533501e-01 5.36937356e-01 1.49485373e+00 6.26433343e-02 3.80771875e-01 -2.02297777e-01 -1.05421662e+00 -6.21668696e-01 -8.94822717e-01 4.11109060e-01 4.99496698e-01 -9.63032059e-03 -4.81363684e-01 -7.87547380e-02 5.26614785e-02 9.54496741e-01 2.83432961e-01 3.82839978e-01 -5.73508143e-01 -5.40492415e-01 -5.06476939e-01 -2.89904982e-01 1.54852837e-01 3.32025558e-01 -5.93627632e-01 -8.04664791e-01 -3.97590786e-01 3.33863437e-01 3.34272385e-01 8.69364515e-02 9.91384685e-01 8.68754745e-01 -5.16048551e-01 -3.09653252e-01 1.03197646e+00 1.67713892e+00 6.06455326e-01 3.97029430e-01 2.40887716e-01 4.86470610e-01 6.76835030e-02 6.19200706e-01 5.44240952e-01 -2.70818740e-01 6.32692039e-01 1.97533995e-01 1.53096944e-01 -4.81608957e-02 2.12953016e-01 2.37796735e-02 1.29650092e+00 -1.50977761e-01 -5.12286305e-01 -6.98012352e-01 1.32059142e-01 -1.52570486e+00 -1.04574752e+00 -5.46068668e-01 2.00859690e+00 2.91716158e-01 -3.08806419e-01 -3.50926489e-01 3.36552143e-01 6.90291882e-01 1.32507935e-01 -4.76504982e-01 1.09263420e-01 -6.59864426e-01 7.34334961e-02 8.46086323e-01 3.66838038e-01 -9.65710461e-01 8.44743252e-02 6.30190897e+00 1.21973681e+00 -1.22581363e+00 -7.32946116e-03 1.46271497e-01 2.73825645e-01 -6.50786906e-02 -3.52916241e-01 -6.38961077e-01 2.70988911e-01 6.79662764e-01 -8.17604288e-02 5.88185191e-01 6.86572671e-01 3.10937405e-01 6.93122596e-02 -3.86159718e-01 1.86866784e+00 3.40775430e-01 -1.08571196e+00 -4.63305339e-02 -4.42700207e-01 6.87138736e-01 -9.97022241e-02 -2.65280847e-02 5.84785230e-02 -3.95814002e-01 -1.00631225e+00 2.60507584e-01 6.67042434e-01 7.60998607e-01 -8.02589059e-01 9.82351601e-01 6.53548419e-01 -1.31635261e+00 -3.25929224e-01 -3.13247502e-01 -2.85178751e-01 2.88039148e-01 9.02788520e-01 -7.95975268e-01 1.22414041e+00 4.54317749e-01 4.95010197e-01 -1.93925574e-01 1.65171528e+00 -3.68935242e-02 7.31604576e-01 -5.77435672e-01 -2.12461516e-01 -5.22968881e-02 -3.88379872e-01 1.10642815e+00 9.62714314e-01 1.19299436e+00 2.51080751e-01 -1.06551692e-01 1.51089281e-02 3.80285919e-01 4.25176054e-01 -5.72027206e-01 3.05128157e-01 8.50407243e-01 1.27027547e+00 -5.89491546e-01 -1.07267294e-02 -3.66637945e-01 4.75561708e-01 -5.99405646e-01 8.06855857e-01 -6.60251081e-01 -8.61993909e-01 5.56026921e-02 -2.58385956e-01 1.63381726e-01 -5.25180161e-01 -3.40673029e-01 -9.76924241e-01 -1.08090863e-01 -1.12912214e+00 1.79975480e-01 -1.14349496e+00 -9.36821580e-01 4.02810723e-01 -9.99876112e-02 -2.05178738e+00 -3.34564626e-01 -5.68802059e-01 -5.74467301e-01 9.99223888e-01 -1.21232414e+00 -9.13584828e-01 -3.23084414e-01 3.81285429e-01 2.57291973e-01 -7.54703343e-01 7.51860499e-01 7.67481148e-01 -4.66092587e-01 6.10572994e-01 6.53870583e-01 1.99051946e-01 6.63533568e-01 -6.47019804e-01 -2.16742367e-01 1.12199950e+00 1.47034973e-01 8.16162109e-01 1.17111731e+00 -2.45852709e-01 -1.66557503e+00 -6.42288983e-01 1.68531567e-01 3.24809879e-01 2.74584770e-01 1.88526899e-01 -4.00509983e-01 5.04765511e-01 2.39542037e-01 -3.94920632e-02 9.43779826e-01 -1.68087911e-02 -1.68027118e-01 -4.74626362e-01 -7.02516198e-01 4.85184491e-01 7.22867787e-01 1.69858620e-01 -3.39238048e-01 5.77357039e-02 4.58719105e-01 -4.94357884e-01 -7.56480336e-01 8.11767936e-01 8.53191733e-01 -8.46101940e-01 1.36540413e+00 -3.08266040e-02 -2.42251664e-01 -8.91619802e-01 -6.87855542e-01 -1.37488103e+00 -6.48347616e-01 -6.35270476e-01 -1.43946245e-01 1.01835907e+00 3.28377098e-01 -9.22431946e-01 3.04989159e-01 -4.76067245e-01 -1.52795136e-01 -4.23879147e-01 -1.16789973e+00 -7.84245491e-01 -6.10301852e-01 -6.32388115e-01 6.83827043e-01 1.24644411e+00 -5.13421476e-01 5.42391956e-01 -9.79490161e-01 1.06488049e+00 8.85519207e-01 2.17610985e-01 9.31824207e-01 -1.15543437e+00 -6.65053487e-01 -3.72912616e-01 -2.63528734e-01 -1.43441510e+00 -4.51470047e-01 -6.10663176e-01 -3.17931652e-01 -1.52960503e+00 -5.40929019e-01 -4.74686801e-01 -3.10966581e-01 -1.82828635e-01 1.07184552e-01 1.42450064e-01 -1.98850125e-01 1.72727793e-01 -2.51496136e-01 1.98177040e-01 1.06033266e+00 2.84197982e-02 -1.05645582e-01 6.92152619e-01 -5.19626975e-01 1.01173067e+00 7.57351756e-01 -3.21944058e-01 -2.19403028e-01 -7.48007178e-01 8.31815600e-01 7.29798436e-01 2.66750269e-02 -1.41036034e+00 5.81422806e-01 2.81164311e-02 9.59441364e-01 -9.17843699e-01 3.81544381e-01 -9.73542809e-01 7.87892282e-01 4.63565260e-01 1.87055971e-02 5.60680218e-02 2.37650201e-01 5.54972470e-01 -5.43411493e-01 -3.71526182e-01 5.68455517e-01 2.67252684e-01 -5.40248156e-01 1.88245684e-01 -3.39625120e-01 -3.82800817e-01 5.61873317e-01 -2.98234910e-01 -1.55945554e-01 -8.67664158e-01 -3.90060097e-01 3.92223429e-03 -5.58988631e-01 1.01881027e-01 7.32906461e-01 -1.81815219e+00 -8.50318432e-01 3.40730190e-01 -3.07059586e-01 -1.14907920e-01 8.38542581e-01 8.62129569e-01 -6.67793632e-01 4.50123876e-01 -1.17791414e-01 -4.24477398e-01 -1.27607000e+00 1.38466686e-01 4.98828799e-01 1.44804686e-01 -2.81795207e-02 7.37223208e-01 -2.10198119e-01 -2.12458804e-01 1.29704833e-01 1.99762031e-01 -8.07130992e-01 1.53587684e-01 8.94908309e-01 6.27591193e-01 -2.88502842e-01 -8.30102324e-01 -2.26857498e-01 1.16352570e+00 6.40910089e-01 -2.43411213e-01 1.09051824e+00 -2.87398189e-01 -2.29073972e-01 2.74996817e-01 1.40297830e+00 1.01983976e+00 -3.47272009e-01 -3.23977172e-01 -3.99600238e-01 -5.99537015e-01 6.96085617e-02 -7.48225331e-01 -6.10679328e-01 7.15563476e-01 1.03438520e+00 2.77460128e-01 1.36101377e+00 -6.20631099e-01 4.71909374e-01 8.56447816e-01 5.44406891e-01 -5.84114075e-01 -9.29843113e-02 4.04182374e-01 1.06863403e+00 -8.56314182e-01 1.48476735e-01 -2.16383740e-01 -9.09499079e-02 1.58742690e+00 2.90946662e-01 4.44671921e-02 6.70736670e-01 3.64886373e-01 3.86796772e-01 2.54809439e-01 2.76238807e-02 2.23825142e-01 2.35189483e-01 5.98751426e-01 1.65311337e-01 -1.35794310e-02 -2.05207869e-01 9.40884769e-01 -3.69782388e-01 -3.54216754e-01 6.16094410e-01 5.50857663e-01 -7.30082393e-01 -7.19829261e-01 -1.28850293e+00 5.36963999e-01 -7.11617231e-01 -1.28190950e-01 5.37621140e-01 4.96262372e-01 7.26012588e-02 1.03353298e+00 -1.53655753e-01 -3.43147099e-01 4.92928505e-01 -2.90950507e-01 1.77490950e-01 8.18655565e-02 1.40594140e-01 5.72545409e-01 1.88675240e-01 -7.64796287e-02 -3.41917664e-01 -5.96878052e-01 -8.84352505e-01 2.82883167e-01 -5.65510154e-01 4.26313788e-01 6.52968287e-01 6.49788558e-01 3.90020341e-01 6.19864762e-01 9.84525502e-01 -5.88371158e-01 -4.54773337e-01 -9.96053696e-01 -7.52772450e-01 -2.10977107e-01 5.97135782e-01 -4.32956308e-01 -6.09379232e-01 -3.63231748e-01]
[6.523139476776123, 1.2866038084030151]
da350003-e487-4771-9cd2-ab6adf4247ef
geometric-laplacian-eigenmap-embedding
1905.09763
null
https://arxiv.org/abs/1905.09763v2
https://arxiv.org/pdf/1905.09763v2.pdf
GLEE: Geometric Laplacian Eigenmap Embedding
Graph embedding seeks to build a low-dimensional representation of a graph G. This low-dimensional representation is then used for various downstream tasks. One popular approach is Laplacian Eigenmaps, which constructs a graph embedding based on the spectral properties of the Laplacian matrix of G. The intuition behind it, and many other embedding techniques, is that the embedding of a graph must respect node similarity: similar nodes must have embeddings that are close to one another. Here, we dispose of this distance-minimization assumption. Instead, we use the Laplacian matrix to find an embedding with geometric properties instead of spectral ones, by leveraging the so-called simplex geometry of G. We introduce a new approach, Geometric Laplacian Eigenmap Embedding (or GLEE for short), and demonstrate that it outperforms various other techniques (including Laplacian Eigenmaps) in the tasks of graph reconstruction and link prediction.
['Kevin S. Chan', 'Tina Eliassi-Rad', 'Leo Torres']
2019-05-23
null
null
null
null
['graph-reconstruction']
['graphs']
[-2.49004453e-01 5.29523551e-01 -2.05243781e-01 7.90066179e-03 -6.36374354e-02 -7.39589989e-01 4.15375113e-01 3.36144626e-01 1.35300428e-01 1.30563676e-01 4.94171947e-01 -5.25696576e-01 -3.26753497e-01 -9.81483400e-01 -3.95259202e-01 -5.55375874e-01 -5.43333173e-01 3.68845820e-01 -2.28387956e-02 -2.52877921e-01 2.87222385e-01 5.35649896e-01 -8.08012545e-01 -2.81132370e-01 5.04846632e-01 3.20616037e-01 -5.25802933e-02 6.74747109e-01 -3.39775115e-01 1.88046336e-01 8.54997858e-02 -5.84640741e-01 2.03731045e-01 -4.21341628e-01 -8.09317410e-01 -4.94484529e-02 1.90749481e-01 6.16987199e-02 -9.25405800e-01 1.07896292e+00 2.32726619e-01 -6.69543371e-02 1.07658589e+00 -1.51645184e+00 -9.53526795e-01 3.54507178e-01 -6.61397278e-01 3.41416858e-02 5.98962605e-01 -4.83586013e-01 1.70726919e+00 -1.02554274e+00 6.82287037e-01 1.28054214e+00 8.75232160e-01 1.94860101e-01 -1.81126046e+00 -1.86868131e-01 -8.42947140e-02 1.51227280e-01 -1.55414009e+00 -9.11252648e-02 1.06188130e+00 -7.19663441e-01 6.13151610e-01 3.22155386e-01 6.56374454e-01 6.13792062e-01 2.09030330e-01 4.54658598e-01 6.78873479e-01 -5.18534005e-01 6.31586984e-02 -7.24954382e-02 1.73289895e-01 1.16476560e+00 3.73266250e-01 -2.71726131e-01 -3.09466153e-01 -4.60379511e-01 5.13866723e-01 1.31918132e-01 -4.28725243e-01 -1.15882635e+00 -1.22131121e+00 1.11882281e+00 7.27354407e-01 4.34115469e-01 -1.06499456e-01 4.02948499e-01 2.07867876e-01 4.05097306e-01 3.81747633e-01 4.13446993e-01 1.71463996e-01 2.00635806e-01 -5.25411546e-01 -2.20126241e-01 9.62462068e-01 7.91809857e-01 1.23105335e+00 -3.21254402e-01 2.37744585e-01 5.90949774e-01 7.34823585e-01 1.25665858e-01 1.84455186e-01 -5.83557427e-01 4.13211793e-01 1.03371656e+00 -3.64092499e-01 -1.90806842e+00 -4.56162989e-01 -4.16324660e-02 -1.03543806e+00 -1.06940173e-01 1.93447754e-01 1.27742380e-01 -5.48796296e-01 1.77969956e+00 2.70987839e-01 4.26549762e-01 -1.74109861e-01 6.96918726e-01 6.26731336e-01 5.18334389e-01 -3.12731713e-01 6.10066429e-02 9.85481620e-01 -5.44961572e-01 -5.19942105e-01 -1.83823064e-03 7.90526509e-01 -4.64643925e-01 1.00769830e+00 -3.24025929e-01 -7.23161936e-01 -3.80853042e-02 -1.20020545e+00 2.75637750e-02 -4.76195008e-01 -2.15115666e-01 6.35602593e-01 6.99111700e-01 -1.43612754e+00 6.83602631e-01 -6.56477869e-01 -6.54834688e-01 -6.70332462e-02 2.50865579e-01 -6.98505700e-01 -4.03605402e-02 -9.29130971e-01 6.01931274e-01 1.78245410e-01 -9.29975435e-02 -2.94238389e-01 -5.57405055e-01 -1.14337552e+00 1.62278160e-01 7.27557391e-02 -5.93355060e-01 3.03134203e-01 -4.86095816e-01 -1.13873792e+00 1.02522385e+00 -2.02773467e-01 -1.15465507e-01 4.05447781e-02 3.01259637e-01 -3.13426882e-01 2.91917443e-01 -1.03404067e-01 3.65310669e-01 9.84111667e-01 -1.17447138e+00 2.72080123e-01 -4.74803716e-01 -2.09918637e-02 1.79364532e-02 -6.88845754e-01 -3.82884771e-01 -4.53000873e-01 -5.43278635e-01 7.44245112e-01 -1.14990497e+00 -2.84248769e-01 1.61560804e-01 -8.12553406e-01 -1.98929951e-01 8.03669453e-01 -5.76066375e-01 1.61565459e+00 -2.20155120e+00 6.47367954e-01 9.48188186e-01 1.01260912e+00 -2.83048991e-02 -4.25427765e-01 1.12703335e+00 -6.36978507e-01 4.97995496e-01 -3.39620948e-01 -1.31564736e-01 1.95963189e-01 1.55122474e-01 -2.30863929e-01 8.38115811e-01 1.93798259e-01 1.07514966e+00 -1.04404831e+00 -4.88202721e-01 1.55451715e-01 5.81805468e-01 -6.33057535e-01 -1.78277008e-02 1.81098700e-01 -1.42697606e-03 -4.45097417e-01 2.17557922e-01 6.07032180e-01 -3.60739082e-01 6.87234104e-01 -6.18881285e-01 1.61206216e-01 8.54298547e-02 -1.30302286e+00 1.67582738e+00 -2.56005228e-01 7.71829069e-01 7.71280332e-03 -1.31088674e+00 1.03477979e+00 1.70064390e-01 8.24997127e-01 -7.53746694e-03 -8.44852999e-02 -2.84811836e-02 5.16935401e-02 -2.43649945e-01 1.67794093e-01 1.32669449e-01 1.29132673e-01 9.44329679e-01 -1.19957216e-01 2.20066264e-01 2.01750159e-01 9.49014068e-01 1.52845263e+00 -3.25111419e-01 2.71706372e-01 -4.87941951e-01 3.96810561e-01 -4.66528773e-01 1.08728759e-01 2.47242227e-01 -4.27967086e-02 4.77758974e-01 1.02216959e+00 -3.66229922e-01 -1.12836397e+00 -1.26687157e+00 5.97834848e-02 8.36404502e-01 2.50226587e-01 -1.18559980e+00 -6.99182689e-01 -7.43392944e-01 3.35143089e-01 8.12833682e-02 -7.19306529e-01 -4.28375065e-01 -3.62889618e-01 -5.91728806e-01 3.33721340e-01 1.16723455e-01 -1.58437744e-01 -4.41554159e-01 8.87153521e-02 4.11002673e-02 5.62180728e-02 -7.70756900e-01 -8.54080737e-01 -1.58255458e-01 -9.49657142e-01 -1.31502390e+00 -5.29127836e-01 -9.81144786e-01 1.12251782e+00 5.47710657e-01 1.24561095e+00 4.75134850e-01 -3.74023259e-01 7.26947069e-01 -3.53213698e-01 4.48535264e-01 -3.51702780e-01 1.32116124e-01 2.06000045e-01 3.11075419e-01 2.95401841e-01 -1.06023419e+00 -5.25642812e-01 2.31469706e-01 -9.50886846e-01 -6.88928962e-02 5.69256186e-01 7.22594619e-01 6.33888781e-01 -9.04270168e-03 3.35286528e-01 -7.96556652e-01 1.00600815e+00 -5.68913519e-01 -2.97273487e-01 5.02671301e-01 -8.09591413e-01 5.22063971e-01 3.53771865e-01 -1.17719717e-01 1.42395303e-01 7.11175650e-02 1.13819912e-01 -5.34663618e-01 3.87589574e-01 7.42363572e-01 -1.46154806e-01 -4.77025151e-01 4.32979614e-01 1.88616380e-01 2.83360243e-01 -5.54664969e-01 8.57408345e-01 3.72039586e-01 8.96464810e-02 -4.07403499e-01 1.24215829e+00 4.26728010e-01 4.32737321e-01 -1.06743956e+00 -4.19130206e-01 -5.58728218e-01 -9.31394279e-01 -1.69713393e-01 8.34368944e-01 -3.87965351e-01 -8.38905871e-01 -3.06986183e-01 -9.80020046e-01 6.44927695e-02 -6.61772639e-02 3.54510605e-01 -5.74410677e-01 8.08376551e-01 -4.44900304e-01 -5.09424210e-01 -1.01817371e-02 -8.27262044e-01 8.48558247e-01 -3.11214566e-01 -1.81899875e-01 -1.50529325e+00 6.06459975e-01 -2.67669976e-01 4.10387740e-02 3.49296838e-01 1.38714790e+00 -4.28358912e-01 -3.54289740e-01 -3.99498671e-01 -5.27526319e-01 -2.14217789e-02 1.97628662e-01 2.00809285e-01 -4.72676694e-01 -5.06170154e-01 -5.08570492e-01 1.69658661e-01 7.96869516e-01 2.04922110e-01 1.00791490e+00 -4.00725275e-01 -6.70445621e-01 7.99317777e-01 1.61458600e+00 -5.91615856e-01 5.45925975e-01 -5.18157706e-02 1.07594669e+00 5.40693700e-01 -1.55749187e-01 2.27833182e-01 4.79384303e-01 6.81678176e-01 4.80921537e-01 -1.29020393e-01 2.62410585e-02 -6.76755309e-01 4.21636850e-01 1.30972195e+00 6.83146939e-02 2.78433925e-03 -1.02507579e+00 4.79579568e-01 -1.84082580e+00 -8.78104568e-01 -2.83473700e-01 2.19138575e+00 4.27701980e-01 -9.28891003e-02 3.99306655e-01 1.32520899e-01 8.92005563e-01 4.09990966e-01 -3.74653935e-01 -5.53532660e-01 -6.73119202e-02 -2.08809339e-02 4.96760845e-01 7.24320829e-01 -9.60226238e-01 6.88470840e-01 7.33426714e+00 4.13321078e-01 -6.05500281e-01 -1.48402423e-01 2.41488382e-01 6.21201217e-01 -7.66701281e-01 2.04766214e-01 -2.56048560e-01 1.81571096e-01 7.72623122e-01 -5.71317375e-01 6.41587496e-01 6.61254942e-01 -7.60755986e-02 3.83558452e-01 -1.24031687e+00 1.05544841e+00 7.38841146e-02 -1.51196849e+00 1.09894037e-01 5.05927026e-01 6.30489051e-01 -4.38062251e-02 1.07417457e-01 -1.05729625e-01 3.61386001e-01 -1.16210771e+00 3.26851606e-02 4.77365732e-01 6.98233604e-01 -7.23585486e-01 3.10677886e-01 3.99920382e-02 -1.88176286e+00 2.81174034e-01 -5.79222679e-01 2.17939273e-01 2.45063841e-01 7.08243132e-01 -8.86437595e-01 6.07596159e-01 1.79619521e-01 1.12821090e+00 -6.40940189e-01 9.37975168e-01 -1.53291106e-01 3.64110291e-01 -2.45372295e-01 1.12240113e-01 2.08139271e-01 -8.90373290e-01 7.65410662e-01 1.10678089e+00 2.96769142e-01 -1.82908073e-01 2.48279095e-01 1.06207132e+00 -2.96832860e-01 3.32201600e-01 -1.19125688e+00 -4.92278278e-01 4.40389305e-01 1.43802798e+00 -6.33610249e-01 1.48088858e-01 -6.32155836e-01 1.15742803e+00 4.81049985e-01 3.75074148e-01 -5.17692745e-01 -5.58394372e-01 1.07616675e+00 1.76190898e-01 1.56625852e-01 -6.46288514e-01 2.62340717e-02 -1.03290486e+00 6.67664781e-02 -5.13393641e-01 3.28136235e-01 -5.33291936e-01 -1.49542320e+00 4.23246235e-01 -2.55255818e-01 -1.14461625e+00 -6.09827787e-02 -8.61539364e-01 -8.97853374e-01 7.89512575e-01 -1.27191329e+00 -9.37144101e-01 -6.91030622e-02 7.14673936e-01 -2.39718154e-01 8.35385546e-02 1.05222213e+00 2.55385548e-01 -5.63185036e-01 5.51462233e-01 5.08408189e-01 3.40500474e-01 3.21106732e-01 -1.63350773e+00 5.29507279e-01 4.36412930e-01 6.96173668e-01 7.29883790e-01 4.90853816e-01 -5.59075475e-01 -2.02431273e+00 -9.68785346e-01 7.87618041e-01 -4.06227589e-01 1.23549116e+00 -3.83776456e-01 -8.24705482e-01 8.63520145e-01 -6.46052435e-02 1.28977612e-01 9.43865478e-01 3.51887971e-01 -6.97649777e-01 9.57940370e-02 -8.24914634e-01 8.14286232e-01 1.31608915e+00 -9.85795259e-01 -2.26575792e-01 5.38921475e-01 6.26666427e-01 3.41862828e-01 -1.15076196e+00 6.41936511e-02 5.31577706e-01 -9.67524827e-01 1.25514972e+00 -8.68146062e-01 -4.37108949e-02 -3.49861056e-01 -3.12510878e-01 -1.52177966e+00 -6.77746117e-01 -9.28140044e-01 -3.32866102e-01 8.70401859e-01 4.49011296e-01 -7.11358249e-01 1.02284050e+00 2.92922646e-01 1.57834291e-01 -8.16482604e-01 -7.72122324e-01 -8.58372569e-01 3.21456901e-04 -1.15575254e-01 4.76499736e-01 1.21693492e+00 5.74096143e-01 6.43312991e-01 -2.07004651e-01 2.13490739e-01 7.40716338e-01 1.21871442e-01 7.19975591e-01 -1.51131785e+00 -1.26327604e-01 -5.09871602e-01 -1.08270490e+00 -1.00686538e+00 4.44612920e-01 -1.53450751e+00 -4.32102412e-01 -1.69794142e+00 9.23755094e-02 -4.93932486e-01 -1.32376686e-01 1.49525180e-01 -3.93556431e-02 4.54581618e-01 1.20417990e-01 3.25084805e-01 -4.86928225e-01 6.71323717e-01 1.09616756e+00 -8.16346854e-02 -2.84286886e-01 -2.89405853e-01 -5.34009039e-01 6.86977088e-01 5.63134372e-01 -2.49737948e-01 -4.35975850e-01 -3.45664531e-01 5.83738923e-01 -2.28011057e-01 2.11650208e-01 -6.74018145e-01 7.48096630e-02 9.55930054e-02 -9.84353572e-02 -3.28530997e-01 2.81372041e-01 -8.22640479e-01 4.08024222e-01 4.22842473e-01 -1.81014419e-01 4.44894642e-01 -2.96925187e-01 9.27383244e-01 -1.33883566e-01 -2.96301991e-01 4.96839702e-01 1.55533612e-01 -4.72186655e-01 6.36292398e-01 -1.39959246e-01 7.74148777e-02 1.03585184e+00 -3.35889965e-01 -6.29451275e-02 -5.90036035e-01 -9.51684356e-01 -6.97273761e-03 7.21407950e-01 1.16856486e-01 8.61275494e-01 -1.89436221e+00 -6.76613212e-01 2.55329669e-01 3.62770736e-01 -6.49879694e-01 -3.15848976e-01 1.14473712e+00 -5.71957469e-01 1.92585379e-01 1.35332003e-01 -4.56603974e-01 -1.10398757e+00 7.87011504e-01 2.75618702e-01 -2.50077546e-01 -9.56418812e-01 6.48170471e-01 2.64066845e-01 -4.91723895e-01 1.60927791e-02 1.49671182e-01 -3.82272042e-02 9.24331173e-02 3.19721907e-01 3.74848366e-01 -1.57159135e-01 -9.13737416e-01 -4.63092148e-01 9.95141506e-01 3.20975870e-01 -2.74215434e-02 1.32561934e+00 -1.36405081e-01 -5.67664862e-01 3.81178170e-01 1.76551223e+00 2.13093236e-01 -6.79755032e-01 -3.08979303e-01 3.70671660e-01 -6.27742589e-01 -2.09075399e-02 1.45226672e-01 -1.09522760e+00 1.00712430e+00 2.92227477e-01 9.38244104e-01 7.08438158e-01 1.98515862e-01 5.61344802e-01 5.05793512e-01 4.13929410e-02 -6.47037268e-01 2.68238097e-01 3.33225459e-01 9.32065129e-01 -9.60267067e-01 2.12784503e-02 -5.81411183e-01 -2.92569011e-01 1.22483051e+00 7.37001449e-02 -4.89235282e-01 1.22494125e+00 -2.05270514e-01 -3.31232280e-01 -5.43716729e-01 -4.43418235e-01 -4.17407513e-01 5.93632638e-01 7.60357261e-01 4.89183307e-01 1.54067785e-01 -2.07997099e-01 -3.92603502e-02 -9.85376090e-02 -8.47146869e-01 2.96410471e-01 4.13529336e-01 -4.47040170e-01 -1.36229193e+00 -1.29881993e-01 5.03348291e-01 1.97080135e-01 -1.36749506e-01 -9.52444017e-01 5.01349688e-01 -4.91477787e-01 7.28326261e-01 1.67057402e-02 -8.55430663e-01 1.23709939e-01 1.58035412e-01 3.74973267e-01 -7.07869291e-01 1.66591659e-01 -2.39982262e-01 -4.86544929e-02 -7.22769737e-01 -1.68581814e-01 -3.92125756e-01 -1.04783487e+00 -7.24250793e-01 -3.40789914e-01 2.88633376e-01 7.84000516e-01 4.92252469e-01 5.84660292e-01 9.48293656e-02 9.98777509e-01 -7.13131964e-01 -3.85439426e-01 -5.67398012e-01 -9.32418287e-01 4.83620793e-01 2.69602478e-01 -6.17913008e-01 -4.85357016e-01 -1.77034855e-01]
[7.129220962524414, 5.837766170501709]
1ffbe726-dd2f-4925-97f0-669bf141d191
a-tensor-based-sub-mode-coordinate-algorithm
1805.07979
null
http://arxiv.org/abs/1805.07979v1
http://arxiv.org/pdf/1805.07979v1.pdf
A Tensor-Based Sub-Mode Coordinate Algorithm for Stock Prediction
The investment on the stock market is prone to be affected by the Internet. For the purpose of improving the prediction accuracy, we propose a multi-task stock prediction model that not only considers the stock correlations but also supports multi-source data fusion. Our proposed model first utilizes tensor to integrate the multi-sourced data, including financial Web news, investors' sentiments extracted from the social network and some quantitative data on stocks. In this way, the intrinsic relationships among different information sources can be captured, and meanwhile, multi-sourced information can be complemented to solve the data sparsity problem. Secondly, we propose an improved sub-mode coordinate algorithm (SMC). SMC is based on the stock similarity, aiming to reduce the variance of their subspace in each dimension produced by the tensor decomposition. The algorithm is able to improve the quality of the input features, and thus improves the prediction accuracy. And the paper utilizes the Long Short-Term Memory (LSTM) neural network model to predict the stock fluctuation trends. Finally, the experiments on 78 A-share stocks in CSI 100 and thirteen popular HK stocks in the year 2015 and 2016 are conducted. The results demonstrate the improvement on the prediction accuracy and the effectiveness of the proposed model.
['Jieyun Huang', 'Jialai Zhang', 'Yunjia Zhang', 'Xi Zhang']
2018-05-21
null
null
null
null
['stock-prediction']
['time-series']
[-9.14225340e-01 -8.50171804e-01 -1.32832453e-01 -1.23669855e-01 -1.77416161e-01 -2.62698025e-01 3.60039592e-01 -2.11931154e-01 -2.47579604e-01 4.36388046e-01 7.71351337e-01 2.44572327e-01 -3.51914674e-01 -1.03671885e+00 -2.65190721e-01 -7.83952415e-01 -1.98333729e-02 -1.28735319e-01 2.72992462e-01 -5.47149360e-01 5.64231873e-01 9.99558493e-02 -1.47500873e+00 2.70317703e-01 7.80813634e-01 1.65521502e+00 2.34209806e-01 -3.30530345e-01 -6.44600034e-01 1.28146493e+00 -3.62771779e-01 -5.15349925e-01 4.71674949e-01 -1.66757067e-03 3.80028388e-03 1.63286403e-01 -2.02448204e-01 -4.14953977e-01 -1.51979610e-01 1.33359551e+00 3.27229261e-01 -1.88259527e-01 7.52251968e-02 -8.91349912e-01 -7.14192152e-01 9.54986453e-01 -6.67057455e-01 4.58754927e-01 -3.75868648e-01 -1.04304023e-01 1.17755246e+00 -1.20339262e+00 3.62002999e-01 1.09846461e+00 5.98460853e-01 -3.93681437e-01 -5.69042623e-01 -8.41727078e-01 1.36369601e-01 4.77555573e-01 -9.68149602e-01 -3.17941755e-02 1.11329877e+00 -6.11562073e-01 3.49300951e-01 7.76442233e-03 1.12037456e+00 4.65609580e-01 4.44766194e-01 7.81082928e-01 1.21939206e+00 2.59434193e-01 -1.48586348e-01 3.49825829e-01 3.49686384e-01 2.58043706e-01 6.23635948e-01 -1.39078721e-01 -9.07033980e-01 -1.52737096e-01 4.42049086e-01 6.53535903e-01 -2.32520059e-01 1.96592912e-01 -1.48577249e+00 7.36793995e-01 4.59479332e-01 8.07948887e-01 -7.81932950e-01 -2.85573542e-01 5.03032863e-01 4.60961670e-01 9.55835164e-01 1.03482567e-01 -6.36682630e-01 -1.74130976e-01 -9.30370688e-01 1.94241151e-01 5.31972706e-01 4.11515772e-01 6.86681569e-01 6.79512382e-01 6.56043813e-02 4.44932014e-01 6.02568090e-01 7.13051736e-01 1.18734717e+00 -6.28857970e-01 7.85183489e-01 9.95727420e-01 1.02983601e-01 -1.71354532e+00 -4.81477559e-01 -1.07351387e+00 -1.06631410e+00 -1.61533147e-01 -4.16119322e-02 -2.94423878e-01 -6.95599914e-02 1.14107072e+00 1.12487882e-01 3.32559705e-01 2.55013078e-01 9.46732044e-01 8.42623830e-01 9.33120847e-01 -3.33614469e-01 -5.64683378e-01 1.23859632e+00 -7.03428447e-01 -1.18063438e+00 1.74998745e-01 4.68208462e-01 -8.39828312e-01 2.61963874e-01 4.16793525e-01 -8.93396676e-01 -6.15357339e-01 -1.06357777e+00 4.11200076e-01 -4.89906430e-01 3.45638573e-01 4.53376144e-01 1.88431233e-01 -6.52494073e-01 7.89007843e-01 -5.89531064e-01 4.96204555e-01 4.72902600e-03 1.50093973e-01 -6.50886446e-02 3.08012992e-01 -1.52893901e+00 9.73771036e-01 7.17217088e-01 5.94808340e-01 6.10521585e-02 -7.09120929e-01 -2.42070809e-01 1.85669586e-01 2.58263975e-01 -1.89861864e-01 7.04640925e-01 -9.54292834e-01 -1.24995100e+00 -2.47320101e-01 2.23105714e-01 -4.36662376e-01 3.44620794e-01 -8.89916793e-02 -7.81682134e-01 -8.94516930e-02 -1.93558354e-02 -2.10941523e-01 5.67591965e-01 -6.86673343e-01 -8.73171151e-01 -6.31042302e-01 -4.27708864e-01 1.63280204e-01 -1.00288320e+00 3.65979597e-02 -1.23183936e-01 -9.72031295e-01 6.24051392e-01 -6.63281322e-01 8.74125510e-02 -7.55450964e-01 -1.18298635e-01 -1.14665367e-01 8.04860473e-01 -1.07224059e+00 1.29608798e+00 -2.08576775e+00 1.30039766e-01 3.72283548e-01 2.30584979e-01 2.01386735e-01 2.17494462e-02 5.23546338e-01 1.24725187e-02 -4.33631241e-02 2.90542722e-01 -2.47030601e-01 -9.15547013e-02 -2.93934084e-02 -7.16601968e-01 3.05608481e-01 -4.53575179e-02 7.87848592e-01 -3.88327628e-01 -3.12247366e-01 -3.66967097e-02 2.16731980e-01 2.44324356e-02 -2.65902430e-02 -1.94407199e-02 8.89746398e-02 -6.71531200e-01 5.75185478e-01 9.48704422e-01 -2.37276971e-01 -1.52429506e-01 -5.63052118e-01 -6.05213881e-01 1.08317547e-01 -1.59674585e+00 1.09638751e+00 -1.07128322e-01 5.40466458e-02 -2.14228421e-01 -7.39954174e-01 1.38510430e+00 2.13556230e-01 7.72747874e-01 -6.70732081e-01 2.09595993e-01 7.06904531e-01 -2.30634995e-02 -4.55675006e-01 4.53331918e-01 -1.09331347e-01 2.53473520e-01 4.59899545e-01 -1.51235014e-01 5.35171747e-01 2.20897585e-01 8.16107243e-02 3.18244994e-01 -6.51280284e-02 -6.53007776e-02 -3.46047729e-01 5.95723569e-01 -2.01205000e-01 9.40067053e-01 -1.71632379e-01 2.12370768e-01 -8.56185518e-03 4.39722985e-01 -5.89155376e-01 -7.20822871e-01 -3.16070616e-01 -3.73668298e-02 4.60860640e-01 6.88048154e-02 -2.39711642e-01 -5.26606068e-02 -1.66262507e-01 4.27655131e-01 4.64002728e-01 -5.30425906e-01 -1.40445411e-01 -1.66782320e-01 -1.27533948e+00 7.38607496e-02 3.79463077e-01 9.83983219e-01 -6.50978386e-01 -1.09255441e-01 4.03891295e-01 -1.98703706e-01 -6.71682537e-01 -2.68038541e-01 -2.72660226e-01 -1.16569364e+00 -7.32027948e-01 -8.95381689e-01 -2.27280498e-01 1.09859943e-01 4.55638677e-01 6.15302265e-01 3.19900252e-02 7.88639605e-01 -1.91245571e-01 -4.55604315e-01 -4.84447598e-01 2.02638179e-01 -7.62917474e-02 2.90884227e-01 8.47403228e-01 4.88522261e-01 -5.33952773e-01 -1.74197569e-01 1.29960611e-01 -7.80752659e-01 6.87026456e-02 7.66764581e-01 8.06937337e-01 3.14750433e-01 5.41196525e-01 8.06952715e-01 -3.22952628e-01 7.24442184e-01 -8.37933600e-01 -8.92105103e-01 1.02246195e-01 -1.10149622e+00 8.21648464e-02 5.53572118e-01 -3.18082064e-01 -1.18606281e+00 -4.86889958e-01 2.88740009e-01 -5.99445403e-01 7.14836001e-01 1.38491297e+00 9.55947116e-02 -1.06737595e-02 -2.39340112e-01 7.52906501e-01 1.91555128e-01 -9.57782388e-01 -1.27246633e-01 5.16194046e-01 -9.17032436e-02 -1.62211917e-02 9.43203330e-01 1.74973622e-01 1.15495913e-01 -3.88989151e-01 -7.79978096e-01 -2.57954448e-01 -4.68675494e-01 -3.11117411e-01 4.79048461e-01 -1.27904141e+00 -8.32743704e-01 9.87445831e-01 -1.12345576e+00 6.49788618e-01 6.47528693e-02 1.06804430e+00 2.44940907e-01 1.90549880e-01 -1.06335723e+00 -9.22678709e-01 -5.32067716e-01 -9.54627454e-01 4.65507179e-01 2.90725827e-01 5.25236189e-01 -8.89275610e-01 2.42634248e-02 1.81601495e-01 8.13811123e-01 7.62463138e-02 5.40772200e-01 -1.02890301e+00 -9.32533026e-01 -2.25554928e-01 -3.20728123e-01 6.25232816e-01 1.46658748e-01 1.09658115e-01 -4.36666191e-01 -1.42524727e-02 6.55262351e-01 1.74705125e-02 1.01469946e+00 2.09174797e-01 4.44103330e-01 -5.66457450e-01 3.30487698e-01 3.65201771e-01 1.42158294e+00 3.86589617e-01 2.87344843e-01 7.57189572e-01 8.08309257e-01 7.34041810e-01 7.25673497e-01 6.77307248e-01 6.59738660e-01 2.24039286e-01 2.53298223e-01 1.83870375e-01 5.03678441e-01 -7.24788755e-02 5.56436837e-01 1.95618558e+00 -4.17991608e-01 3.56150717e-01 -6.28820121e-01 1.84778422e-01 -2.01646733e+00 -1.22790408e+00 -4.74120557e-01 1.70837498e+00 5.17595112e-01 2.62829006e-01 2.28123844e-01 3.17443490e-01 6.04901254e-01 4.61122006e-01 -4.98396605e-01 4.92294192e-01 -6.00511849e-01 -5.24133623e-01 7.68363476e-01 -3.74560691e-02 -8.69637728e-01 3.47454816e-01 4.88521767e+00 7.46355832e-01 -1.40208972e+00 1.17106728e-01 6.72280490e-01 -2.64663696e-01 -6.45014882e-01 -2.82486320e-01 -9.42798376e-01 1.01373744e+00 7.91702867e-01 -6.67440474e-01 2.82761514e-01 8.88967156e-01 4.68457580e-01 1.49264678e-01 -1.23201951e-01 8.51895750e-01 7.73185790e-02 -1.52956414e+00 1.03196926e-01 2.34464005e-01 6.77059054e-01 2.95016021e-01 3.33257675e-01 7.01891407e-02 -1.53520122e-01 -1.11280873e-01 1.05405629e+00 1.23443437e+00 -4.69999984e-02 -9.51368034e-01 1.40224600e+00 4.11224067e-01 -1.47577131e+00 -3.53444993e-01 -4.96347040e-01 -2.50225574e-01 2.68130534e-04 1.11914897e+00 -3.61588672e-02 9.82629716e-01 7.57824659e-01 1.49150598e+00 -7.04774439e-01 9.21237171e-01 2.37029627e-01 3.43791813e-01 -1.27199218e-01 -5.02527475e-01 3.57208669e-01 -9.28406477e-01 3.56993496e-01 5.14466465e-01 8.48601818e-01 1.35755748e-01 1.29038952e-02 6.62536383e-01 1.72264501e-01 5.69347382e-01 -2.92631269e-01 -1.70911565e-01 2.63183653e-01 1.16152513e+00 -2.84050375e-01 -5.62645316e-01 -6.76550329e-01 2.14745179e-01 -7.20048919e-02 2.43870154e-01 -4.49752241e-01 -3.15045148e-01 4.84888941e-01 -1.71704382e-01 5.17097354e-01 -3.01021695e-01 -4.94984567e-01 -1.67950368e+00 4.20246124e-01 -7.13652551e-01 8.67705718e-02 -6.47986054e-01 -1.42971492e+00 4.87963289e-01 -3.42826694e-01 -1.87257552e+00 -7.57626221e-02 -5.78786552e-01 -6.06394291e-01 1.05991042e+00 -1.76719177e+00 -8.66361678e-01 1.48068801e-01 5.56442738e-01 8.13196823e-02 -8.32576752e-01 2.98182189e-01 5.35496056e-01 -8.30550551e-01 -2.33431816e-01 7.19646573e-01 2.86583364e-01 3.37119907e-01 -9.11098480e-01 3.02774198e-02 9.95554090e-01 -1.01812139e-01 6.33939028e-01 2.64884651e-01 -9.91655052e-01 -1.60412276e+00 -1.05941689e+00 8.33748698e-01 3.14993829e-01 1.29099393e+00 2.84156650e-01 -9.99367476e-01 5.96979320e-01 3.78909744e-02 -2.06095517e-01 6.96670890e-01 4.11940999e-02 -3.79038930e-01 -8.41313779e-01 -8.56183469e-01 2.31352746e-01 7.86858723e-02 -3.11890215e-01 -6.88131750e-01 1.37593165e-01 8.30384314e-01 -5.16158789e-02 -1.23771179e+00 2.43244514e-01 4.50447351e-01 -1.16473758e+00 7.03331709e-01 -3.56597781e-01 3.81610185e-01 -4.51667994e-01 -1.82053640e-01 -1.47099781e+00 -6.57814026e-01 6.59999698e-02 -1.68646201e-01 1.50124955e+00 3.04166853e-01 -9.92594302e-01 3.99899125e-01 6.72530890e-01 7.57372454e-02 -6.90489233e-01 -1.02931011e+00 -6.58306599e-01 -2.72304952e-01 -2.64345855e-01 1.07524824e+00 1.00143611e+00 1.57383919e-01 1.80651635e-01 -7.49900401e-01 -3.28853279e-02 5.85001171e-01 5.09606481e-01 4.05263662e-01 -1.40005147e+00 -1.32994294e-01 -6.01590455e-01 -2.55911291e-01 -6.40554607e-01 -2.49898508e-02 -6.60239697e-01 -7.36232340e-01 -1.20925117e+00 -7.35197728e-03 -1.66037291e-01 -7.27393389e-01 1.84879184e-01 -1.34434640e-01 -5.89892566e-02 6.57673717e-01 8.54032636e-01 -3.19395870e-01 9.29179728e-01 1.31235063e+00 -1.32282123e-01 -8.68341997e-02 2.34783530e-01 -6.90357149e-01 6.90398335e-01 7.27538466e-01 -1.52188152e-01 4.28156555e-03 -4.22580212e-01 5.51357031e-01 3.41365784e-01 -7.14845806e-02 -8.29000235e-01 2.78704733e-01 -2.08664805e-01 5.32873154e-01 -9.88773823e-01 3.79722983e-01 -9.45170879e-01 3.54556769e-01 8.02376449e-01 -2.70386487e-02 4.57394689e-01 -5.69723286e-02 3.21463883e-01 -8.44817817e-01 -1.36063233e-01 3.43416125e-01 -1.84792995e-01 -5.26222348e-01 4.36499417e-01 -9.85477492e-02 -4.44151193e-01 7.40645528e-01 3.12572777e-01 -5.72638929e-01 -2.70759165e-01 -3.31623763e-01 4.27180022e-01 -1.11595951e-01 4.98748541e-01 5.94158530e-01 -1.81904089e+00 -9.51566935e-01 1.91161126e-01 -2.14212939e-01 -5.63822567e-01 3.08265239e-01 1.05510008e+00 -3.44595909e-01 5.38286746e-01 -2.44496614e-01 -1.23427764e-01 -7.61081100e-01 5.68694413e-01 2.84955978e-01 -3.30188781e-01 -8.88817683e-02 4.37624186e-01 -3.57613325e-01 5.47818281e-02 -1.54327482e-01 -4.09637123e-01 -9.33726788e-01 9.03961897e-01 8.25853765e-01 6.27120495e-01 1.50522366e-01 -1.00721300e+00 -1.10408813e-01 8.21723938e-01 4.10700850e-02 -1.02478109e-01 1.84446323e+00 -2.60884613e-01 -7.20490396e-01 1.11932468e+00 1.21652710e+00 2.14716122e-01 -6.70594513e-01 -7.06857383e-01 3.22649062e-01 -5.84365070e-01 3.28049749e-01 -4.05434638e-01 -1.74780989e+00 6.40055358e-01 4.79920298e-01 6.76269412e-01 1.10885108e+00 -6.29240215e-01 1.04365683e+00 4.45023835e-01 3.29093367e-01 -1.08570588e+00 2.88780481e-02 4.43193942e-01 8.68538141e-01 -1.01035762e+00 1.19915858e-01 -2.34324373e-02 -8.55236173e-01 1.41382051e+00 2.64912635e-01 -2.76022643e-01 1.13472545e+00 -4.69296128e-02 2.26144210e-01 -1.61652938e-01 -7.64378607e-01 9.87976231e-03 4.48430330e-01 -3.80346417e-01 1.34883732e-01 -6.02550805e-02 -4.98257548e-01 1.01568317e+00 -3.49776983e-01 8.59772190e-02 6.55907810e-01 5.32910764e-01 -5.52591264e-01 -8.81488502e-01 -6.51328444e-01 7.47534454e-01 -6.75671518e-01 -1.74782231e-01 9.08283889e-02 1.51009783e-01 2.72907853e-01 7.83798158e-01 8.14457685e-02 -8.94966304e-01 2.95171112e-01 2.17166916e-02 -5.39312363e-01 -1.20262958e-01 -5.59312165e-01 5.22245228e-01 -1.95000187e-01 -3.44300658e-01 -7.14736104e-01 -9.87342358e-01 -8.44134033e-01 -4.50862527e-01 -3.70600104e-01 3.27093989e-01 7.23843753e-01 1.10574448e+00 3.95627409e-01 5.48792958e-01 1.31197262e+00 -6.59682989e-01 -7.73135126e-01 -1.17256236e+00 -1.35016656e+00 1.95587412e-01 3.41513813e-01 -7.60507166e-01 -6.30747676e-01 -2.26315901e-01]
[4.538118839263916, 4.181317329406738]
85743fc2-7625-4eb4-9fc5-1427c6c07590
a-deep-convolutional-neural-network-using
1610.09736
null
http://arxiv.org/abs/1610.09736v3
http://arxiv.org/pdf/1610.09736v3.pdf
A deep convolutional neural network using directional wavelets for low-dose X-ray CT reconstruction
Due to the potential risk of inducing cancers, radiation dose of X-ray CT should be reduced for routine patient scanning. However, in low-dose X-ray CT, severe artifacts usually occur due to photon starvation, beamhardening, etc, which decrease the reliability of diagnosis. Thus, high quality reconstruction from low-dose X-ray CT data has become one of the important research topics in CT community. Conventional model-based denoising approaches are, however, computationally very expensive, and image domain denoising approaches hardly deal with CT specific noise patterns. To address these issues, we propose an algorithm using a deep convolutional neural network (CNN), which is applied to wavelet transform coefficients of low-dose CT images. Specifically, by using a directional wavelet transform for extracting directional component of artifacts and exploiting the intra- and inter-band correlations, our deep network can effectively suppress CT specific noises. Moreover, our CNN is designed to have various types of residual learning architecture for faster network training and better denoising. Experimental results confirm that the proposed algorithm effectively removes complex noise patterns of CT images, originated from the reduced X-ray dose. In addition, we show that wavelet domain CNN is efficient in removing the noises from low-dose CT compared to an image domain CNN. Our results were rigorously evaluated by several radiologists and won the second place award in 2016 AAPM Low-Dose CT Grand Challenge. To the best of our knowledge, this work is the first deep learning architecture for low-dose CT reconstruction that has been rigorously evaluated and proven for its efficacy.
['junhong Min', 'Jong Chul Ye', 'Eunhee Kang']
2016-10-31
null
null
null
null
['low-dose-x-ray-ct-reconstruction']
['medical']
[ 2.47089893e-01 -2.14484259e-01 2.24908274e-02 -3.61167729e-01 -9.24774051e-01 8.72175917e-02 4.55769300e-02 1.27401158e-01 -5.72311044e-01 3.93387824e-01 4.37726378e-01 -2.26397008e-01 -3.87287289e-01 -1.01680470e+00 -6.40524328e-01 -1.01824117e+00 6.27754927e-02 9.87309217e-02 1.74099773e-01 -3.31239879e-01 -2.27620170e-01 5.48194468e-01 -9.37233984e-01 5.53484440e-01 5.98451793e-01 9.14055824e-01 4.47619557e-01 4.90878195e-01 1.19891204e-01 8.14233124e-01 -4.30680752e-01 2.54008882e-02 1.45836711e-01 -6.36620820e-01 -5.10451138e-01 -1.04938813e-01 -4.55618687e-02 -7.21887648e-01 -8.73203278e-01 1.14792752e+00 9.88867939e-01 1.58466116e-01 2.71691024e-01 -4.66886550e-01 -4.43797112e-01 5.62124491e-01 -4.78419751e-01 6.55651331e-01 -8.44670907e-02 3.13774616e-01 4.33438122e-01 -7.19464660e-01 4.13027793e-01 8.00658226e-01 8.80704045e-01 6.90054774e-01 -9.37864423e-01 -6.36906266e-01 -3.03161263e-01 2.60265678e-01 -7.29959846e-01 1.73787057e-01 1.00503099e+00 -6.06195740e-02 6.42207146e-01 3.04788709e-01 8.29780221e-01 1.21837294e+00 5.70602953e-01 4.90772814e-01 1.10752058e+00 -1.86824977e-01 8.80516917e-02 -4.99515593e-01 -1.17918625e-02 5.59286356e-01 2.66583025e-01 3.57766867e-01 -1.73829794e-01 3.68267260e-02 8.86086643e-01 3.49076092e-01 -8.89791727e-01 7.46271387e-02 -1.07092130e+00 7.09181309e-01 7.46593416e-01 8.43840778e-01 -6.16285205e-01 3.74118567e-01 6.33039773e-01 3.30389440e-02 4.78424072e-01 6.66736439e-02 -2.06124827e-01 -5.83787039e-02 -6.38028741e-01 -1.63161606e-02 3.14713717e-01 3.04882705e-01 -4.84661721e-02 1.07801676e-01 -1.08957961e-01 9.00807977e-01 8.69283900e-02 2.10773140e-01 7.75902629e-01 -7.19464481e-01 2.65430897e-01 1.00441597e-01 -5.42461336e-01 -8.32628250e-01 -7.25747406e-01 -9.29016292e-01 -1.64163184e+00 1.70785442e-01 3.01308632e-01 3.89846344e-03 -1.10654259e+00 1.51475811e+00 1.32452890e-01 4.04226273e-01 -8.75863805e-02 1.24925423e+00 1.28940475e+00 5.41423142e-01 2.78933853e-01 -4.60623801e-01 1.44830561e+00 -5.74571311e-01 -9.68567312e-01 4.76605259e-03 3.80050153e-01 -8.30772460e-01 9.55969393e-01 6.48496985e-01 -1.33966768e+00 -6.15454912e-01 -1.28487015e+00 -9.88771245e-02 3.89651954e-01 4.55506658e-03 5.35613716e-01 5.76665342e-01 -6.77279651e-01 9.44000423e-01 -1.24863625e+00 1.57269593e-02 6.91885412e-01 2.80436635e-01 -1.94909468e-01 -4.89553958e-01 -1.25466561e+00 8.34313571e-01 -2.40739528e-02 3.50658685e-01 -7.60200381e-01 -1.21499467e+00 -5.52637935e-01 9.46727246e-02 3.01596969e-01 -7.07736313e-01 1.34146559e+00 -5.71955144e-01 -1.30144024e+00 4.37576294e-01 1.89645484e-01 -4.47887599e-01 5.38692534e-01 -3.34885977e-02 -5.93341112e-01 5.19503117e-01 -5.18476963e-02 3.63257937e-02 6.44014657e-01 -1.22287846e+00 -3.02673817e-01 -3.57362241e-01 -3.07272941e-01 -4.21877168e-02 -2.86755294e-01 -1.36802405e-01 -3.80575627e-01 -9.85593617e-01 6.23201549e-01 -5.91426015e-01 -5.08423686e-01 6.48545399e-02 -6.43204972e-02 1.73318774e-01 6.61184907e-01 -7.99681962e-01 1.10208309e+00 -2.31846547e+00 -2.47291192e-01 1.99676543e-01 4.88601446e-01 9.06626433e-02 3.92713137e-02 4.98988330e-02 -6.19736135e-01 1.41092166e-01 -4.29705739e-01 3.86920609e-02 -5.88008404e-01 3.15951139e-01 4.12169211e-02 6.00491464e-01 -6.79985955e-02 7.31081188e-01 -9.04286385e-01 -3.04064006e-01 3.49493742e-01 7.70044327e-01 -3.34429383e-01 -7.66301528e-02 1.96556970e-01 7.66639709e-01 -5.50361633e-01 3.65436524e-01 1.04758310e+00 -1.17132135e-01 4.95821796e-02 -7.95956433e-01 -1.24394689e-02 8.29973668e-02 -7.37259865e-01 1.67539036e+00 -6.38491035e-01 4.01909292e-01 2.86646485e-01 -9.94315982e-01 4.64481562e-01 5.90287507e-01 9.24298525e-01 -1.22634780e+00 4.35009480e-01 3.99989069e-01 3.08935076e-01 -7.42752790e-01 -6.61185682e-02 -8.23317528e-01 3.83988738e-01 3.37594330e-01 -2.65251130e-01 -4.64921266e-01 -2.07641080e-01 -2.23557632e-02 1.40551341e+00 -4.46882099e-01 -8.03996623e-02 -8.44789371e-02 3.96385163e-01 -7.08691180e-02 4.77266967e-01 6.94331706e-01 -2.46102974e-01 1.01841056e+00 2.31608674e-01 -7.04258502e-01 -1.15617812e+00 -9.46390331e-01 -5.30935109e-01 3.66541982e-01 6.84850737e-02 1.09309703e-02 -6.56507790e-01 -2.69534916e-01 -4.26126599e-01 4.32343602e-01 -5.36888659e-01 -3.94985884e-01 -1.27823329e+00 -1.21834683e+00 4.76270080e-01 6.50121748e-01 8.09647918e-01 -1.03602827e+00 -5.05328417e-01 5.60445905e-01 -5.45411587e-01 -9.60525572e-01 -4.47305620e-01 5.37201583e-01 -1.29400885e+00 -1.15096736e+00 -8.54290843e-01 -8.24645400e-01 5.73587596e-01 4.55774963e-01 1.12777233e+00 5.79695106e-01 -5.82323134e-01 1.95863873e-01 -3.01027268e-01 -2.57725984e-01 -4.85393465e-01 -2.53893971e-01 -3.47078651e-01 -4.86049533e-01 2.82792468e-02 -6.63349092e-01 -1.07224178e+00 6.35679215e-02 -1.35613406e+00 -1.56999737e-01 7.31110036e-01 1.26561248e+00 8.47478271e-01 7.46412873e-01 3.46585244e-01 -9.55083787e-01 7.95893312e-01 -3.49039167e-01 -1.66040629e-01 -1.78304195e-01 -2.23691717e-01 -1.61386847e-01 8.26777220e-01 -1.80629030e-01 -1.28502965e+00 -1.98603913e-01 -8.83166730e-01 -1.69472709e-01 -8.43161866e-02 5.98618090e-01 1.89844295e-01 -2.53934652e-01 7.81040549e-01 3.08111161e-01 -1.03918560e-01 -3.06720912e-01 -2.44032413e-01 1.31917730e-01 5.55756330e-01 -3.23216379e-01 7.20343113e-01 8.27187955e-01 4.54619288e-01 -7.26301849e-01 -8.78296494e-01 -3.72211218e-01 -1.61597818e-01 -2.51493275e-01 1.08453012e+00 -8.76728714e-01 -7.79404640e-01 5.06604552e-01 -9.37954128e-01 -6.28736094e-02 -3.15654069e-01 1.01289976e+00 -2.54407734e-01 6.70444787e-01 -1.20884407e+00 -1.22102976e-01 -5.67226887e-01 -1.38758004e+00 6.05207801e-01 1.56969130e-02 2.15844899e-01 -7.71490633e-01 -3.98688912e-02 2.17464164e-01 5.78454554e-01 4.82468575e-01 1.21047354e+00 -7.50968978e-02 -5.90222538e-01 -1.62553206e-01 -1.81470990e-01 6.04485989e-01 1.06413223e-01 -3.65275592e-01 -7.46694267e-01 -3.25796336e-01 1.01590848e+00 -8.13325867e-02 1.00342381e+00 1.06719446e+00 1.78708541e+00 3.16550940e-01 1.04311630e-02 9.96693611e-01 1.70074713e+00 3.05871904e-01 8.34181786e-01 3.96379352e-01 4.17776614e-01 1.67583615e-01 3.15597653e-01 4.18605506e-01 -1.85987592e-01 2.13432848e-01 6.59768403e-01 -5.85901916e-01 -4.30896759e-01 2.39780441e-01 -3.81842941e-01 8.78009379e-01 -4.46794331e-01 -1.90991402e-01 -8.77623737e-01 3.82693022e-01 -1.27625155e+00 -8.86712074e-01 -7.81418502e-01 1.79561388e+00 7.14702368e-01 1.87253341e-01 -3.41694832e-01 3.05785149e-01 4.62480754e-01 3.08460109e-02 -3.63195539e-01 4.11219597e-02 -1.70226067e-01 1.00456381e+00 6.19593561e-01 1.47730216e-01 -9.39963758e-01 -3.36948596e-02 5.66606998e+00 8.62284780e-01 -1.22621572e+00 5.94446242e-01 7.46222913e-01 -1.75745502e-01 -2.29711801e-01 -4.68920439e-01 3.27596813e-02 4.01503682e-01 5.17029822e-01 5.35758883e-02 6.33987933e-02 5.44678450e-01 5.91947079e-01 -1.53449178e-01 -6.94743156e-01 9.98271585e-01 -2.74631321e-01 -1.32883334e+00 -3.17520738e-01 -5.33668548e-02 5.91680586e-01 6.84570000e-02 1.52270079e-01 1.28966838e-01 -5.40031902e-02 -1.08231366e+00 2.91105300e-01 2.81368047e-01 5.61086595e-01 -1.04915857e+00 1.17180276e+00 2.34371275e-01 -8.56522858e-01 2.77117789e-02 -4.10502523e-01 3.30217183e-01 4.19987559e-01 1.08935666e+00 -3.47548306e-01 7.35513449e-01 1.01051605e+00 5.27389646e-01 -1.24847583e-01 9.71295416e-01 -1.71153709e-01 8.01466644e-01 -1.89121619e-01 4.07208115e-01 3.21755409e-01 -4.60988618e-02 3.72514874e-01 1.00803792e+00 4.86639977e-01 7.97558963e-01 -5.29977605e-02 3.63247752e-01 -3.06667745e-01 -1.86652198e-01 -2.74079561e-01 6.14428818e-01 -1.79784149e-01 1.11041641e+00 -8.12009513e-01 -3.27508509e-01 -5.16999066e-01 6.44631803e-01 -3.35064083e-01 2.60798007e-01 -8.80941391e-01 2.41381507e-02 1.61172107e-01 3.06037635e-01 -1.35841090e-02 -2.02699140e-01 -7.61417389e-01 -9.79188383e-01 1.81523100e-01 -8.48441064e-01 4.38198954e-01 -6.32179379e-01 -1.39203334e+00 7.63735771e-01 -1.40642077e-01 -1.42879725e+00 3.50956231e-01 -4.36424047e-01 -7.47779548e-01 7.73098588e-01 -1.77385557e+00 -7.72161186e-01 -7.73650825e-01 7.32224762e-01 5.21883070e-01 2.90376782e-01 5.18586993e-01 7.04085350e-01 -2.58155942e-01 2.21578568e-01 2.25930214e-01 1.38551354e-01 6.05982244e-01 -1.04867029e+00 -9.23268199e-02 7.18573391e-01 -5.50252736e-01 4.23881888e-01 6.60058200e-01 -6.53618157e-01 -1.39966071e+00 -1.08377755e+00 2.94472247e-01 2.45690346e-01 4.42882687e-01 2.24449728e-02 -1.17844713e+00 3.47146094e-01 2.64892519e-01 4.16676164e-01 6.65030062e-01 -4.40582633e-01 2.29041100e-01 -1.78425431e-01 -1.31655777e+00 4.15416300e-01 8.34757984e-01 -2.23128587e-01 -4.43325728e-01 6.61209047e-01 6.29032910e-01 -8.18530500e-01 -1.12818766e+00 6.59354687e-01 2.25523412e-01 -1.10024190e+00 1.42491424e+00 -1.00322440e-01 8.78081977e-01 -4.31115255e-02 1.30334556e-01 -1.36331761e+00 -5.65057158e-01 6.30115196e-02 3.61947715e-01 6.14985466e-01 -5.12979291e-02 -4.83842760e-01 9.11168337e-01 2.56206036e-01 -6.65799618e-01 -7.07168818e-01 -1.09846210e+00 -5.73095560e-01 2.74604082e-01 -6.87394023e-01 3.10185462e-01 8.58181894e-01 -6.19747341e-01 -1.85083941e-01 -2.09707230e-01 2.53733456e-01 8.19937885e-01 -1.90431729e-01 1.35590509e-01 -8.06353092e-01 -5.12400925e-01 -4.64434057e-01 -1.15662321e-01 -7.81755865e-01 -3.10334533e-01 -7.61104524e-01 7.40541369e-02 -1.59325373e+00 2.42549539e-01 -3.54532003e-01 -3.75375152e-01 1.77063555e-01 -1.10416107e-01 5.10346293e-01 1.22537557e-02 1.76458970e-01 2.21318364e-01 5.19237757e-01 2.03667450e+00 -2.86353379e-01 1.48320183e-01 1.64630443e-01 -3.67285281e-01 8.70637655e-01 7.76160359e-01 -7.90539205e-01 -2.24777669e-01 -5.89380801e-01 1.36776984e-01 4.61946845e-01 5.90035796e-01 -1.12084866e+00 2.68851966e-01 3.85329574e-02 7.45292902e-01 -7.90117085e-01 1.74273312e-01 -1.09409404e+00 2.65115738e-01 1.01603568e+00 -1.13178290e-01 -2.08599716e-01 2.10190803e-01 4.18075621e-01 -6.81786954e-01 -4.04977560e-01 1.21917105e+00 -5.63835382e-01 -2.74927169e-01 3.76844734e-01 -5.34705102e-01 -1.75287455e-01 8.07292879e-01 -5.73615544e-02 -2.04567239e-03 -2.57210642e-01 -8.83556128e-01 -1.49153724e-01 -5.46116978e-02 -1.60723209e-01 9.66962039e-01 -1.24785566e+00 -7.79758334e-01 7.35148713e-02 -3.05579096e-01 4.01902169e-01 9.81981277e-01 1.12000895e+00 -9.12921965e-01 6.16233051e-02 -1.62321746e-01 -7.08978951e-01 -1.06706786e+00 3.86701912e-01 4.83231127e-01 -6.86879098e-01 -1.14486229e+00 7.85384417e-01 2.72118092e-01 -1.23285867e-01 -1.71957389e-01 -7.10513413e-01 9.74068195e-02 -3.89655590e-01 4.67084587e-01 2.60856569e-01 7.88333535e-01 -1.69324398e-01 -1.50758967e-01 6.72282398e-01 -1.46728367e-01 1.90657780e-01 1.85001755e+00 1.35236025e-01 -1.42624170e-01 -1.88073799e-01 1.21662593e+00 -2.28366822e-01 -9.74544585e-01 -1.33011401e-01 -3.55349481e-01 -4.35970932e-01 5.87335110e-01 -6.90049946e-01 -1.58510697e+00 9.47778523e-01 8.35526526e-01 5.76650128e-02 1.77709568e+00 -1.85411870e-01 1.34361446e+00 7.59361759e-02 7.72355348e-02 -8.97275448e-01 4.38790709e-01 1.94397047e-01 9.38915968e-01 -1.12816465e+00 3.10564578e-01 -5.75283945e-01 -1.93542451e-01 1.50699770e+00 5.12784183e-01 -3.28735828e-01 8.10169160e-01 5.39178491e-01 1.37522295e-01 -4.35910493e-01 -3.93219173e-01 9.12953913e-02 3.45267821e-03 4.38254505e-01 5.53397894e-01 -1.46628603e-01 -5.72761059e-01 5.81125498e-01 5.53000271e-02 1.34458661e-01 5.33415675e-01 1.04458475e+00 -3.36006105e-01 -1.10414064e+00 -6.75204158e-01 4.96380180e-01 -8.54399860e-01 -1.35907367e-01 3.70470971e-01 9.79034543e-01 2.37830549e-01 8.00949693e-01 -3.34168255e-01 1.33629948e-01 7.02217579e-01 -5.41273296e-01 5.27061105e-01 -3.89210761e-01 -1.08686900e+00 4.18168545e-01 -2.84578711e-01 -4.87657309e-01 -5.06727636e-01 -2.67439425e-01 -1.40343142e+00 -3.48897129e-01 -1.29576728e-01 -2.05857769e-01 8.02999794e-01 7.23209858e-01 -2.94354945e-01 1.31895447e+00 5.69033802e-01 -5.92237771e-01 -4.80163515e-01 -9.24814105e-01 -5.60351014e-01 7.32771695e-01 3.94667864e-01 -3.58346671e-01 -3.65642607e-01 -1.71265125e-01]
[13.457512855529785, -2.5393857955932617]
f07e9216-55dd-42ef-b56c-fb3abae35d80
image-restoration-from-parametric
2005.14036
null
https://arxiv.org/abs/2005.14036v2
https://arxiv.org/pdf/2005.14036v2.pdf
Image Restoration from Parametric Transformations using Generative Models
When images are statistically described by a generative model we can use this information to develop optimum techniques for various image restoration problems as inpainting, super-resolution, image coloring, generative model inversion, etc. With the help of the generative model it is possible to formulate, in a natural way, these restoration problems as Statistical estimation problems. Our approach, by combining maximum a-posteriori probability with maximum likelihood estimation, is capable of restoring images that are distorted by transformations even when the latter contain unknown parameters. The resulting optimization is completely defined with no parameters requiring tuning. This must be compared with the current state of the art which requires exact knowledge of the transformations and contains regularizer terms with weights that must be properly defined. Finally, we must mention that we extend our method to accommodate mixtures of multiple images where each image is described by its own generative model and we are able of successfully separating each participating image from a single mixture.
['Kalliopi Basioti', 'George V. Moustakides']
2020-05-27
null
null
null
null
['transparency-separation', 'blind-image-deblurring']
['computer-vision', 'computer-vision']
[ 6.36112452e-01 6.02807365e-02 2.45935544e-01 -1.35754913e-01 -7.22467721e-01 -3.66787702e-01 7.83813119e-01 -2.94193476e-01 -4.24623221e-01 9.16496038e-01 1.18914284e-01 1.39318228e-01 -2.79701054e-01 -7.88106620e-01 -6.25557125e-01 -1.07816410e+00 3.94184291e-01 7.83526003e-01 1.69817284e-01 -2.82985955e-01 2.49594882e-01 7.43622661e-01 -1.68797946e+00 -4.07956503e-02 1.09424520e+00 3.56585711e-01 4.64932889e-01 8.66952300e-01 -1.56102955e-01 5.88787138e-01 -6.10461175e-01 -1.68449849e-01 1.53535396e-01 -5.81486106e-01 -7.32479811e-01 7.84185767e-01 3.22715998e-01 -9.91986617e-02 -8.74096006e-02 1.18576658e+00 2.32978314e-01 1.43789560e-01 9.77966130e-01 -7.65993237e-01 -5.48635721e-01 1.66891515e-01 -4.18372095e-01 2.36879755e-02 2.82803774e-01 -3.19796026e-01 3.76697183e-01 -7.08253622e-01 7.45023072e-01 1.28986120e+00 2.22825721e-01 5.33005238e-01 -1.97456503e+00 -8.42054039e-02 -1.55649960e-01 2.27712736e-01 -1.33996701e+00 -6.43708944e-01 8.83506715e-01 -5.73468089e-01 5.21610975e-01 3.34690571e-01 2.88576037e-01 7.65841901e-01 1.80009920e-02 4.83537644e-01 1.29186332e+00 -8.24679554e-01 2.89215505e-01 3.71617138e-01 -1.27089024e-01 5.65085709e-01 2.31142595e-01 3.03792842e-02 -2.47624472e-01 -6.75226003e-02 8.53481174e-01 -2.38819003e-01 -5.59051514e-01 -5.06076217e-01 -1.07803297e+00 7.76451588e-01 3.43858898e-02 6.21944547e-01 -3.67140561e-01 2.23737434e-02 -6.76901564e-02 1.18732065e-01 5.51652908e-01 3.35640788e-01 1.40772089e-01 3.04297090e-01 -1.21705008e+00 1.71860546e-01 9.11256135e-01 7.73786306e-01 1.00227618e+00 3.12253654e-01 2.10177302e-01 9.10943568e-01 3.50131154e-01 5.74497402e-01 2.19641834e-01 -1.23850489e+00 -5.97405396e-02 2.34893579e-02 3.04600805e-01 -7.69355059e-01 1.47065043e-01 -4.73011881e-01 -9.77776170e-01 8.14457476e-01 3.73150080e-01 1.08802885e-01 -9.36318457e-01 1.75503874e+00 2.17101604e-01 2.36924648e-01 2.14606687e-01 5.76780021e-01 3.58624518e-01 6.93724036e-01 -3.16501707e-01 -6.60298884e-01 1.04742324e+00 -3.97599816e-01 -1.02371490e+00 -3.49296294e-02 4.31323163e-02 -1.25587630e+00 6.10597432e-01 6.74173474e-01 -1.18355465e+00 -6.28542900e-01 -1.18773901e+00 2.05727369e-02 -8.35162550e-02 2.13718459e-01 2.48116523e-01 7.02245414e-01 -1.28047478e+00 7.50047982e-01 -8.83636117e-01 -3.13125312e-01 1.85044985e-02 4.49123800e-01 -4.46990311e-01 -3.57664749e-02 -6.20173395e-01 1.26857829e+00 4.22299922e-01 2.09733576e-01 -8.62435639e-01 -1.36295140e-01 -8.14727962e-01 -1.85342520e-01 1.59753099e-01 -8.45572174e-01 7.37595081e-01 -1.09598613e+00 -1.55293310e+00 9.45945323e-01 -4.70624298e-01 -1.65777460e-01 6.48604095e-01 8.57710168e-02 -3.27072233e-01 3.61927778e-01 -1.23506822e-01 3.29536885e-01 1.24957919e+00 -1.87743700e+00 -1.92089126e-01 -1.79976314e-01 -2.15655088e-01 8.96187648e-02 1.88126475e-01 -5.16619831e-02 -3.40400755e-01 -5.11999011e-01 4.42299545e-01 -8.63910496e-01 -3.32933605e-01 -1.84466571e-01 -4.58020598e-01 3.78849924e-01 6.24915838e-01 -8.49194169e-01 6.57186449e-01 -2.05635405e+00 8.12137723e-01 2.59256303e-01 9.61385742e-02 7.28436857e-02 -2.45420393e-02 3.82873446e-01 -2.81215906e-01 -9.21700895e-02 -5.42478442e-01 -7.14999795e-01 -9.76360887e-02 6.43795073e-01 -2.59182185e-01 7.33856857e-01 1.81765303e-01 3.56698304e-01 -4.53243941e-01 -6.34501934e-01 6.84767187e-01 8.84800434e-01 -4.28257197e-01 1.75976574e-01 -1.54161230e-01 6.17076159e-01 -3.20891112e-01 1.68872610e-01 9.73277450e-01 6.84046466e-03 1.80140465e-01 -1.70808077e-01 -1.76796719e-01 -2.66047746e-01 -1.58053792e+00 1.46667826e+00 -3.33562553e-01 5.52638531e-01 3.60117257e-01 -1.20204198e+00 1.02541435e+00 4.08546031e-01 5.32952905e-01 -2.37474769e-01 5.30954963e-03 3.85732263e-01 -8.42953697e-02 -3.28919470e-01 5.41601360e-01 -5.84690630e-01 4.62545484e-01 2.62029231e-01 2.66347736e-01 -3.06603819e-01 5.09674132e-01 1.49565086e-01 7.22239017e-01 1.71604529e-01 1.42799258e-01 -3.74135703e-01 7.28106499e-01 -2.46336088e-01 2.18858942e-01 6.11089945e-01 3.76681000e-01 9.53223586e-01 3.77103984e-01 -1.71178821e-02 -1.33884263e+00 -1.08308971e+00 -4.14867729e-01 1.13022909e-01 -1.12176556e-02 7.27456138e-02 -8.67418170e-01 -1.45173535e-01 -3.80657226e-01 6.59045458e-01 -5.39437354e-01 1.32466584e-01 -6.06016397e-01 -1.17660654e+00 6.68847188e-03 -2.20673919e-01 3.11931789e-01 -9.37970579e-01 -1.52918011e-01 3.54092181e-01 -1.95683599e-01 -1.10635972e+00 -1.81027148e-02 1.22906081e-01 -1.05993724e+00 -9.57203388e-01 -9.52654123e-01 -6.09939635e-01 9.85333502e-01 1.99801028e-01 1.02001750e+00 1.16567291e-01 -3.17554235e-01 4.76582110e-01 -1.63606897e-01 3.65176164e-02 -9.63422239e-01 -3.03939790e-01 -1.69334635e-01 3.62970501e-01 -2.45699808e-01 -1.13724124e+00 -2.48551980e-01 1.74618259e-01 -1.50503063e+00 5.84098175e-02 5.48694551e-01 7.83864975e-01 7.27867126e-01 2.73533404e-01 1.23677775e-01 -9.49690104e-01 3.27374667e-01 -7.86042660e-02 -6.70402646e-01 2.73442745e-01 -5.14513910e-01 5.13199866e-01 4.96448219e-01 -4.37533140e-01 -1.35851121e+00 2.69379526e-01 -1.63824022e-01 -1.23013161e-01 -5.12217760e-01 1.56687424e-01 -5.63029110e-01 -4.26882267e-01 6.55777156e-01 5.33710003e-01 2.21459359e-01 -7.79429197e-01 5.34814417e-01 2.94908196e-01 7.80065417e-01 -5.41240036e-01 9.61800218e-01 6.06860340e-01 3.03485006e-01 -1.11467981e+00 -3.36887717e-01 -3.33928168e-01 -6.96328759e-01 -2.04129025e-01 8.29472601e-01 -6.25862539e-01 -4.01362836e-01 4.76783663e-01 -1.28851569e+00 6.32205978e-02 -4.01318222e-01 5.72553396e-01 -8.64651740e-01 7.01904714e-01 -4.78590161e-01 -9.20000315e-01 2.54770786e-01 -1.27138150e+00 8.73500526e-01 1.35467529e-01 1.67627275e-01 -1.20720148e+00 1.38308495e-01 3.80710304e-01 2.95335084e-01 2.72090673e-01 1.00322938e+00 -1.69170260e-01 -8.47705066e-01 -1.77492693e-01 -4.80327243e-03 6.86185360e-01 2.27843255e-01 1.18594259e-01 -9.41989958e-01 -2.66785562e-01 4.73152816e-01 3.18943053e-01 9.84176874e-01 6.22670650e-01 7.48736799e-01 -2.30751321e-01 -7.56884217e-02 6.49350226e-01 1.78831875e+00 -1.52621733e-03 1.14267135e+00 2.54812747e-01 5.33426523e-01 6.24326169e-01 1.67987674e-01 2.52077729e-01 -1.07631180e-02 9.81721282e-01 5.20322263e-01 -2.44409651e-01 -2.63953567e-01 2.07373008e-01 1.90868795e-01 6.44469321e-01 -4.81496334e-01 -3.70047897e-01 -5.11756659e-01 3.73127013e-01 -1.67084098e+00 -1.29232919e+00 -4.11877126e-01 2.40053058e+00 7.47896433e-01 -1.12536564e-01 -1.93683252e-01 3.75018954e-01 8.35329771e-01 1.88941017e-01 -9.37901065e-02 -1.21154681e-01 -3.65427971e-01 3.11561227e-01 4.89400029e-01 1.16642046e+00 -9.36199903e-01 4.87849116e-01 7.16029596e+00 8.31425309e-01 -7.66677380e-01 7.10547566e-02 4.53529775e-01 1.66712448e-01 -6.42568111e-01 4.26519334e-01 -5.32028198e-01 4.54402715e-01 6.52237117e-01 -3.46439183e-02 5.99898458e-01 2.24715024e-01 2.78638244e-01 -5.66524267e-01 -8.27184558e-01 8.39715302e-01 3.81824046e-01 -1.09727311e+00 1.55195802e-01 4.01307374e-01 7.79961228e-01 -4.69511777e-01 3.11508798e-03 -3.99904728e-01 1.34750262e-01 -1.00448155e+00 5.05088031e-01 9.17495370e-01 4.26732540e-01 -7.62173951e-01 4.99559969e-01 5.03394067e-01 -5.79367280e-01 4.73314524e-01 -3.08629692e-01 2.23524645e-01 5.31227827e-01 8.90190721e-01 -5.26470661e-01 7.56420970e-01 -2.40598642e-03 5.18940210e-01 -2.53280342e-01 1.16319811e+00 -1.54321298e-01 3.97458225e-01 -3.44030619e-01 5.13357580e-01 -2.42468715e-01 -8.24381113e-01 8.49066317e-01 9.23381925e-01 5.56588352e-01 8.83735344e-02 5.69420420e-02 1.02668893e+00 4.45042014e-01 1.53146893e-01 -3.68896812e-01 3.30745041e-01 -2.59373575e-01 1.10130882e+00 -9.41373706e-01 -3.80660504e-01 -1.16917059e-01 1.14064002e+00 -2.12525547e-01 6.69987679e-01 -4.04248029e-01 -1.20447367e-01 3.86545241e-01 1.89224780e-01 2.40966037e-01 -3.59545231e-01 -4.81858402e-02 -1.45810139e+00 -2.75387578e-02 -7.81711042e-01 -1.79488271e-01 -9.76220548e-01 -8.72971058e-01 8.53646696e-01 3.13921988e-01 -1.07516944e+00 -6.78730190e-01 -5.70586324e-01 -2.83479869e-01 1.21102154e+00 -1.51724899e+00 -1.11618471e+00 -5.89790978e-02 5.85945368e-01 3.22266340e-01 -1.43395156e-01 9.07124102e-01 3.25845361e-01 -3.37472975e-01 -9.43579599e-02 3.91651034e-01 -3.51288050e-01 5.71384609e-01 -1.42021000e+00 -2.72866756e-01 1.29077375e+00 2.46192440e-01 4.29964095e-01 1.51135206e+00 -4.55296844e-01 -1.04295123e+00 -4.95871931e-01 9.66133058e-01 -1.05986305e-01 3.72432351e-01 -1.36028796e-01 -1.06199312e+00 5.30281186e-01 2.72042722e-01 1.16699484e-04 4.54160869e-01 -1.53720871e-01 3.61778550e-02 4.02838886e-02 -1.18795443e+00 2.96737462e-01 5.32196939e-01 -4.19209510e-01 -4.68565166e-01 4.83522177e-01 9.35561359e-02 -2.81494439e-01 -6.42973065e-01 4.30836231e-02 1.68459043e-01 -1.02914822e+00 1.15647340e+00 -1.47993103e-01 2.97403693e-01 -6.12099767e-01 -7.68166333e-02 -1.24528360e+00 -1.50582269e-01 -6.25267386e-01 1.80815279e-01 1.30824590e+00 1.67519420e-01 -5.79070985e-01 5.69266617e-01 5.84082425e-01 -4.02060105e-03 9.95907411e-02 -9.92265284e-01 -6.82874560e-01 -2.68137485e-01 -9.61872414e-02 2.03979656e-01 7.49149144e-01 -4.26731735e-01 5.22954948e-02 -8.11632574e-01 2.46785730e-01 1.14675117e+00 1.04850091e-01 6.12156510e-01 -1.34172881e+00 -7.32063293e-01 -3.29945952e-01 -4.68111008e-01 -8.90243590e-01 2.48479411e-01 -5.99360704e-01 2.12030098e-01 -1.50741053e+00 3.13073725e-01 -2.21689269e-01 8.22647959e-02 4.36776169e-02 9.69338268e-02 4.17532325e-01 1.42013505e-01 3.09501678e-01 -9.58543792e-02 4.12730157e-01 1.28975344e+00 4.07387502e-02 6.70660706e-03 2.71796823e-01 -4.71218020e-01 7.76687264e-01 5.27344644e-01 -6.13480210e-01 -3.44256163e-01 -2.19790533e-01 7.27730319e-02 3.51268739e-01 4.68684196e-01 -8.22951555e-01 1.11504093e-01 -1.32082894e-01 4.21618462e-01 -1.96043789e-01 6.11988246e-01 -8.58187497e-01 7.31204927e-01 1.66980729e-01 -1.98942766e-01 -2.66263783e-01 -9.59660485e-02 4.66927320e-01 -4.78563279e-01 -8.28319848e-01 1.06370795e+00 -3.01056027e-01 -5.19272029e-01 -6.84157982e-02 -6.40010834e-01 -4.24266189e-01 6.69582844e-01 -3.27361435e-01 -5.29915802e-02 -5.85187018e-01 -1.28544879e+00 -3.80240768e-01 8.05323958e-01 -2.30179816e-01 5.21989226e-01 -1.16667008e+00 -8.93981934e-01 1.51726142e-01 -2.74321437e-01 -1.15084380e-01 5.73130012e-01 7.56174862e-01 -8.06636035e-01 -5.01008891e-02 -3.14927816e-01 -5.55347383e-01 -1.43693209e+00 6.07928038e-01 4.40056026e-01 -2.00250641e-01 -5.02375305e-01 2.93970436e-01 1.91539705e-01 2.63866782e-02 -1.10855997e-01 3.56719047e-02 -2.62555122e-01 -1.94755182e-01 5.20244420e-01 3.40355098e-01 -3.75306755e-02 -1.04413009e+00 -8.73309523e-02 8.21547449e-01 1.68928429e-01 -5.97249269e-01 1.36532688e+00 -2.30808631e-01 -4.08953190e-01 2.35127077e-01 1.02119100e+00 3.13411564e-01 -1.16967726e+00 -1.25458136e-01 -2.56079286e-01 -5.58451355e-01 2.34380543e-01 -5.76993644e-01 -1.04008472e+00 8.11069846e-01 5.69001257e-01 2.86688209e-01 1.45325398e+00 -1.59590706e-01 4.57768999e-02 2.82430393e-03 1.65616482e-01 -7.73673356e-01 -1.10323697e-01 5.46703376e-02 9.85499382e-01 -1.00256598e+00 2.78928757e-01 -6.64795399e-01 -1.55312553e-01 1.25793552e+00 -1.08330734e-01 -5.70192218e-01 3.96017998e-01 3.87184232e-01 -1.11193798e-01 1.50880933e-01 -5.02224684e-01 -5.03907263e-01 5.14282465e-01 7.42001116e-01 2.71687776e-01 -4.69653644e-02 -6.42662823e-01 -2.14959279e-01 1.01346679e-01 -1.41152829e-01 9.02410388e-01 6.85166240e-01 -5.79746842e-01 -1.88418961e+00 -8.26952696e-01 7.45531823e-03 -4.48064178e-01 3.48130614e-03 -6.86763693e-03 8.29653680e-01 8.92513022e-02 8.82118225e-01 -2.58460402e-01 1.33114636e-01 1.88634664e-01 1.83747243e-02 8.89495909e-01 -5.62313080e-01 -4.71979864e-02 7.76157200e-01 -9.46755484e-02 -2.51542091e-01 -8.63878667e-01 -9.12973225e-01 -8.52966726e-01 -1.33531928e-01 -3.27338040e-01 2.89522558e-01 8.97651434e-01 1.07448471e+00 -1.05620787e-01 5.29412746e-01 4.29231703e-01 -1.12661481e+00 -3.95102441e-01 -7.64518023e-01 -1.02033114e+00 2.70049423e-01 4.00360554e-01 -5.69872320e-01 -5.76368213e-01 5.83147466e-01]
[11.511175155639648, -2.46928334236145]
bb60d39d-f3e5-4146-8549-ca0efc8091eb
robust-learning-at-noisy-labeled-medical
1901.07759
null
http://arxiv.org/abs/1901.07759v2
http://arxiv.org/pdf/1901.07759v2.pdf
Robust Learning at Noisy Labeled Medical Images: Applied to Skin Lesion Classification
Deep neural networks (DNNs) have achieved great success in a wide variety of medical image analysis tasks. However, these achievements indispensably rely on the accurately-annotated datasets. If with the noisy-labeled images, the training procedure will immediately encounter difficulties, leading to a suboptimal classifier. This problem is even more crucial in the medical field, given that the annotation quality requires great expertise. In this paper, we propose an effective iterative learning framework for noisy-labeled medical image classification, to combat the lacking of high quality annotated medical data. Specifically, an online uncertainty sample mining method is proposed to eliminate the disturbance from noisy-labeled images. Next, we design a sample re-weighting strategy to preserve the usefulness of correctly-labeled hard samples. Our proposed method is validated on skin lesion classification task, and achieved very promising results.
['Cheng Xue', 'Qi Dou', 'Hao Chen', 'Pheng Ann Heng', 'Xueying Shi']
2019-01-23
null
null
null
null
['skin-lesion-classification']
['medical']
[ 3.79471272e-01 3.33018929e-01 -2.29013696e-01 -5.47089398e-01 -8.91457498e-01 1.06541999e-01 8.75876546e-02 2.70761162e-01 -5.84484279e-01 9.26931679e-01 -9.45230499e-02 -7.67805949e-02 -5.58797777e-01 -7.21170008e-01 -2.45799333e-01 -1.08016264e+00 2.68126458e-01 2.92609841e-01 -1.11264512e-01 2.77636260e-01 -2.20804326e-02 1.18590377e-01 -1.48481274e+00 2.45147452e-01 1.13977516e+00 1.26336432e+00 2.37330794e-01 1.20423540e-01 -3.01919490e-01 8.32239628e-01 -6.08941734e-01 -5.21676004e-01 -5.98464999e-03 -3.87995362e-01 -7.20646322e-01 4.65725571e-01 -2.19317839e-01 -3.30391794e-01 7.68364873e-04 1.53759074e+00 5.65042853e-01 -1.78934172e-01 5.54317653e-01 -9.94079947e-01 -3.37287575e-01 6.68368459e-01 -3.27302039e-01 -7.45545700e-02 -3.66898924e-01 -1.01449691e-01 7.65139639e-01 -8.45109820e-01 4.83456433e-01 7.75843680e-01 5.02321541e-01 6.98497951e-01 -8.02587628e-01 -4.33776230e-01 -2.38779280e-02 4.74407554e-01 -1.26825833e+00 -2.39628002e-01 8.13773572e-01 -2.95387089e-01 -4.71219122e-02 6.29359558e-02 5.46773076e-01 1.22557044e+00 1.68430254e-01 8.05117309e-01 1.02982807e+00 -2.69808829e-01 4.84824240e-01 3.02178055e-01 2.00669810e-01 3.81151199e-01 3.93966138e-01 1.65873114e-02 -2.29747683e-01 5.15897088e-02 2.96675503e-01 2.45406672e-01 -4.75518733e-01 -2.58045822e-01 -8.57009888e-01 6.01328015e-01 3.72714847e-01 5.79928100e-01 -5.30784130e-01 -3.11453819e-01 6.53817177e-01 4.79608439e-02 5.96560597e-01 3.43993515e-01 -4.82313782e-01 3.22458260e-02 -9.51040566e-01 -3.23743969e-01 4.17906821e-01 5.01470566e-01 3.49038661e-01 1.81633160e-02 -2.26566315e-01 8.67801309e-01 2.81836540e-01 1.06431708e-01 5.90534866e-01 -6.99450195e-01 1.75855339e-01 6.53760433e-01 3.33519652e-02 -1.19972515e+00 -5.26237786e-01 -9.34863329e-01 -1.53782833e+00 5.96115887e-02 5.32723486e-01 -5.74645996e-02 -9.83130932e-01 1.44936848e+00 4.54020441e-01 -4.56654243e-02 8.85068327e-02 9.93808448e-01 8.12266588e-01 3.28455865e-01 2.60682255e-01 -5.75807750e-01 1.26987910e+00 -5.35409391e-01 -1.23476243e+00 -4.99146208e-02 4.48182046e-01 -4.05788571e-01 9.59886670e-01 8.92942071e-01 -7.42750108e-01 -4.98456419e-01 -1.00225484e+00 2.18813494e-01 -2.48778556e-02 5.73954403e-01 5.12727559e-01 5.68249583e-01 -5.78223705e-01 6.30205274e-01 -6.93032920e-01 4.64568809e-02 1.02838588e+00 2.16541633e-01 -2.25212231e-01 -3.16235870e-01 -1.24669421e+00 6.63359404e-01 7.47108936e-01 5.86950958e-01 -8.88361275e-01 -4.87857640e-01 -6.88323438e-01 5.40840849e-02 7.13582754e-01 -4.38910723e-01 1.21894181e+00 -1.10700631e+00 -1.18537498e+00 6.52872741e-01 1.46566942e-01 -4.33779627e-01 7.91502655e-01 -1.27470180e-01 -4.85486805e-01 1.84499606e-01 -1.21363051e-01 4.15060848e-01 1.09814608e+00 -1.43850434e+00 -6.21183038e-01 -5.90098202e-01 -3.40436488e-01 1.72266290e-02 -8.15920174e-01 -4.09045577e-01 -2.91745067e-01 -6.75116122e-01 3.30303729e-01 -4.03914958e-01 -5.73976398e-01 4.44440031e-03 -4.89389569e-01 -1.94813430e-01 5.40950298e-01 -7.47136414e-01 1.26131415e+00 -2.19559741e+00 -8.92905071e-02 1.87945902e-01 4.66011405e-01 5.05612969e-01 1.30565763e-01 -2.74316788e-01 1.78545676e-02 1.47304133e-01 -4.69365060e-01 -1.99302852e-01 -2.59853780e-01 3.42688233e-01 -2.52606925e-02 3.21969330e-01 3.35877955e-01 5.05957127e-01 -8.75133872e-01 -8.33109260e-01 1.17856368e-01 3.20621461e-01 -8.69997889e-02 1.19978815e-01 -2.68351376e-01 5.55468500e-01 -6.26807570e-01 7.85942256e-01 7.14614213e-01 -4.37330514e-01 1.89747170e-01 -4.65105474e-01 3.58318627e-01 -3.41579795e-01 -1.06825030e+00 1.56988502e+00 -3.63306582e-01 1.33869186e-01 2.06120834e-01 -1.25830793e+00 8.59489620e-01 4.87221390e-01 7.03768671e-01 -6.20473564e-01 5.71545124e-01 4.62356299e-01 -1.01267338e-01 -1.00074470e+00 2.04020336e-01 -2.39033088e-01 1.72769308e-01 -5.97638525e-02 -2.41020489e-02 7.22920895e-02 1.28118441e-01 -4.18058671e-02 7.62341261e-01 -9.85755250e-02 3.43005329e-01 -5.71580715e-02 6.21603012e-01 9.06752050e-02 1.00924051e+00 4.52294290e-01 -6.35124564e-01 5.99838734e-01 4.37713206e-01 -5.38354337e-01 -9.89696681e-01 -6.51753187e-01 -4.31950808e-01 4.48513746e-01 1.40012100e-01 -8.29201415e-02 -8.44288945e-01 -1.01180398e+00 -3.60141337e-01 4.38026786e-01 -5.05994618e-01 -3.35328341e-01 -1.61642388e-01 -1.21105516e+00 2.96119183e-01 3.75809044e-01 6.34887695e-01 -1.01735234e+00 -3.69647831e-01 2.69326419e-01 -3.23387682e-01 -9.52374220e-01 -1.66634209e-02 2.45845437e-01 -8.92218649e-01 -1.17478693e+00 -7.32487023e-01 -7.42950618e-01 9.69725907e-01 -1.05758114e-02 8.58541012e-01 1.63254917e-01 -4.80726510e-01 -2.65100718e-01 -4.70403045e-01 -5.42485714e-01 -5.09804428e-01 6.29141778e-02 2.30017062e-02 2.37960994e-01 4.08989966e-01 -2.42603749e-01 -5.36999822e-01 3.59306157e-01 -1.21295989e+00 -1.30776644e-01 7.31219351e-01 1.37679923e+00 8.18151295e-01 6.60537183e-01 1.03917527e+00 -8.76216769e-01 5.07301986e-01 -2.68442601e-01 -5.76115131e-01 4.03648913e-01 -8.16580892e-01 -3.21913734e-02 8.57615948e-01 -3.15486282e-01 -1.23450887e+00 1.60990685e-01 -4.17111337e-01 -3.36726993e-01 -2.88735390e-01 6.04733288e-01 -3.25485229e-01 2.20749497e-01 6.89824581e-01 -2.97195520e-02 2.70698015e-02 -3.53102982e-01 -1.82182789e-02 9.54273582e-01 4.25087690e-01 -2.72751242e-01 3.69064599e-01 4.68904942e-01 3.49004194e-02 -4.91911948e-01 -1.16296828e+00 -2.95522541e-01 -2.97804415e-01 -4.08897310e-01 8.00178170e-01 -6.31664395e-01 -4.91614074e-01 6.34730279e-01 -9.34936523e-01 7.58915395e-02 -3.86105001e-01 7.82856524e-01 -1.56479165e-01 5.83686054e-01 -4.36248243e-01 -9.87164438e-01 -3.87185276e-01 -1.20211518e+00 7.08560288e-01 4.06931072e-01 1.47881821e-01 -7.01756418e-01 -3.54671866e-01 4.27710682e-01 2.04456627e-01 4.02325131e-02 8.65279317e-01 -7.04472661e-01 -4.32989210e-01 -2.44388670e-01 -3.27948719e-01 8.41767967e-01 2.40409404e-01 -1.24108516e-01 -9.73876774e-01 -1.20566733e-01 4.25733924e-01 -5.82014918e-01 9.85134840e-01 5.26881397e-01 1.73646581e+00 -5.99885061e-02 -1.70320541e-01 3.74315888e-01 1.33955026e+00 2.72585511e-01 4.31134552e-01 1.24273539e-01 4.88857955e-01 7.65841246e-01 8.90950322e-01 5.27814507e-01 1.32661052e-02 1.35458663e-01 5.96207619e-01 -9.12532732e-02 1.20619118e-01 1.08929135e-01 -1.89843535e-01 8.55410635e-01 1.07956231e-01 -4.13800240e-01 -7.70393729e-01 6.76406205e-01 -1.89805114e+00 -5.80856502e-01 -1.26430392e-02 2.10402608e+00 9.86373365e-01 2.20333755e-01 -3.79693180e-01 5.69861531e-01 7.99409389e-01 -1.61370471e-01 -6.85475290e-01 3.19177538e-01 1.25066414e-02 -1.39886849e-02 2.51997620e-01 4.18134257e-02 -1.25599170e+00 4.16486651e-01 5.27899551e+00 1.24918723e+00 -9.35123920e-01 2.12150887e-01 1.17406404e+00 1.41381845e-01 -2.41471946e-01 -5.02100825e-01 -5.15705943e-01 4.86727893e-01 4.85181093e-01 1.17922261e-01 -2.67336201e-02 1.02021849e+00 3.50880682e-01 -3.08440894e-01 -8.15595329e-01 1.14776087e+00 -4.51482944e-02 -1.11927712e+00 -2.20615610e-01 -1.68464214e-01 5.99531651e-01 -4.95061517e-01 1.88822195e-01 1.38227224e-01 -5.88036440e-02 -9.16999698e-01 2.46610940e-01 6.53085828e-01 6.43051803e-01 -9.58230853e-01 1.26925766e+00 5.83313465e-01 -4.82551336e-01 -2.47539386e-01 -4.97668594e-01 3.64436388e-01 1.27040222e-01 1.53225529e+00 -8.75442326e-01 8.51726413e-01 7.70530164e-01 4.82211471e-01 -3.50343347e-01 1.19365478e+00 -2.83141673e-01 5.52330494e-01 -4.88762669e-02 -8.45152698e-03 -6.75286651e-02 -7.77289495e-02 1.87128320e-01 6.06731534e-01 3.30225080e-01 2.03358293e-01 1.06215134e-01 6.24617696e-01 -2.38592461e-01 2.88623154e-01 -2.88792670e-01 -1.06517904e-01 1.56310961e-01 1.45669198e+00 -8.07233155e-01 -2.66490340e-01 -1.06311522e-01 6.77325368e-01 8.61561596e-02 7.61878714e-02 -6.85649157e-01 -1.52622715e-01 9.19413269e-02 -2.65226871e-01 -1.69404730e-01 1.66353613e-01 -5.88724256e-01 -1.10215938e+00 1.36475086e-01 -7.68815160e-01 4.58305448e-01 -4.55546230e-01 -1.55555701e+00 6.86797440e-01 -2.79586643e-01 -1.29638577e+00 7.74883628e-02 -5.13882339e-01 -1.11906566e-01 4.29898620e-01 -1.57160914e+00 -7.24344254e-01 -5.00127077e-01 3.64650846e-01 5.29272377e-01 -7.00881332e-02 6.51003301e-01 8.12008560e-01 -8.03462446e-01 6.18941188e-01 1.53596088e-01 1.30567998e-01 7.97420800e-01 -9.97810364e-01 -4.65165079e-01 7.93744922e-01 -7.42146596e-02 2.37501830e-01 5.23892939e-01 -6.82124972e-01 -7.32124090e-01 -1.25636804e+00 7.13992655e-01 7.58615881e-02 2.34022647e-01 1.42749362e-02 -1.08261740e+00 1.83549300e-02 -4.14093398e-02 1.98102832e-01 6.57285035e-01 -7.10790455e-02 1.70798331e-01 -3.74649376e-01 -1.44236469e+00 4.11984950e-01 8.08501542e-01 -1.53427288e-01 -4.19983447e-01 5.09646058e-01 5.87216496e-01 -3.12918603e-01 -8.67955506e-01 8.53719831e-01 2.61137366e-01 -8.33504617e-01 6.55025721e-01 -4.79912043e-01 4.08325642e-01 -2.36839086e-01 8.59544426e-02 -1.31543148e+00 7.72887189e-03 1.36453323e-02 2.25733314e-02 1.05760121e+00 3.07372600e-01 -2.74362653e-01 1.02034533e+00 4.52211380e-01 -2.04193458e-01 -9.46110308e-01 -9.16117370e-01 -4.59855974e-01 -4.24413681e-01 -4.19728369e-01 2.90344357e-01 9.19090807e-01 -7.76458159e-02 1.58477686e-02 -5.36271334e-01 2.14072004e-01 7.84563899e-01 -2.75087595e-01 1.01455346e-01 -1.40129793e+00 -1.54039338e-01 -2.48307571e-01 -3.02350849e-01 -4.06791955e-01 3.07749421e-03 -5.01845837e-01 3.43026847e-01 -1.49331939e+00 2.15459526e-01 -4.74641711e-01 -4.20419455e-01 5.20255744e-01 -3.96088004e-01 2.36036420e-01 -2.20220476e-01 2.05010757e-01 -7.15707958e-01 7.16636956e-01 1.42488837e+00 -3.61536384e-01 5.70705906e-02 2.73948848e-01 -5.26117682e-01 7.85796046e-01 9.68467057e-01 -5.11290073e-01 -5.83371043e-01 -3.20550233e-01 3.12941670e-01 -4.06344682e-02 2.71360159e-01 -1.03909242e+00 1.89835429e-01 -7.60353655e-02 4.15847152e-01 -6.52617693e-01 3.32676470e-02 -1.13164699e+00 1.90520603e-02 6.25798881e-01 -3.94745648e-01 -5.32273591e-01 -1.93483859e-01 6.47148371e-01 -4.65337515e-01 -5.13648093e-01 9.42317784e-01 -2.76184171e-01 -6.23470008e-01 4.52335209e-01 -2.10922241e-01 -1.72807857e-01 1.09053361e+00 4.12675589e-02 -1.38303069e-02 -7.90582374e-02 -1.17939329e+00 3.15554529e-01 4.94960770e-02 2.41225466e-01 7.37345636e-01 -1.20761228e+00 -5.76179624e-01 1.42183691e-01 1.74498007e-01 4.69523519e-01 7.72419512e-01 7.31592178e-01 -2.16832802e-01 4.91143465e-02 8.43305513e-02 -7.06603467e-01 -1.10561824e+00 7.03371227e-01 3.16119373e-01 -4.21723127e-01 -3.41900110e-01 6.31515980e-01 -2.37740561e-01 -2.85039842e-01 6.49574578e-01 -3.84009451e-01 -3.70699555e-01 3.20572585e-01 6.56011820e-01 3.66586000e-01 5.07270932e-01 -1.72050655e-01 -7.70725086e-02 1.72622949e-01 -1.56324133e-01 1.77686170e-01 1.27641356e+00 -1.75654367e-01 -1.35833949e-01 2.39463404e-01 8.32586944e-01 -4.39001977e-01 -1.07455373e+00 -2.89743662e-01 1.72406033e-01 -3.82824808e-01 2.81907141e-01 -8.69759142e-01 -1.26667428e+00 9.29753602e-01 8.09483111e-01 2.44770706e-01 1.37395918e+00 -2.35547021e-01 8.02911639e-01 4.87110764e-01 4.65454608e-01 -1.33708405e+00 1.90063134e-01 3.30503508e-02 3.85873437e-01 -1.64073038e+00 -2.54524667e-02 -4.22458649e-01 -6.62819982e-01 1.11412811e+00 6.24919713e-01 3.93716544e-01 7.29322314e-01 1.19191736e-01 8.76268223e-02 -1.64158225e-01 -3.22225064e-01 -4.20191921e-02 -4.48278822e-02 4.50737000e-01 6.64050952e-02 1.75224151e-02 -7.23715782e-01 1.02621078e+00 3.82432491e-01 4.67183352e-01 3.68467391e-01 6.56638265e-01 -4.51067984e-01 -1.06346262e+00 -4.58864301e-01 6.33406520e-01 -7.25655437e-01 9.37711746e-02 5.08495383e-02 4.88315672e-01 5.57931304e-01 9.49577332e-01 -2.78673708e-01 -3.87259811e-01 5.21238409e-02 -1.43240774e-02 2.76510179e-01 -4.42462474e-01 -1.91451386e-01 1.35444507e-01 2.06520520e-02 -3.23362708e-01 -7.01715469e-01 -3.76438916e-01 -1.27667916e+00 2.01691106e-01 -5.66304445e-01 2.14361399e-01 6.66480780e-01 9.14964378e-01 2.04540029e-01 7.97944665e-01 7.64289141e-01 -2.77075768e-01 -8.30064476e-01 -9.11458313e-01 -6.98718369e-01 4.01556909e-01 2.94591129e-01 -5.32433212e-01 -3.92631203e-01 -2.04020217e-02]
[14.738873481750488, -2.2300031185150146]
537d1927-576c-4f91-a3d3-7fbf96d4fed8
learning-large-neighborhood-search-for
2302.13797
null
https://arxiv.org/abs/2302.13797v1
https://arxiv.org/pdf/2302.13797v1.pdf
Learning Large Neighborhood Search for Vehicle Routing in Airport Ground Handling
Dispatching vehicle fleets to serve flights is a key task in airport ground handling (AGH). Due to the notable growth of flights, it is challenging to simultaneously schedule multiple types of operations (services) for a large number of flights, where each type of operation is performed by one specific vehicle fleet. To tackle this issue, we first represent the operation scheduling as a complex vehicle routing problem and formulate it as a mixed integer linear programming (MILP) model. Then given the graph representation of the MILP model, we propose a learning assisted large neighborhood search (LNS) method using data generated based on real scenarios, where we integrate imitation learning and graph convolutional network (GCN) to learn a destroy operator to automatically select variables, and employ an off-the-shelf solver as the repair operator to reoptimize the selected variables. Experimental results based on a real airport show that the proposed method allows for handling up to 200 flights with 10 types of operations simultaneously, and outperforms state-of-the-art methods. Moreover, the learned method performs consistently accompanying different solvers, and generalizes well on larger instances, verifying the versatility and scalability of our method.
['Zhenghua Chen', 'Jie Zhang', 'Wen Song', 'Zhiguang Cao', 'Yaoxin Wu', 'Jianan Zhou']
2023-02-27
null
null
null
null
['metaheuristic-optimization']
['methodology']
[ 3.35544534e-02 1.84255153e-01 -2.82345235e-01 -7.23278448e-02 -6.28975570e-01 -8.32195997e-01 -7.55788237e-02 1.90519080e-01 -2.70038366e-01 8.47700775e-01 -4.06121016e-01 -6.06589377e-01 -7.94042885e-01 -1.01433611e+00 -9.32453096e-01 -5.83548605e-01 -5.54228127e-01 1.13184106e+00 2.50379741e-02 -6.61594272e-01 -9.95725170e-02 7.50709832e-01 -1.38276422e+00 7.16431960e-02 1.05115068e+00 1.25093067e+00 2.81387895e-01 2.67682165e-01 2.12722588e-02 3.36457461e-01 -5.18559754e-01 -3.22821528e-01 9.05058861e-01 1.12339363e-01 -6.72322214e-01 3.95851314e-01 6.99006543e-02 1.83053911e-01 -2.55934954e-01 8.18938732e-01 1.61699384e-01 7.93293893e-01 2.09918186e-01 -1.83598459e+00 -2.62634158e-01 3.72958362e-01 -2.82259256e-01 -1.34001493e-01 3.34345222e-01 3.89372081e-01 1.19263399e+00 -3.60122919e-01 6.39877021e-01 9.01569426e-01 2.58556873e-01 1.87872112e-01 -1.00796580e+00 -5.71925163e-01 8.00142229e-01 6.80390224e-02 -1.40074277e+00 2.60261685e-01 5.91601253e-01 -1.90175623e-01 1.12849295e+00 3.12973261e-01 8.64504397e-01 5.30770779e-01 4.14738029e-01 6.59311533e-01 4.62796301e-01 1.08047530e-01 4.25576657e-01 1.58820339e-02 -2.90852129e-01 1.09581876e+00 2.62256384e-01 1.66919664e-01 -5.22351936e-02 8.22776034e-02 3.72086406e-01 2.64960110e-01 -1.21171333e-01 -5.27966022e-01 -1.18193066e+00 9.95287716e-01 6.74622953e-01 -1.39095262e-01 -5.71227372e-01 1.77812681e-01 4.82627422e-01 4.56762463e-01 3.35758716e-01 7.99679875e-01 -6.02016866e-01 5.82621805e-02 -8.84302139e-01 6.32415652e-01 1.11358213e+00 1.34722722e+00 7.42034793e-01 3.19792956e-01 -4.33453359e-02 3.77145588e-01 -1.99548259e-01 3.98479968e-01 -1.70085788e-01 -6.12125874e-01 1.05120802e+00 8.64064038e-01 3.27198088e-01 -1.18053293e+00 -7.48971760e-01 -5.78535557e-01 -7.44967878e-01 5.41351456e-03 -1.18066417e-02 -4.07021493e-01 -8.25042963e-01 1.32704663e+00 3.72409165e-01 4.96710956e-01 -1.85320541e-01 1.21155894e+00 3.49390447e-01 9.85721529e-01 -3.31670940e-01 -3.14839989e-01 1.02241278e+00 -1.44335926e+00 -5.32570302e-01 -3.43805641e-01 6.99661672e-01 -2.16904566e-01 5.04221678e-01 4.03383821e-01 -1.22685885e+00 -3.20518970e-01 -9.20534432e-01 5.58969259e-01 -6.66973233e-01 -6.29228279e-02 8.60450625e-01 1.86397284e-01 -8.71809602e-01 4.78989273e-01 -5.75051248e-01 1.17667913e-01 1.39647678e-01 9.55659628e-01 -4.83593464e-01 -3.41187239e-01 -1.33137190e+00 6.66876495e-01 4.95528668e-01 7.51259267e-01 -1.07407296e+00 -8.61305296e-01 -1.14608598e+00 3.49180192e-01 1.18091702e+00 -6.74342036e-01 8.47669542e-01 -6.97796702e-01 -1.26614678e+00 2.86483526e-01 1.79173961e-01 -3.50948125e-01 3.66926402e-01 2.95856357e-01 -6.78696215e-01 4.11245227e-02 -8.90422389e-02 3.43014687e-01 8.02912474e-01 -1.21433353e+00 -9.55870926e-01 2.30474994e-02 7.04639912e-01 3.12009212e-02 3.43996845e-02 -3.00033987e-01 -4.14424866e-01 -2.93507278e-01 -1.61461160e-01 -1.31323647e+00 -5.50536036e-01 -6.71184659e-01 -3.69823754e-01 -1.28958642e-01 4.28270280e-01 -6.06601834e-01 1.39428532e+00 -1.87555659e+00 7.80875325e-01 7.69161105e-01 4.46680672e-02 2.56423146e-01 -5.73844671e-01 7.45078385e-01 -2.00402558e-01 -1.14542753e-01 -3.16746473e-01 -2.42763847e-01 3.89595300e-01 4.40973312e-01 -2.91278720e-01 3.77840519e-01 3.04129362e-01 1.03799880e+00 -1.20389915e+00 -8.97380039e-02 2.10977376e-01 2.07471419e-02 -8.03347111e-01 2.90441841e-01 -5.47223866e-01 3.83572698e-01 -5.63897073e-01 1.09442008e+00 7.84877062e-01 -6.12442009e-03 4.02774662e-01 -4.57542874e-02 -5.60347699e-02 -3.64071727e-02 -1.61226058e+00 1.69886386e+00 -9.40420508e-01 2.03740031e-01 4.75461930e-01 -1.37462175e+00 7.29147375e-01 -9.73233506e-02 6.51206672e-01 -6.11614943e-01 1.26104593e-01 1.79033771e-01 -1.71688825e-01 -5.74477732e-01 5.58102250e-01 8.27231854e-02 -4.42205817e-01 1.58580646e-01 -1.09270342e-01 -2.99557239e-01 6.96913242e-01 -4.06462885e-02 1.23954165e+00 -1.45181909e-01 -2.20290512e-01 -4.69417721e-02 6.44460797e-01 4.81299937e-01 7.93024004e-01 3.96677613e-01 1.83583930e-01 2.97738183e-02 8.04122210e-01 -8.83393168e-01 -4.41656530e-01 -7.86042988e-01 3.83264929e-01 1.01853275e+00 4.92922753e-01 -2.44852245e-01 -6.87075496e-01 -8.63098800e-01 4.62844431e-01 7.90949643e-01 -4.12064552e-01 -1.80670559e-01 -7.15603530e-01 -6.64657891e-01 -1.21532016e-01 4.03529167e-01 1.71658784e-01 -9.17996109e-01 -3.23655546e-01 3.49052459e-01 -3.75840883e-03 -1.33196640e+00 -6.71664059e-01 4.54180576e-02 -3.45156252e-01 -1.11868548e+00 -1.47283584e-01 -1.10144341e+00 1.01225495e+00 2.80655086e-01 1.19167960e+00 4.02784616e-01 -5.65380454e-01 3.42031300e-01 -2.28312463e-01 5.31358384e-02 -1.39665440e-01 4.72470134e-01 1.92462355e-01 3.79449517e-01 -2.82490224e-01 -2.01712906e-01 -3.06894332e-01 5.46868145e-01 -9.51193750e-01 -1.58517644e-01 5.57451427e-01 9.62571383e-01 6.97465837e-01 8.99030089e-01 4.05316085e-01 -9.67371464e-01 7.20975757e-01 -7.06391931e-01 -1.45720685e+00 4.39330250e-01 -5.11592686e-01 -1.44339323e-01 9.63832915e-01 -1.65048510e-01 -5.51324546e-01 7.23299533e-02 3.34473610e-01 -6.50804937e-01 1.21873908e-01 1.01202548e+00 -1.06129602e-01 -4.23664361e-01 -3.38085622e-01 3.66349071e-02 -3.29925753e-02 8.19620937e-02 8.13648105e-02 1.36976838e-01 9.30801705e-02 -5.68902194e-01 1.17719936e+00 1.43804669e-01 6.11504734e-01 -3.00255805e-01 -8.53744864e-01 -3.50208044e-01 -5.57638526e-01 -2.95398653e-01 7.28868783e-01 -7.36491501e-01 -1.55011237e+00 -1.88718632e-01 -8.88448656e-01 -4.10939544e-01 1.42496999e-03 3.38068724e-01 -5.75541794e-01 -2.12218300e-01 -2.29989439e-01 -7.58475006e-01 1.73278227e-01 -1.53756547e+00 1.10443330e+00 -5.66529781e-02 2.74249524e-01 -1.30157816e+00 -1.67071313e-01 6.48616552e-01 2.96143591e-01 6.35263741e-01 9.46666837e-01 -4.28922474e-01 -1.12225974e+00 -7.12702453e-01 -1.67388488e-02 8.29302818e-02 1.75645817e-02 -1.40620470e-01 -1.03739746e-01 -8.72705579e-01 -4.93195087e-01 -4.55498062e-02 4.79237437e-01 2.16429174e-01 1.25258255e+00 -5.16371667e-01 -5.67958117e-01 9.26029801e-01 1.64529157e+00 1.75038159e-01 3.83706480e-01 2.76124448e-01 7.25373328e-01 8.70468915e-01 1.22313893e+00 5.35793245e-01 5.04083276e-01 6.75624371e-01 1.08690047e+00 -2.23719642e-01 6.00648582e-01 -1.35806531e-01 3.41293842e-01 6.68469191e-01 -9.75321382e-02 -6.81753278e-01 -9.76106822e-01 4.55889910e-01 -2.15643239e+00 -7.78421521e-01 2.71841794e-01 2.16594577e+00 1.30892277e-01 1.38328463e-01 -2.90972297e-03 -2.16898292e-01 6.24582112e-01 1.55140936e-01 -5.82744479e-01 -6.44086599e-01 2.45405465e-01 1.07651405e-01 1.03141153e+00 6.16046369e-01 -9.80199456e-01 1.03207147e+00 5.60543299e+00 7.21081257e-01 -8.97960484e-01 -3.01457852e-01 3.02059084e-01 -5.64365566e-01 -1.62829340e-01 -1.26467288e-01 -7.74703264e-01 4.43950891e-01 9.52380896e-01 -1.82262808e-01 1.42025578e+00 6.85194731e-01 2.31697485e-01 2.11121455e-01 -1.10556650e+00 7.22277343e-01 3.22841257e-02 -1.63376474e+00 -2.68070586e-02 2.45770607e-02 1.04798901e+00 -3.01784873e-01 1.80149451e-01 6.44790649e-01 2.35120922e-01 -1.17691422e+00 5.06211340e-01 3.24037284e-01 4.67980713e-01 -1.27563465e+00 9.71682191e-01 3.06992680e-01 -1.58785117e+00 -6.83739066e-01 -3.34042907e-01 5.91310784e-02 4.64648247e-01 2.78232336e-01 -9.09064531e-01 1.13039243e+00 4.22474414e-01 6.01041377e-01 -7.17799813e-02 8.92533183e-01 -8.78905132e-02 1.18590124e-01 -1.25677124e-01 9.64962598e-03 7.85711408e-01 -6.98712468e-01 6.04872644e-01 7.94084311e-01 2.85443097e-01 -6.14698343e-02 8.10006976e-01 8.85049582e-01 -1.36252403e-01 -1.29736975e-01 -5.83350599e-01 -3.68104428e-01 2.80525357e-01 1.62885463e+00 -7.21315682e-01 1.56839311e-01 -4.04845595e-01 7.40641534e-01 3.51326913e-01 6.41401827e-01 -1.38138235e+00 -5.67214191e-01 9.21149194e-01 1.48502216e-01 4.11891967e-01 -3.98917109e-01 3.11856180e-01 -9.24823463e-01 2.04288065e-01 -8.55715692e-01 4.45207417e-01 -5.92838645e-01 -9.96160924e-01 7.83121049e-01 -9.43556875e-02 -1.53361058e+00 -2.55022675e-01 -8.91006887e-01 -5.98207474e-01 7.81074345e-01 -2.04842186e+00 -1.02762258e+00 -3.43610287e-01 3.98776591e-01 5.04604280e-01 -3.52363467e-01 3.88462216e-01 4.76317674e-01 -9.96732533e-01 4.47397172e-01 -8.78464580e-02 -2.11266786e-01 2.06439346e-01 -9.98805285e-01 2.24116951e-01 6.43953323e-01 -3.95293057e-01 3.34656209e-01 4.20682251e-01 -6.31486952e-01 -2.28256536e+00 -1.45056343e+00 6.40215635e-01 -1.16642751e-01 9.84194398e-01 -6.58931911e-01 -4.71193671e-01 8.80306065e-01 4.05187309e-02 1.93693846e-01 2.92019874e-01 -8.38724375e-02 4.79097873e-01 -4.05759156e-01 -1.10247934e+00 5.16520262e-01 1.09602654e+00 1.12018287e-01 -2.05998167e-01 1.02926362e+00 1.15041828e+00 -9.92692888e-01 -8.19201946e-01 3.53086233e-01 -2.28063300e-01 -4.24946189e-01 9.47291851e-01 -1.04393363e+00 7.78399780e-02 -4.92893606e-01 9.54907201e-03 -1.80091286e+00 -3.94786566e-01 -1.05720973e+00 4.35591340e-02 8.62229645e-01 7.10535824e-01 -9.73695159e-01 6.44602478e-01 9.52582657e-01 -5.19489765e-01 -1.07609451e+00 -1.12099040e+00 -1.02574122e+00 -4.38245445e-01 -2.92254090e-01 1.20107889e+00 8.53165507e-01 -8.23906660e-02 -2.66686082e-01 -3.14669490e-01 9.06804264e-01 3.10753495e-01 6.37459278e-01 5.66431105e-01 -1.05779731e+00 -2.31917828e-01 -1.96331158e-01 -4.31439459e-01 -6.87517703e-01 8.95750999e-01 -1.21756768e+00 6.31189868e-02 -1.69615924e+00 -6.77648783e-01 -7.43253529e-01 -5.14059782e-01 5.72566032e-01 2.56310761e-01 -2.64611512e-01 2.32742220e-01 -4.60153997e-01 -8.31529140e-01 5.73157191e-01 1.38521302e+00 -5.70547581e-01 -3.14863473e-01 2.00587392e-01 -4.42922831e-01 3.57020140e-01 5.31788528e-01 -3.15074652e-01 -5.17032921e-01 -5.73279083e-01 7.40704000e-01 5.81035435e-01 1.95157394e-01 -8.07503462e-01 4.45113957e-01 -7.87166059e-01 -2.58356512e-01 -3.98970485e-01 3.58624905e-01 -1.32183182e+00 1.72223672e-01 5.34378231e-01 -1.61356792e-01 7.17830598e-01 5.47618389e-01 5.80738306e-01 -4.00635213e-01 -1.36209309e-01 1.52167588e-01 5.70417009e-02 -1.01564920e+00 8.13354671e-01 -1.81384563e-01 1.33495897e-01 1.56499040e+00 4.72500324e-02 -6.04825988e-02 -5.92221878e-02 -6.08973205e-01 1.16431141e+00 1.79736018e-01 5.77948987e-01 6.76716387e-01 -1.11839163e+00 -5.48575580e-01 5.30705929e-01 1.28332609e-02 2.45910734e-01 4.44645941e-01 8.37209344e-01 -7.53862321e-01 6.36994839e-01 -3.92097421e-02 -2.52483308e-01 -6.42196894e-01 1.07709885e+00 2.17179775e-01 -5.02347052e-01 -3.65555286e-01 7.62371659e-01 -1.52486473e-01 -6.98247254e-01 2.72281617e-01 -5.73533177e-01 8.00957009e-02 2.06295967e-01 2.24180028e-01 6.33797705e-01 3.33274275e-01 -2.84026057e-01 -5.36440253e-01 4.67961758e-01 1.83105946e-01 4.99989629e-01 1.48663485e+00 1.06944032e-01 -2.60687470e-01 -2.16179654e-01 9.32740152e-01 -1.46433517e-01 -1.18516898e+00 3.42195034e-01 -3.26397151e-01 -5.40949404e-01 2.06799895e-01 -6.92583859e-01 -1.77465439e+00 4.07165974e-01 -8.12543631e-02 3.43829364e-01 1.17045057e+00 -5.20476878e-01 1.12790799e+00 5.88848591e-01 7.80791521e-01 -1.37501729e+00 -3.37120086e-01 5.37051857e-01 8.00837636e-01 -1.10952795e+00 -2.07490206e-01 -6.96677089e-01 -8.70300949e-01 1.14330602e+00 6.13527477e-01 -3.27424288e-01 5.26449144e-01 1.94843650e-01 -2.78653204e-01 -1.94013715e-01 -1.07393420e+00 -1.72556490e-01 3.44596893e-01 2.99504906e-01 -3.59075606e-01 2.61207521e-01 2.05552995e-01 5.58276474e-01 -9.56092961e-03 -8.08191672e-02 4.93584037e-01 9.18974340e-01 1.12850936e-02 -8.95019054e-01 -9.44981501e-02 3.05382699e-01 3.30603719e-01 2.56959684e-02 1.27343357e-01 9.95242774e-01 1.65203378e-01 9.67190981e-01 2.14395404e-01 -4.87514257e-01 7.87981629e-01 -2.14844257e-01 2.61399835e-01 -6.49664998e-01 -9.73556280e-01 -5.58672428e-01 8.78264681e-02 -1.02402806e+00 1.09281741e-01 -3.94473463e-01 -1.45195758e+00 -4.51895565e-01 -2.32811227e-01 4.13890570e-01 6.47276163e-01 8.56505156e-01 6.08003974e-01 1.12386572e+00 1.07751632e+00 -9.20180082e-01 -5.62824667e-01 1.70575138e-02 -9.44351614e-01 1.83337212e-01 2.99801052e-01 -1.11751676e+00 -5.73246002e-01 -4.58132595e-01]
[5.15142297744751, 2.7994680404663086]
962499f2-47d6-4423-9e9e-098594c45851
segment-mask-and-predict-augmenting-chinese
null
null
https://aclanthology.org/2021.emnlp-main.158
https://aclanthology.org/2021.emnlp-main.158.pdf
Segment, Mask, and Predict: Augmenting Chinese Word Segmentation with Self-Supervision
Recent state-of-the-art (SOTA) effective neural network methods and fine-tuning methods based on pre-trained models (PTM) have been used in Chinese word segmentation (CWS), and they achieve great results. However, previous works focus on training the models with the fixed corpus at every iteration. The intermediate generated information is also valuable. Besides, the robustness of the previous neural methods is limited by the large-scale annotated data. There are a few noises in the annotated corpus. Limited efforts have been made by previous studies to deal with such problems. In this work, we propose a self-supervised CWS approach with a straightforward and effective architecture. First, we train a word segmentation model and use it to generate the segmentation results. Then, we use a revised masked language model (MLM) to evaluate the quality of the segmentation results based on the predictions of the MLM. Finally, we leverage the evaluations to aid the training of the segmenter by improved minimum risk training. Experimental results show that our approach outperforms previous methods on 9 different CWS datasets with single criterion training and multiple criteria training and achieves better robustness.
['Maosong Sun', 'Huanbo Luan', 'Ji Zhang', 'Kaiyu Huang', 'Gang Chen', 'Yuanhang Zheng', 'Yang Liu', 'Mieradilijiang Maimaiti']
null
null
null
null
emnlp-2021-11
['chinese-word-segmentation']
['natural-language-processing']
[ 4.39346313e-01 5.15659451e-02 -2.41059065e-01 -5.90518892e-01 -1.01409864e+00 -2.61323482e-01 1.25041068e-01 -1.36077553e-01 -8.65301013e-01 5.16986609e-01 1.48629509e-02 -5.42844236e-01 3.85909259e-01 -6.86699510e-01 -4.68512505e-01 -5.40801823e-01 5.00724614e-01 5.66954672e-01 7.37196326e-01 -2.35678524e-01 4.33189392e-01 -1.15626335e-01 -1.01885080e+00 3.11052740e-01 1.58393633e+00 9.44111764e-01 5.21907032e-01 5.14147103e-01 -4.27873909e-01 3.95434558e-01 -8.50322127e-01 -3.33488464e-01 1.47554189e-01 -5.54349422e-01 -9.89796579e-01 7.50235617e-02 -1.74103126e-01 -2.30416447e-01 2.37861454e-01 1.05035484e+00 6.29825175e-01 2.15122342e-01 3.00897002e-01 -5.87714255e-01 -7.18126059e-01 1.31136584e+00 -4.58487362e-01 -1.13892436e-01 -4.21240702e-02 1.43407866e-01 9.84999537e-01 -8.68907988e-01 4.78911608e-01 1.17427099e+00 6.03681326e-01 8.16363037e-01 -8.28041732e-01 -6.26629233e-01 5.94556749e-01 5.54125160e-02 -1.17915714e+00 -2.29966447e-01 7.61615098e-01 -9.02715325e-02 9.69026327e-01 7.43902177e-02 5.57787359e-01 8.38735223e-01 -3.27956051e-01 1.34433818e+00 9.95665312e-01 -8.79260778e-01 2.29855239e-01 1.08380355e-01 6.10482752e-01 4.94896770e-01 1.43286411e-03 -3.51073444e-01 -6.86439201e-02 3.34552288e-01 3.25094700e-01 -2.72948921e-01 -1.99870378e-01 2.43383065e-01 -8.83791804e-01 9.27947581e-01 2.61655420e-01 7.04131305e-01 -2.68308342e-01 -2.22281992e-01 2.61383414e-01 5.57495207e-02 6.60764277e-01 5.78663647e-01 -7.83155203e-01 -7.01064337e-03 -1.25975430e+00 -1.70410171e-01 7.19599783e-01 9.50840175e-01 5.90817809e-01 2.35811070e-01 -4.32976902e-01 1.01550102e+00 4.81750846e-01 2.50736207e-01 9.71734941e-01 -5.61878681e-01 7.82218993e-01 8.23484361e-01 -6.04826845e-02 -7.61005759e-01 -4.31322008e-01 -4.24690187e-01 -6.45773053e-01 -2.20273122e-01 2.96110779e-01 -3.91106486e-01 -1.39287412e+00 1.50734508e+00 1.55781806e-01 1.20787449e-01 4.99568991e-02 8.36860061e-01 6.91232026e-01 7.97007740e-01 1.48846015e-01 -1.14843443e-01 9.43831742e-01 -1.40156746e+00 -9.25077260e-01 -4.80047494e-01 7.90100813e-01 -8.51919651e-01 1.27827227e+00 3.99613857e-01 -1.06530285e+00 -7.45112121e-01 -1.03601146e+00 2.25635290e-01 -4.84861493e-01 4.96417969e-01 3.67357165e-01 9.14355934e-01 -8.73630285e-01 7.42782056e-01 -9.62218821e-01 -1.15077578e-01 3.87170047e-01 4.38315272e-01 3.15232098e-01 2.16804594e-02 -1.56577981e+00 8.31687987e-01 9.22350347e-01 5.97855151e-01 -6.25611126e-01 -2.82293648e-01 -7.60530829e-01 -1.45726457e-01 7.43463635e-01 -3.85331698e-02 1.44204831e+00 -1.16114569e+00 -1.78057766e+00 5.55629373e-01 8.24024603e-02 -4.27724332e-01 5.42897224e-01 -4.33308035e-01 -4.62214738e-01 -3.57566588e-02 1.14003457e-01 7.05767095e-01 5.68144917e-01 -1.21493864e+00 -6.90283954e-01 -9.37953666e-02 -1.36903793e-01 1.28190026e-01 -3.40793818e-01 2.78039753e-01 -1.05651891e+00 -7.36273110e-01 1.00357547e-01 -7.67044842e-01 -6.34777486e-01 -6.89282417e-01 -8.86522412e-01 -2.93085247e-01 4.89497662e-01 -9.57223356e-01 1.72212338e+00 -2.09785676e+00 -7.06470162e-02 3.34461331e-01 -2.23996550e-01 9.61261332e-01 -4.24281985e-01 1.53343394e-01 2.92775929e-01 5.60669005e-01 -5.96197665e-01 -6.19278848e-01 -3.78966369e-02 1.76444188e-01 -1.61517579e-02 -9.60678533e-02 5.47632873e-01 8.82668793e-01 -6.37746692e-01 -8.13745081e-01 -1.33749142e-01 1.88911911e-02 -5.15412152e-01 5.20452917e-01 -4.56960946e-01 3.78439009e-01 -6.14429295e-01 6.90938056e-01 6.57422006e-01 -1.25894323e-01 1.76715985e-01 1.19943507e-01 -1.68717608e-01 2.12809145e-01 -1.30859089e+00 1.60194254e+00 -2.67293453e-01 2.21174628e-01 -1.93194956e-01 -1.13902223e+00 9.39761817e-01 5.72841503e-02 -3.74197662e-02 -6.75113082e-01 4.28834766e-01 4.11670029e-01 1.84961528e-01 -5.91660619e-01 6.27389669e-01 2.49221861e-01 -1.21712647e-01 4.05216247e-01 7.27839172e-02 -3.94595154e-02 4.25835758e-01 1.23478714e-02 6.41202986e-01 4.49196398e-01 2.41849199e-02 3.22104767e-02 7.07851708e-01 2.41304174e-01 9.67835009e-01 9.32628512e-01 -2.33881205e-01 9.51131880e-01 2.67851144e-01 -2.09499031e-01 -8.49064469e-01 -4.73986685e-01 7.79525563e-02 1.28073430e+00 1.21192537e-01 -1.53205141e-01 -1.39311576e+00 -7.87966013e-01 -5.25290966e-01 9.44725871e-01 -5.18330693e-01 -1.20874718e-01 -8.75711679e-01 -1.14723110e+00 7.63014436e-01 9.11400259e-01 7.37315118e-01 -1.50832796e+00 -4.56429005e-01 6.00883663e-01 -2.68766582e-01 -1.22268152e+00 -6.09865367e-01 1.58467397e-01 -8.54284346e-01 -7.70216227e-01 -7.34280050e-01 -1.00798213e+00 6.91995323e-01 -2.67233215e-02 7.97096431e-01 4.00717258e-01 -5.07906079e-02 -3.30090255e-01 -6.69762909e-01 -4.53230381e-01 -5.74168146e-01 5.31793833e-01 -2.55081177e-01 1.14129797e-01 4.65660214e-01 2.25121737e-01 -3.88478279e-01 4.55483735e-01 -1.04526532e+00 2.71836847e-01 9.03911352e-01 8.82647395e-01 6.27622306e-01 -7.73578510e-02 7.52257347e-01 -1.36467433e+00 8.70056570e-01 -1.82485789e-01 -6.71846867e-01 6.15144968e-01 -8.71252894e-01 3.36163603e-02 5.34863412e-01 -6.51468992e-01 -1.43948519e+00 2.73272190e-02 -4.97826010e-01 4.97763306e-02 -2.71534771e-01 8.15747559e-01 -2.80300081e-01 2.14280069e-01 4.66121435e-01 1.38340443e-01 -3.31981719e-01 -7.37854004e-01 3.47551376e-01 9.57060397e-01 3.06160241e-01 -4.80230093e-01 5.19010723e-01 -7.07780719e-02 -9.05447125e-01 -3.69141281e-01 -1.14022911e+00 -4.44060296e-01 -1.01411414e+00 -4.55646142e-02 1.07716525e+00 -6.19016707e-01 -9.07156169e-02 8.63334179e-01 -1.11878169e+00 -7.33507991e-01 -5.58045367e-03 3.75348747e-01 8.08528811e-02 4.14945871e-01 -8.02902102e-01 -9.49839354e-01 -5.82325459e-01 -1.26931262e+00 8.07718277e-01 5.26544571e-01 -5.43404855e-02 -8.34495842e-01 -2.09465727e-01 5.43389142e-01 5.55547953e-01 1.25634195e-02 7.05073953e-01 -1.03754973e+00 -3.59319121e-01 -2.02588007e-01 -1.26906529e-01 6.17751300e-01 -1.01810880e-01 1.69142693e-01 -9.30784583e-01 4.48620543e-02 -8.56507570e-02 -3.63395602e-01 1.10214865e+00 4.18243170e-01 1.16411722e+00 -7.11354315e-02 -2.47291073e-01 3.91777635e-01 1.20375907e+00 5.17270267e-01 4.38093662e-01 4.62762028e-01 7.48833537e-01 6.31832957e-01 9.54106510e-01 1.12633765e-01 3.06264579e-01 2.46050775e-01 9.86665934e-02 -3.49836767e-01 1.09563507e-02 -8.02814588e-02 3.10707301e-01 1.28407514e+00 -6.63793311e-02 -3.12417239e-01 -1.14808667e+00 6.35222197e-01 -1.94310379e+00 -3.84576201e-01 -1.58309996e-01 1.73529983e+00 1.28525078e+00 6.74538732e-01 -1.78539470e-01 1.87725127e-01 8.90914917e-01 3.63878943e-02 -6.03481650e-01 -3.47641051e-01 -1.49587244e-01 2.72679508e-01 3.22182596e-01 4.38366503e-01 -1.26621163e+00 1.62144887e+00 6.00744200e+00 1.07955778e+00 -1.06959915e+00 1.76151291e-01 9.23693419e-01 2.45703772e-01 -2.87796617e-01 -1.51414806e-02 -1.13033831e+00 4.25280631e-01 9.46802318e-01 3.90734375e-01 2.04836071e-01 8.95733297e-01 1.36524215e-01 -2.39424720e-01 -5.01287699e-01 3.26517493e-01 1.43792510e-01 -1.06778264e+00 -8.12582150e-02 -4.26634043e-01 9.69340026e-01 3.52330692e-02 -1.12435624e-01 5.94319820e-01 4.07258272e-01 -7.89560199e-01 7.50268281e-01 3.23694468e-01 5.93901813e-01 -7.01515377e-01 1.21495867e+00 5.20350277e-01 -9.98682678e-01 -6.13227068e-03 -5.02809167e-01 1.95615202e-01 2.60106266e-01 5.28480232e-01 -8.26247334e-01 5.73485076e-01 4.92238164e-01 2.39879876e-01 -8.58742237e-01 1.00974119e+00 -6.63787305e-01 1.18579257e+00 -2.09814608e-01 -4.05251861e-01 6.75111711e-01 -3.10390890e-01 1.28407767e-02 1.51815200e+00 -9.08055529e-02 5.43045886e-02 4.06513661e-01 9.17766631e-01 -7.08600059e-02 4.89157289e-01 1.61170632e-01 -1.27706289e-01 3.30439895e-01 1.22370267e+00 -1.10678363e+00 -4.27378535e-01 -4.32860792e-01 6.94735825e-01 3.33365470e-01 4.20973718e-01 -7.22625315e-01 -5.65612912e-01 -4.40103449e-02 -3.46903473e-01 3.49151373e-01 -7.83544853e-02 -4.99899000e-01 -1.02459061e+00 -9.38185006e-02 -8.99276078e-01 4.60313052e-01 -4.87724274e-01 -1.10486972e+00 9.48569119e-01 -2.47966439e-01 -9.28481162e-01 -2.20886230e-01 -5.40531278e-01 -7.13689327e-01 8.38026285e-01 -1.81278467e+00 -1.22622108e+00 -6.67167753e-02 2.88710624e-01 9.33765888e-01 -2.49085322e-01 6.82190001e-01 1.98208570e-01 -1.14077187e+00 8.05741191e-01 6.14887616e-03 7.56183207e-01 6.34352207e-01 -1.25198102e+00 6.63875759e-01 1.23243475e+00 -6.88390955e-02 6.00141764e-01 2.58648276e-01 -9.12098289e-01 -8.09969068e-01 -1.17563784e+00 8.36704493e-01 -4.42801751e-02 4.43513751e-01 -3.40952307e-01 -1.15193951e+00 4.74077731e-01 2.26889044e-01 -2.10855633e-01 7.33530164e-01 2.62130257e-02 8.44995528e-02 1.62479103e-01 -8.91349852e-01 5.56343675e-01 7.21260369e-01 -7.41442665e-02 -8.36786866e-01 2.52228659e-02 1.18535793e+00 -6.16199434e-01 -3.81781787e-01 3.12226802e-01 2.92612880e-01 -5.39392710e-01 3.73326987e-01 -6.58821881e-01 2.51970083e-01 -2.33931124e-01 1.53845921e-01 -1.23606586e+00 -5.38974106e-02 -4.60976422e-01 4.37058419e-01 1.48097205e+00 9.54948366e-01 -4.26731557e-01 6.61131322e-01 7.96365917e-01 -3.91075701e-01 -9.10881102e-01 -6.09544277e-01 -5.89409947e-01 3.55658233e-02 -7.61469364e-01 7.62899458e-01 8.50970149e-01 -2.58359641e-01 2.33945802e-01 -2.42736846e-01 4.92440239e-02 1.91399723e-01 2.13687360e-01 3.66867542e-01 -1.09299374e+00 -1.62393183e-01 -4.99966800e-01 3.05566847e-01 -1.02933741e+00 1.49385735e-01 -5.89251339e-01 5.76688051e-01 -1.47109199e+00 3.41007635e-02 -5.46682954e-01 -4.18184251e-01 7.08419025e-01 -7.25122690e-01 4.32626568e-02 2.72227496e-01 8.54174048e-02 -8.89910340e-01 3.71332616e-01 1.19637930e+00 6.39306307e-02 -5.83554566e-01 3.46869856e-01 -6.71353579e-01 8.86542022e-01 1.02552795e+00 -5.02999604e-01 -7.19051957e-02 -7.34838486e-01 -4.46270145e-02 -2.64288038e-01 -3.38231206e-01 -8.31778765e-01 3.87323231e-01 -1.43389031e-01 2.60470271e-01 -8.09500098e-01 -9.67238545e-02 -5.23670673e-01 -4.79426593e-01 4.86621827e-01 -5.04414260e-01 -1.01624034e-01 1.04019150e-01 3.41481060e-01 -2.83922732e-01 -7.38847494e-01 8.52406561e-01 -2.60056466e-01 -8.55345130e-01 1.63362041e-01 -2.60901630e-01 4.01130259e-01 7.24549294e-01 -1.98793828e-01 1.38784185e-01 -3.89127061e-02 -5.51801920e-01 5.05497038e-01 -2.33664420e-02 4.70346987e-01 5.37649155e-01 -1.01824009e+00 -7.17151701e-01 2.16404900e-01 -7.30070323e-02 4.42262709e-01 2.38022469e-02 6.23404205e-01 -4.87771720e-01 3.53888094e-01 8.91356021e-02 -4.76530403e-01 -9.13541734e-01 3.65194857e-01 1.55283779e-01 -6.22721076e-01 -2.43697390e-01 9.14842010e-01 -3.46211374e-01 -6.77574396e-01 4.46367323e-01 -5.39767146e-01 -6.53196394e-01 2.31155992e-01 4.52425331e-01 7.99500644e-02 -1.30253425e-02 -4.39348310e-01 -1.20777950e-01 5.85069954e-01 -2.92083144e-01 -2.17102870e-01 1.48958099e+00 -2.86479462e-02 3.91253270e-02 3.51648569e-01 7.73249149e-01 -1.33127138e-01 -1.13725567e+00 -5.90081394e-01 4.22424912e-01 -1.15607753e-01 -2.83561014e-02 -9.26741004e-01 -1.25499368e+00 9.77315545e-01 3.83374661e-01 1.17631502e-01 1.18262637e+00 -3.88089031e-01 1.08523428e+00 3.03380936e-01 7.48301744e-02 -1.84581327e+00 7.62767792e-02 6.71238303e-01 4.49373424e-01 -1.55133033e+00 -3.44376981e-01 -3.88626695e-01 -1.04106021e+00 1.10138571e+00 8.72924328e-01 -1.21231861e-02 5.64365208e-01 1.42805934e-01 4.47072923e-01 3.16386074e-01 -3.45344424e-01 -3.91895205e-01 4.55074936e-01 3.89823258e-01 2.81063229e-01 -1.08707808e-02 -5.34429908e-01 1.21020222e+00 -5.57321981e-02 -7.10379034e-02 3.08130652e-01 8.75751078e-01 -5.71594894e-01 -1.26900780e+00 -2.54990041e-01 3.35956663e-01 -6.04150653e-01 -3.91517520e-01 -4.68413055e-01 7.06540227e-01 1.73922777e-01 1.17061567e+00 -2.31986418e-01 -4.31190759e-01 3.43926251e-01 2.16503531e-01 -2.06860706e-01 -8.11503768e-01 -8.84998024e-01 4.17972356e-01 8.12406391e-02 -4.26285684e-01 -4.47454572e-01 -5.94247282e-01 -1.53797412e+00 3.38060290e-01 -8.24891984e-01 1.61672086e-01 6.35464668e-01 1.21673965e+00 4.27125394e-02 5.92627883e-01 5.64893007e-01 -5.32356918e-01 -6.94661319e-01 -1.27937984e+00 -2.17039123e-01 3.89332801e-01 1.11153480e-02 -1.03047930e-01 -1.48185238e-01 -3.38557363e-02]
[10.02922534942627, 10.113456726074219]
6d748cf7-9180-4f50-b93a-d2f0a657e2f9
human-and-machine-practicable-mechanisms-for
2302.13937
null
https://arxiv.org/abs/2302.13937v2
https://arxiv.org/pdf/2302.13937v2.pdf
Human and Machine Intelligence in n-Person Games with Partial Knowledge
In this note, I introduce a new framework called n-person games with partial knowledge, in which players have only limited knowledge about the aspects of the game -- including actions, outcomes, and other players. For example, playing an actual game of chess is a game of partial knowledge. To analyze these games, I introduce a set of new concepts and mechanisms for measuring the intelligence of players, with a focus on the interplay between human- and machine-based decision-making. Specifically, I introduce two main concepts: firstly, the Game Intelligence (GI) mechanism, which quantifies a player's demonstrated intelligence in a game by considering not only the game's outcome but also the "mistakes" made during the game according to the reference machine's intelligence. Secondly, I define gaming-proofness, a practical and computational concept of strategy-proofness. The GI mechanism provides a practicable way to assess players and can potentially be applied to a wide range of games, from chess and backgammon to AI systems. To illustrate the concept, I apply the GI mechanism to a selection of top-level chess games.
['Mehmet S. Ismail']
2023-02-27
null
null
null
null
['game-of-chess']
['playing-games']
[-6.92843422e-02 2.66079694e-01 2.64873654e-01 2.55518943e-01 5.24228290e-02 -8.90777469e-01 3.43071014e-01 -2.52575781e-02 -6.56110108e-01 7.20456839e-01 -3.26033503e-01 -3.10161561e-01 -6.63176179e-01 -1.28551483e+00 -8.80021155e-02 -2.95438051e-01 1.12340907e-02 4.46973324e-01 4.42023814e-01 -9.50323164e-01 6.44880056e-01 8.96830708e-02 -1.56452882e+00 -3.84613201e-02 7.03085423e-01 7.50699401e-01 2.02821285e-01 9.85623181e-01 1.83412865e-01 1.37166345e+00 -9.16524231e-01 -1.05555761e+00 3.42503756e-01 -6.74037933e-01 -1.15491414e+00 -2.95903414e-01 -7.10352063e-01 -2.68534482e-01 -1.07175298e-01 1.19731653e+00 1.12622708e-01 2.14883059e-01 3.92054081e-01 -1.29124236e+00 8.17882568e-02 8.34857464e-01 1.19845174e-01 1.54670596e-01 7.69555330e-01 3.25193316e-01 1.22746217e+00 1.18876778e-01 5.41708171e-01 8.50775301e-01 6.01045728e-01 4.61818069e-01 -7.63174832e-01 -2.42464110e-01 -1.39150396e-01 2.70488203e-01 -1.30266714e+00 2.08947346e-01 5.29004991e-01 -7.05740094e-01 7.01636732e-01 5.69759369e-01 1.06249785e+00 2.69215554e-01 1.64239481e-01 5.94292521e-01 1.03081322e+00 -5.91615856e-01 5.73528767e-01 7.08819851e-02 4.76183183e-02 4.30349141e-01 4.04245853e-01 5.53281307e-01 -6.51706934e-01 1.10706665e-01 9.81232882e-01 -4.62177843e-01 7.67569691e-02 -3.17848980e-01 -9.99956489e-01 7.92519212e-01 -1.42109394e-01 6.28951609e-01 -4.67893332e-01 5.41138723e-02 2.16928408e-01 6.28189802e-01 -2.98335925e-02 1.17728436e+00 -2.04655424e-01 -8.39604735e-01 -7.91997194e-01 5.22366285e-01 1.20292354e+00 3.87080401e-01 6.12099051e-01 -1.68554351e-01 9.96619165e-02 7.51152113e-02 -5.51860705e-02 1.66950712e-03 4.18445826e-01 -1.23700035e+00 2.56076634e-01 1.00129521e+00 3.76886010e-01 -1.13466609e+00 -4.48575974e-01 -5.45593023e-01 -5.65102756e-01 7.06213176e-01 5.61977625e-01 -1.89572930e-01 -6.38511777e-02 1.91037083e+00 3.24962214e-02 -1.56130522e-01 1.10057868e-01 8.06799471e-01 6.70370102e-01 2.02673554e-01 -2.01290786e-01 -1.52195618e-01 9.51103091e-01 -6.21949077e-01 -2.04922915e-01 -2.30275303e-01 6.50389552e-01 -9.72854048e-02 8.60109687e-01 7.24882722e-01 -1.65183377e+00 -3.68820459e-01 -1.03584945e+00 4.30591434e-01 -3.68412852e-01 -2.58118123e-01 9.05424178e-01 1.02064502e+00 -1.22276366e+00 7.66283035e-01 -4.69076514e-01 -1.17684178e-01 -1.20919943e-01 5.36652982e-01 -3.16695720e-01 3.26043457e-01 -1.37931657e+00 1.09046102e+00 8.31375897e-01 -2.15461895e-01 -7.52328575e-01 -3.97388756e-01 -5.72443664e-01 3.85769963e-01 9.54630435e-01 -6.92906082e-01 1.38723111e+00 -9.97400105e-01 -1.61169255e+00 1.11324000e+00 3.91556114e-01 -3.09885859e-01 9.96022761e-01 2.97620952e-01 2.00273693e-02 -1.75313666e-01 2.43272305e-01 -5.02296090e-02 1.66479260e-01 -7.68042326e-01 -1.00563979e+00 -3.72201324e-01 1.19317472e+00 6.21884406e-01 1.40231848e-01 8.35291818e-02 -3.42399985e-01 -2.57188350e-01 4.78849262e-02 -6.00193620e-01 -4.50604171e-01 -5.20662785e-01 -2.88508058e-01 -8.70708600e-02 -3.99951458e-01 -2.17326179e-01 1.52427292e+00 -2.00031710e+00 4.06697720e-01 4.98956710e-01 5.89676201e-01 2.99710333e-01 -5.57994284e-02 6.22953773e-01 1.59989566e-01 2.49272555e-01 1.71741188e-01 2.28492796e-01 3.69006425e-01 8.83667395e-02 3.21569085e-01 -7.96842352e-02 -2.40323603e-01 1.10444140e+00 -8.99637520e-01 -9.54479631e-03 2.74980009e-01 -3.57962459e-01 -5.68679750e-01 1.84023455e-01 -6.13875035e-03 1.68485582e-01 -5.81425428e-01 1.88098460e-01 2.09701970e-01 -1.53471038e-01 3.68794292e-01 8.06376696e-01 -2.02262774e-01 2.41307482e-01 -1.39459336e+00 1.12774289e+00 -1.60975084e-01 3.19094360e-01 -8.90534371e-02 -9.41487074e-01 4.92629021e-01 1.20875411e-01 -3.90368216e-02 -6.93815768e-01 3.37386042e-01 7.48961046e-02 6.19122922e-01 -2.97876388e-01 6.44435048e-01 -3.57683688e-01 -5.63336730e-01 8.44675899e-01 -5.20302355e-02 -4.53871906e-01 8.90455246e-01 3.38960320e-01 1.47918022e+00 -1.91368714e-01 9.13139939e-01 -2.25241393e-01 7.67798364e-01 1.69528931e-01 5.46145797e-01 1.44548655e+00 -1.54088631e-01 8.04663971e-02 1.10250735e+00 -4.56752390e-01 -7.26010025e-01 -9.47210610e-01 7.59086132e-01 1.23476005e+00 4.06541586e-01 -4.81991708e-01 -9.79188502e-01 -9.06017646e-02 -1.66464284e-01 7.53995299e-01 -8.08142364e-01 -2.85431147e-01 -1.65168360e-01 -4.40401256e-01 6.36674345e-01 3.17830205e-01 7.31524527e-01 -1.08225358e+00 -1.11400068e+00 1.70771360e-01 -3.07969213e-01 -7.18434334e-01 1.04061991e-01 -1.66860729e-01 -5.05083025e-01 -1.36536932e+00 -1.67050242e-01 -2.60012716e-01 2.01371670e-01 -2.65494753e-02 1.44254506e+00 6.44979298e-01 -4.28176038e-02 5.94920993e-01 -5.94888151e-01 -4.46912438e-01 -6.11396372e-01 4.49985592e-03 1.70725547e-02 -5.77805936e-01 1.43315256e-01 -5.04270911e-01 -2.91597784e-01 4.84875649e-01 -6.19818389e-01 3.20925176e-01 3.95939976e-01 6.35677218e-01 9.43954661e-02 8.30085278e-01 -6.88508004e-02 -7.14275181e-01 1.09331858e+00 -1.93110585e-01 -6.59052789e-01 3.34514767e-01 -3.52207959e-01 -1.80323169e-01 4.89774197e-01 -2.13685855e-01 -6.78696930e-01 -4.90183592e-01 5.57248555e-02 4.87488896e-01 5.03654405e-02 7.66755044e-01 -4.57397223e-01 -3.42057139e-01 5.23171008e-01 5.38290858e-01 -1.04307540e-01 3.60484868e-02 1.34887576e-01 1.56290963e-01 6.28103137e-01 -7.41203189e-01 7.66617298e-01 5.77791966e-02 1.85824677e-01 -3.69731128e-01 -5.15102923e-01 -2.53584415e-01 -4.46850240e-01 -6.41868472e-01 3.98390174e-01 -4.07154977e-01 -1.78246593e+00 8.47182751e-01 -8.24682593e-01 -4.60667878e-01 -4.56641674e-01 3.13634276e-01 -9.06823456e-01 3.29369307e-01 -5.20910144e-01 -1.17942905e+00 1.25775754e-01 -1.04089546e+00 1.96804047e-01 5.88060498e-01 -4.05000240e-01 -9.75010633e-01 1.01038821e-01 4.88865465e-01 2.46009201e-01 7.15075135e-02 7.35682845e-01 -9.38722014e-01 -4.64886725e-01 -4.20403540e-01 1.96888685e-01 3.24728578e-01 -2.28509024e-01 -3.06852013e-01 -3.54942083e-01 4.24413159e-02 2.37585798e-01 -1.40096515e-01 1.05317898e-01 1.64119676e-01 5.28156042e-01 -8.75639617e-02 1.34660125e-01 9.35488865e-02 1.36422503e+00 5.20813823e-01 9.69237983e-01 7.24271715e-01 1.77714545e-02 6.24079466e-01 7.02375412e-01 6.51257634e-01 4.84856278e-01 6.39878094e-01 4.59687889e-01 3.50938797e-01 4.82574582e-01 -3.62879992e-01 2.96439920e-02 3.98834020e-01 -8.72559488e-01 -3.09030354e-01 -8.56417418e-01 1.84835911e-01 -1.84724891e+00 -1.27224302e+00 1.41319237e-03 2.37711143e+00 4.55634117e-01 4.80575114e-01 4.91077870e-01 6.97465539e-01 8.01350892e-01 -2.73756593e-01 -2.35819221e-01 -4.90120471e-01 -1.09917149e-01 4.32195038e-01 2.38524631e-01 6.11683190e-01 -4.94826019e-01 1.12096298e+00 7.21286631e+00 1.06249845e+00 -5.25538206e-01 7.64456764e-02 4.13836747e-01 -1.00876294e-01 -1.78953737e-01 9.46553499e-02 -1.29141584e-01 5.38349986e-01 4.92499441e-01 -7.12779105e-01 7.80317843e-01 6.64943039e-01 1.37232527e-01 -7.56432116e-01 -9.67957079e-01 7.00771213e-01 -2.85820991e-01 -1.25801098e+00 -1.59800708e-01 3.81757349e-01 3.43120843e-01 -8.11176240e-01 4.10310142e-02 5.10874450e-01 7.63444901e-01 -1.22604334e+00 1.03175652e+00 5.20271897e-01 5.51018655e-01 -9.60215330e-01 1.07600796e+00 8.12789202e-01 -8.97521973e-01 -1.45431131e-01 -5.12861349e-02 -1.27491832e+00 -7.37932548e-02 1.05888858e-01 -4.35775757e-01 7.30508924e-01 3.65106732e-01 -1.64467484e-01 -1.79408908e-01 9.85347390e-01 -5.32272756e-01 2.71391243e-01 -2.76542783e-01 -4.09496516e-01 2.81878352e-01 -3.94690931e-01 4.82893646e-01 4.47566599e-01 2.46410221e-01 8.87935102e-01 -1.80070311e-01 9.46386635e-01 4.47900057e-01 -1.37861907e-01 -1.20126411e-01 -2.44317904e-01 3.98076236e-01 1.01988959e+00 -8.75170767e-01 -2.02901587e-01 -3.83445993e-02 8.60521793e-01 1.07419319e-01 -1.69293880e-01 -5.75898707e-01 -3.08865279e-01 6.90245330e-01 7.75121674e-02 -1.51356533e-01 -1.34642273e-01 -4.85437751e-01 -8.36113989e-01 -9.10035595e-02 -1.00108910e+00 3.01698238e-01 -8.94980133e-01 -7.89567947e-01 3.57424229e-01 -1.24839425e-01 -8.23702157e-01 -7.01171339e-01 -8.62065554e-01 -9.68227684e-01 8.08715940e-01 -7.16405213e-01 -7.50080168e-01 -3.24863881e-01 5.37787259e-01 -9.04503241e-02 -3.97422999e-01 6.20063365e-01 -3.43323052e-01 -3.13122004e-01 5.22047281e-01 -2.24903449e-01 2.41953671e-01 -2.51111835e-01 -1.26460230e+00 4.66064304e-01 8.11699212e-01 -1.36833459e-01 4.98473674e-01 8.07653964e-01 -3.85352284e-01 -1.20206094e+00 -2.09830418e-01 7.52207100e-01 -5.20701706e-01 7.00841010e-01 3.24806720e-02 -1.41245142e-01 4.36066926e-01 -4.70792323e-01 -8.58698249e-01 7.64632404e-01 3.72024983e-01 7.56368712e-02 2.47158393e-01 -1.18835628e+00 8.10860932e-01 1.25955915e+00 -4.71460700e-01 -8.42983127e-01 -1.38929829e-01 1.54905915e-01 -5.53041756e-01 -6.06869817e-01 1.43921465e-01 6.76020801e-01 -1.76617515e+00 7.33933628e-01 -6.54575765e-01 4.37455148e-01 -3.18732291e-01 -3.04797255e-02 -1.20964837e+00 -6.84843004e-01 -8.43264580e-01 4.32592928e-01 7.52541482e-01 2.42847934e-01 -7.60051489e-01 9.86125410e-01 9.47581530e-01 2.68915087e-01 -4.50614989e-01 -9.40107048e-01 -7.95208573e-01 2.73935586e-01 -9.05918419e-01 8.21479619e-01 6.38763189e-01 9.13268864e-01 -1.18512005e-01 -1.74604356e-01 -1.12166308e-01 3.67607862e-01 -1.18921101e-01 8.71679008e-01 -1.45900214e+00 -7.57542312e-01 -8.62344980e-01 -1.17582369e+00 -6.91575289e-01 -1.51815265e-01 -5.25744081e-01 -1.91147998e-01 -1.29822671e+00 4.24325496e-01 -2.09268093e-01 -1.72451004e-01 1.78973377e-01 -3.32017168e-02 2.86877483e-01 5.24939895e-01 3.85892764e-02 -1.02731693e+00 -1.82611614e-01 1.16941416e+00 2.44992569e-01 -1.22662112e-01 3.88972998e-01 -1.08223677e+00 8.65750492e-01 7.80871809e-01 -1.54083654e-01 -3.64954859e-01 4.40248325e-02 1.13853025e+00 3.23425293e-01 3.54209542e-01 -1.20890701e+00 4.93540525e-01 -6.65994823e-01 -1.79732323e-01 1.68924689e-01 2.01466218e-01 -2.61198908e-01 5.34200847e-01 8.53743672e-01 -2.24214673e-01 -3.40464339e-02 -1.34306818e-01 6.37732223e-02 -3.23751420e-01 -6.76955581e-01 3.43125165e-01 -6.72388673e-01 -9.14727986e-01 -1.67756513e-01 -7.71186233e-01 1.70703873e-01 1.28393507e+00 -7.16946602e-01 -5.46029396e-02 -8.08268607e-01 -8.16273570e-01 1.79942384e-01 6.49286270e-01 -2.28698865e-01 2.62491196e-01 -8.03335249e-01 -6.40283704e-01 -5.50806485e-02 2.44216970e-03 -5.14103591e-01 4.28912520e-01 5.59027195e-01 -8.49785686e-01 3.69462252e-01 -4.75690454e-01 1.69876561e-01 -1.20600128e+00 3.19053560e-01 5.95024168e-01 -6.63290083e-01 -6.99853199e-03 9.05012429e-01 4.50492144e-01 -3.18740398e-01 -2.28269383e-01 8.62023532e-02 -3.10412943e-01 -2.74891943e-01 7.31842995e-01 5.48923254e-01 -9.25068259e-02 -4.05480772e-01 -3.57518733e-01 2.66774803e-01 2.43673891e-01 -5.40870786e-01 9.49651361e-01 -2.03749716e-01 -3.72025847e-01 3.79331976e-01 -3.00008301e-02 -9.91992056e-02 -5.71855247e-01 3.27482894e-02 -7.73107959e-03 -4.28975254e-01 -2.90748447e-01 -1.15565801e+00 -6.85283244e-01 5.93099415e-01 -6.84571564e-02 6.73311889e-01 1.16661298e+00 -2.57017285e-01 3.18742506e-02 4.26645756e-01 1.16399765e+00 -1.30900466e+00 -9.29234028e-02 9.27179098e-01 3.42943370e-01 -6.69074714e-01 -2.70078152e-01 -3.21146667e-01 -8.79578531e-01 9.76255774e-01 5.44240236e-01 2.56834528e-03 2.17052057e-01 2.22193196e-01 -1.86149061e-01 -1.14208333e-01 -7.36376047e-01 -6.52900934e-01 -2.03595147e-03 7.69240737e-01 9.33521390e-02 4.76008266e-01 -7.04577029e-01 1.45909321e+00 -1.05535734e+00 3.72225225e-01 1.00749421e+00 7.97390401e-01 -6.13120258e-01 -1.01310849e+00 -3.90294641e-01 3.80204231e-01 -3.52743328e-01 -7.15698078e-02 -8.35422218e-01 8.55054498e-01 4.27674919e-01 1.29616058e+00 -9.55949351e-02 -7.99855888e-01 3.29702407e-01 -3.47937733e-01 6.95593297e-01 -6.26040697e-01 -8.90003622e-01 -7.83921897e-01 1.03396781e-01 -7.06771910e-01 -1.83795065e-01 -5.11385858e-01 -8.40697050e-01 -1.12450862e+00 -2.16656297e-01 5.80409408e-01 3.88885051e-01 1.33557844e+00 -1.35909691e-01 2.97508299e-01 3.59760582e-01 -4.93299842e-01 -4.61834162e-01 -5.43089032e-01 -1.10361755e+00 8.57849792e-02 -5.10180891e-01 -5.98007679e-01 -4.75887060e-01 -6.29111230e-01]
[3.4410412311553955, 1.4797251224517822]
4855306e-bfc5-4033-944f-463413b4bcae
any-motion-detector-learning-class-agnostic
2004.11647
null
https://arxiv.org/abs/2004.11647v1
https://arxiv.org/pdf/2004.11647v1.pdf
Any Motion Detector: Learning Class-agnostic Scene Dynamics from a Sequence of LiDAR Point Clouds
Object detection and motion parameters estimation are crucial tasks for self-driving vehicle safe navigation in a complex urban environment. In this work we propose a novel real-time approach of temporal context aggregation for motion detection and motion parameters estimation based on 3D point cloud sequence. We introduce an ego-motion compensation layer to achieve real-time inference with performance comparable to a naive odometric transform of the original point cloud sequence. Not only is the proposed architecture capable of estimating the motion of common road participants like vehicles or pedestrians but also generalizes to other object categories which are not present in training data. We also conduct an in-deep analysis of different temporal context aggregation strategies such as recurrent cells and 3D convolutions. Finally, we provide comparison results of our state-of-the-art model with existing solutions on KITTI Scene Flow dataset.
['Viacheslav Murashkin', 'Artem Filatov', 'Andrey Rykov']
2020-04-24
null
null
null
null
['motion-detection']
['computer-vision']
[-2.52068967e-01 -3.65946978e-01 1.91063151e-01 -4.50363249e-01 -3.01981360e-01 -4.83788610e-01 8.05882335e-01 -3.61136526e-01 -8.05953085e-01 4.66267437e-01 1.01063594e-01 -5.16226351e-01 2.56854534e-01 -7.96230376e-01 -6.95191801e-01 -6.71902418e-01 -2.17124417e-01 5.24481356e-01 8.40904534e-01 -3.07903439e-01 3.81484449e-01 8.87900472e-01 -1.96191728e+00 1.74315959e-01 7.28014588e-01 8.13809395e-01 4.52797115e-01 1.09992027e+00 -1.15149856e-01 9.07656848e-01 -3.22093815e-01 -1.80608258e-01 2.57748634e-01 1.83354720e-01 -5.71629107e-01 -7.06611723e-02 1.09509528e+00 -6.23734355e-01 -7.12184727e-01 7.85592854e-01 3.88732374e-01 5.56181192e-01 7.19099402e-01 -1.25057459e+00 1.31904744e-02 -7.14579597e-02 -3.08483124e-01 7.27867424e-01 4.97716404e-02 6.31892860e-01 5.40392876e-01 -1.01563883e+00 8.51759076e-01 1.46169186e+00 6.03076816e-01 5.93969703e-01 -8.17791402e-01 -6.83558464e-01 5.78128874e-01 8.64917696e-01 -1.25306594e+00 -6.01039469e-01 5.56421161e-01 -6.61993384e-01 1.44158769e+00 8.72863531e-02 8.63913834e-01 1.16440260e+00 3.07071358e-01 1.04256487e+00 3.71607333e-01 2.41269767e-01 2.67240465e-01 -2.29280785e-01 -1.61371492e-02 6.89235151e-01 2.17605546e-01 3.66705894e-01 -2.92490840e-01 3.26325089e-01 7.03841448e-01 -2.16308773e-01 1.85825333e-01 -4.84857678e-01 -1.47316647e+00 6.80250108e-01 5.14142811e-01 -6.07073493e-02 -4.05223072e-01 8.14167261e-01 3.05433214e-01 -9.61676762e-02 5.11989832e-01 -3.05231690e-01 -3.72213930e-01 -1.01439282e-01 -1.07588589e+00 6.83529317e-01 4.92880255e-01 1.09685707e+00 9.92556036e-01 3.63577664e-01 -2.11970299e-01 2.17575520e-01 4.77219164e-01 7.95118988e-01 9.33171138e-02 -1.45060134e+00 6.03802025e-01 3.12129140e-01 3.42125863e-01 -8.66986096e-01 -6.24875486e-01 -4.95783895e-01 -5.43971837e-01 5.77442229e-01 5.58588624e-01 4.22440544e-02 -9.65112329e-01 1.56360173e+00 6.06406510e-01 1.06682086e+00 6.27802685e-02 1.02859950e+00 8.93403530e-01 7.08032250e-01 2.82849938e-01 3.48015994e-01 1.33461022e+00 -1.07909107e+00 -5.53004265e-01 -6.74502313e-01 7.21549571e-01 -4.06410575e-01 4.76459682e-01 -1.46214783e-01 -9.10204709e-01 -7.71684468e-01 -9.06996131e-01 -4.98204410e-01 -4.69670564e-01 7.73333535e-02 7.15526164e-01 5.91848850e-01 -1.18458748e+00 4.12117243e-01 -9.88968372e-01 -3.61157209e-01 5.51794410e-01 2.95413703e-01 -3.56866151e-01 -4.80255447e-02 -1.00366700e+00 1.30454314e+00 1.14887327e-01 3.52794975e-01 -1.02775252e+00 -8.91216695e-01 -1.10766923e+00 -1.97872922e-01 1.19502153e-02 -1.24455869e+00 1.27312410e+00 -3.86668801e-01 -1.36468601e+00 8.60494912e-01 -7.76982844e-01 -8.32770109e-01 8.25524092e-01 -3.56823385e-01 -2.60208845e-01 1.39933079e-01 2.57161200e-01 1.25014830e+00 7.46512294e-01 -9.60757077e-01 -1.27090359e+00 -1.90902844e-01 -5.71209677e-02 4.27855283e-01 5.20394921e-01 -4.27253902e-01 -4.83249038e-01 -2.68777609e-01 7.26060942e-02 -1.16187441e+00 -6.76946998e-01 4.03969169e-01 -5.38881011e-02 -1.37699097e-01 1.15501094e+00 -5.98148227e-01 6.84663057e-01 -1.82183766e+00 -1.53369427e-01 -1.36584207e-01 2.21593883e-02 3.87499899e-01 -1.13510802e-01 -1.25044256e-01 2.50415087e-01 -3.35984886e-01 -1.92325145e-01 -7.10904956e-01 -3.08721350e-03 3.59339029e-01 -3.89739633e-01 5.88564157e-01 3.96279901e-01 1.31627309e+00 -1.01915860e+00 -4.67772722e-01 9.59546626e-01 6.81039989e-01 -5.91174960e-01 -1.68723434e-01 -1.71356872e-01 6.14151895e-01 -2.51578599e-01 4.93795872e-01 9.44590330e-01 6.64200410e-02 -4.81837273e-01 -3.52190472e-02 -5.10655046e-01 4.64443892e-01 -1.24693501e+00 1.70524240e+00 -5.05349159e-01 1.37980628e+00 -2.40378708e-01 -5.87187290e-01 5.95260441e-01 -2.51072049e-02 3.24256063e-01 -6.83948159e-01 -6.63305749e-04 4.44804430e-02 -1.36760950e-01 -5.84872782e-01 1.02919745e+00 2.62651235e-01 1.85378358e-01 -1.15233772e-01 -8.10465496e-03 -1.80512354e-01 2.86432683e-01 4.71502803e-02 1.00950027e+00 3.44283789e-01 -7.40372092e-02 -2.45900810e-01 7.92702794e-01 2.41347194e-01 4.79845792e-01 8.15180540e-01 -6.04522705e-01 5.89662433e-01 4.76457588e-02 -6.21803105e-01 -1.16026759e+00 -1.18608403e+00 -6.79045767e-02 6.07710361e-01 4.97685969e-01 3.67672890e-02 -3.73583078e-01 -6.28419697e-01 1.93627194e-01 8.91798735e-01 -6.12616837e-01 2.81940904e-02 -9.96356010e-01 -4.75083113e-01 3.68866593e-01 5.89310050e-01 7.91086972e-01 -8.50794196e-01 -1.06991482e+00 5.12618363e-01 -4.63621795e-01 -1.66106355e+00 -4.39492881e-01 -2.50329912e-01 -9.30451632e-01 -8.50963652e-01 -4.62618262e-01 -7.03944862e-01 2.33775422e-01 6.93318784e-01 1.07176447e+00 -1.66074783e-01 -2.15218782e-01 1.70665920e-01 1.32149279e-01 -4.35196996e-01 -2.49074042e-01 1.51934354e-02 -8.70501325e-02 1.11031547e-01 5.10677457e-01 -4.94792581e-01 -1.27181423e+00 4.22571868e-01 -3.88297230e-01 1.85434565e-01 3.84574354e-01 2.06427425e-01 3.09554338e-01 -2.63696194e-01 1.78217337e-01 -3.82948220e-01 -2.37692803e-01 -5.04177094e-01 -8.53722274e-01 -2.86347628e-01 -5.43658324e-02 -5.04558487e-03 -4.04558629e-02 -3.87610227e-01 -1.28521121e+00 4.64378178e-01 -5.75279772e-01 -5.22721052e-01 -3.51431698e-01 -2.24464193e-01 5.00225350e-02 -2.01728284e-01 5.59762657e-01 1.94645375e-02 -1.68748289e-01 2.87949387e-02 4.93441552e-01 1.12786844e-01 8.15884054e-01 -1.47564858e-02 8.50923181e-01 1.28522289e+00 3.60371888e-01 -1.10549653e+00 -3.59141231e-01 -9.02378500e-01 -7.75827408e-01 -5.61353862e-01 1.10303116e+00 -1.39336884e+00 -1.01513004e+00 5.96284926e-01 -1.46994841e+00 -6.25155926e-01 -1.35364562e-01 7.03310072e-01 -7.83292532e-01 3.59208733e-01 -5.04748940e-01 -9.02603805e-01 -8.07308182e-02 -1.10788441e+00 1.32377589e+00 2.08690688e-01 -7.17321783e-02 -1.14030397e+00 1.31882429e-02 1.69486344e-01 3.21032435e-01 2.22891241e-01 1.80290967e-01 6.42162487e-02 -1.29683053e+00 -4.82888035e-02 -3.13167214e-01 -1.58341393e-01 -3.57072145e-01 1.07804738e-01 -1.23709714e+00 1.18395954e-01 -1.90279722e-01 5.01054645e-01 1.40976155e+00 9.08167064e-01 5.23154974e-01 2.35231631e-02 -6.79438710e-01 1.02249873e+00 1.10519242e+00 9.70376730e-02 9.71247613e-01 4.21282232e-01 9.99677420e-01 9.00367916e-01 9.34617877e-01 9.53098536e-02 9.06158268e-01 7.49103844e-01 7.50962317e-01 8.11186880e-02 -3.81806850e-01 -2.60839611e-01 3.29554230e-01 3.37663293e-01 -8.71101841e-02 -2.33804628e-01 -8.73599589e-01 9.94151056e-01 -2.11816812e+00 -1.48505199e+00 -7.96796560e-01 1.84708357e+00 -2.01680493e-02 3.04590821e-01 4.22217101e-02 -2.12011039e-01 5.67392051e-01 1.78844154e-01 -6.40270412e-01 -1.34124875e-01 -2.87037075e-01 -2.50573635e-01 9.19916928e-01 9.93177712e-01 -1.45241451e+00 1.30535817e+00 6.26554251e+00 5.27624369e-01 -1.09037733e+00 1.85313731e-01 2.54422843e-01 -1.44016147e-01 -1.59631118e-01 1.80156320e-01 -1.24761689e+00 4.10074443e-01 9.62463856e-01 2.33096451e-01 -1.12143055e-01 7.62281001e-01 6.59110069e-01 -3.56588393e-01 -9.21287715e-01 1.04594481e+00 -1.82224691e-01 -1.70318115e+00 -6.91535994e-02 1.65724427e-01 7.23334968e-01 6.11623347e-01 7.50598311e-02 2.48341441e-01 2.73488313e-01 -6.98818088e-01 1.14471722e+00 6.46091342e-01 1.84645325e-01 -6.58010900e-01 5.21426558e-01 5.07330120e-01 -1.46034241e+00 -1.74175143e-01 -3.59532654e-01 -1.56311005e-01 6.88256800e-01 4.48933363e-01 -9.74748492e-01 3.52681279e-01 7.13966906e-01 1.04562974e+00 -6.71522856e-01 1.37269557e+00 -9.89503041e-02 2.15384319e-01 -4.09154296e-01 7.82547444e-02 5.45413315e-01 -1.05153993e-01 1.00207162e+00 1.66129756e+00 4.94062483e-01 -1.59421980e-01 -2.48467967e-01 8.38955402e-01 5.13554454e-01 -4.19185460e-01 -8.89480948e-01 7.59574234e-01 3.14145893e-01 1.15632391e+00 -7.39926040e-01 -6.08430505e-01 -3.08429271e-01 8.55882823e-01 1.53208479e-01 7.07722723e-01 -8.85355055e-01 -3.75304148e-02 1.49503160e+00 2.23896503e-01 7.47480869e-01 -8.69262397e-01 -2.70427555e-01 -1.17794168e+00 -1.19208805e-01 1.77523047e-01 5.22484481e-02 -8.91300142e-01 -8.08931947e-01 2.88740396e-01 2.29230925e-01 -1.45172870e+00 -4.95967537e-01 -7.62716949e-01 -8.47517848e-01 8.51370513e-01 -2.02974319e+00 -1.14900470e+00 -4.95012790e-01 4.84160841e-01 7.23769724e-01 9.70932767e-02 1.40078291e-01 4.92066920e-01 -2.42329583e-01 4.03254963e-02 -2.85118014e-01 -6.76499754e-02 3.74908686e-01 -9.97254848e-01 1.34101963e+00 1.20377576e+00 -1.22016862e-01 1.94422960e-01 9.08268273e-01 -6.93362236e-01 -1.16677356e+00 -1.49606419e+00 1.06006539e+00 -9.03123140e-01 3.18648219e-01 -4.74337041e-01 -8.83831501e-01 7.81740606e-01 -1.30819067e-01 2.35568777e-01 -9.47507322e-02 -4.88378853e-01 -2.14796476e-02 -1.01388484e-01 -9.11210001e-01 9.06177104e-01 1.53019714e+00 -3.72913182e-01 -2.95156300e-01 1.29492814e-02 6.59496844e-01 -6.53587878e-01 -8.68325830e-02 5.25389493e-01 5.60572147e-01 -1.00872898e+00 1.49327374e+00 -3.21578830e-01 6.36978308e-03 -8.02382588e-01 -6.92220181e-02 -8.64508331e-01 -4.07362938e-01 -4.66216415e-01 -3.84373128e-01 6.50850296e-01 1.30543143e-01 -5.65257013e-01 1.20460820e+00 4.10440147e-01 -3.88605326e-01 -5.10097332e-02 -1.35890269e+00 -6.34787619e-01 -6.11609854e-02 -1.05459738e+00 4.19420004e-01 5.68829894e-01 -7.06848621e-01 2.46364713e-01 -1.67569265e-01 5.31277716e-01 8.73746157e-01 -1.87272638e-01 1.04496765e+00 -1.09455585e+00 3.36192548e-01 -5.82891762e-01 -9.85398591e-01 -1.51721084e+00 4.24149543e-01 -8.26264441e-01 9.23436582e-02 -1.74216449e+00 -5.09535730e-01 -2.20103279e-01 2.25150347e-01 -1.73944026e-01 -1.45619646e-01 3.42672467e-01 1.17350683e-01 3.80486734e-02 -5.33900082e-01 5.69261909e-01 1.30154550e+00 -2.27130264e-01 -2.66074330e-01 2.51946300e-01 1.12883009e-01 6.49680078e-01 6.03730798e-01 -3.76123667e-01 -4.52844054e-01 -4.85592932e-01 1.60384417e-01 1.02402136e-01 1.15067565e+00 -1.25998521e+00 5.85610509e-01 -1.24897450e-01 3.94400835e-01 -1.49343407e+00 7.63718665e-01 -6.79640532e-01 1.19985618e-01 6.15023136e-01 1.25430584e-01 2.19708741e-01 5.13414383e-01 6.78952396e-01 9.97241288e-02 3.59016545e-02 6.65092230e-01 -2.49710396e-01 -1.53088379e+00 4.39012140e-01 -9.28895652e-01 -1.53788894e-01 9.67513740e-01 -6.43061757e-01 -3.96360099e-01 -4.18743223e-01 -6.17703915e-01 4.96005863e-01 2.67519504e-01 7.62342989e-01 7.87375152e-01 -1.28686976e+00 -8.35628211e-01 1.56181410e-01 1.25553966e-01 2.32362315e-01 7.55906582e-01 7.86783457e-01 -9.21200395e-01 7.09098756e-01 -2.27352455e-01 -1.20385027e+00 -9.97080982e-01 5.49588978e-01 5.91891587e-01 5.34096733e-02 -8.93774569e-01 6.63492620e-01 6.28845870e-01 -4.45117533e-01 -7.46974498e-02 -7.07374454e-01 -2.99543887e-01 1.70231760e-02 5.83557785e-01 5.12998044e-01 1.54771879e-01 -1.04104435e+00 -5.77377439e-01 8.17827642e-01 2.85812974e-01 -4.06754076e-01 8.75011861e-01 -6.15032017e-01 4.73324239e-01 4.49442327e-01 1.20887434e+00 -4.16273594e-01 -1.93482125e+00 1.31144663e-02 1.89118572e-02 -5.27011633e-01 8.63115788e-02 -2.49067530e-01 -8.37589204e-01 1.11603785e+00 8.52306664e-01 -3.96275789e-01 5.09035110e-01 -1.83090925e-01 6.67232037e-01 4.94296879e-01 2.33223945e-01 -9.62698519e-01 -1.70093924e-01 9.65273798e-01 5.22534788e-01 -1.35589218e+00 -2.09764451e-01 -4.75564033e-01 -5.78120887e-01 9.52245295e-01 6.86976135e-01 -3.49999726e-01 6.32067800e-01 8.04909170e-02 1.85120448e-01 -1.77084636e-02 -1.03279746e+00 -8.48171294e-01 3.47270131e-01 9.69907343e-01 -2.15997398e-01 -2.03633785e-01 3.56703430e-01 -2.11017996e-01 -2.08815157e-01 -9.22982916e-02 4.55532610e-01 7.28128314e-01 -5.37568390e-01 -5.14337778e-01 -2.02046871e-01 -2.23782454e-02 4.83607613e-02 7.26115704e-02 2.63241351e-01 9.47773576e-01 1.84518293e-01 9.14841175e-01 5.75024843e-01 -1.83403030e-01 4.48035061e-01 -6.09929375e-02 4.30963308e-01 -2.05864757e-01 -3.72905463e-01 -2.46860281e-01 2.29325756e-01 -6.83217347e-01 -4.45728153e-01 -1.02818859e+00 -1.31558096e+00 -4.85234797e-01 8.94290060e-02 -5.39717138e-01 9.97534692e-01 9.02535498e-01 4.35120702e-01 5.20072877e-01 2.25851581e-01 -1.58990121e+00 2.14951456e-01 -5.83952785e-01 -2.54644342e-02 2.94275224e-01 8.69215131e-01 -7.53193736e-01 -2.72553891e-01 8.35811868e-02]
[8.20074462890625, -1.5871716737747192]
69900121-95ba-4691-ab60-75ed5f77334a
counterfactual-data-augmentation-via
2210.16838
null
https://arxiv.org/abs/2210.16838v1
https://arxiv.org/pdf/2210.16838v1.pdf
Counterfactual Data Augmentation via Perspective Transition for Open-Domain Dialogues
The construction of open-domain dialogue systems requires high-quality dialogue datasets. The dialogue data admits a wide variety of responses for a given dialogue history, especially responses with different semantics. However, collecting high-quality such a dataset in most scenarios is labor-intensive and time-consuming. In this paper, we propose a data augmentation method to automatically augment high-quality responses with different semantics by counterfactual inference. Specifically, given an observed dialogue, our counterfactual generation model first infers semantically different responses by replacing the observed reply perspective with substituted ones. Furthermore, our data selection method filters out detrimental augmented responses. Experimental results show that our data augmentation method can augment high-quality responses with different semantics for a given dialogue history, and can outperform competitive baselines on multiple downstream tasks.
['Jie zhou', 'Yang Feng', 'Jinchao Zhang', 'Jiao Ou']
2022-10-30
null
null
null
null
['counterfactual-inference']
['miscellaneous']
[ 3.40551525e-01 6.25698924e-01 -2.18215548e-02 -7.29217708e-01 -1.04706979e+00 -7.55123138e-01 6.86309159e-01 4.47725914e-02 -4.29003537e-01 1.20313907e+00 1.09490705e+00 -1.57210708e-01 4.59866673e-01 -8.57146740e-01 -1.63101584e-01 -1.52010769e-01 4.54832643e-01 8.50653648e-01 -1.49584934e-02 -8.03035915e-01 3.65609936e-02 -2.01579005e-01 -9.97031569e-01 7.83182025e-01 9.26696599e-01 5.76299787e-01 1.24282762e-01 7.34212756e-01 -4.48377788e-01 9.02437329e-01 -9.08744633e-01 -9.91030097e-01 1.58472359e-01 -8.43122840e-01 -1.46369708e+00 1.66586012e-01 -7.00369552e-02 -6.71729863e-01 -3.25524986e-01 7.07844079e-01 5.11523247e-01 4.14586604e-01 3.97011369e-01 -1.09010565e+00 -6.68927431e-01 1.20341933e+00 3.14577781e-02 -2.01398298e-01 8.73896480e-01 5.31764746e-01 1.41717434e+00 -7.13654637e-01 6.05741620e-01 1.61241376e+00 3.08414847e-01 1.14998424e+00 -1.46818364e+00 -5.04965127e-01 1.43065363e-01 -1.52488217e-01 -4.77216870e-01 -5.16652584e-01 9.52763081e-01 -1.12616286e-01 6.52226329e-01 4.86108691e-01 4.08130348e-01 1.59358954e+00 -6.89535975e-01 1.07015085e+00 1.17149842e+00 -2.28889853e-01 9.22646374e-02 1.80522546e-01 3.11282218e-01 2.64567792e-01 -3.45247656e-01 -3.52588147e-01 -5.88748097e-01 -5.55105329e-01 3.93076360e-01 -2.01898694e-01 -4.93502438e-01 1.02294698e-01 -1.26704383e+00 1.22675633e+00 1.90493584e-01 8.10882226e-02 -3.86065543e-01 -3.32519442e-01 6.07955635e-01 8.15894723e-01 5.91276765e-01 1.16659403e+00 -9.48673606e-01 -3.30060720e-01 -2.65339911e-01 6.95976377e-01 1.29532182e+00 9.16229129e-01 7.66257465e-01 -2.99506098e-01 -5.26806474e-01 1.43115103e+00 1.71742253e-02 2.64941216e-01 6.70801818e-01 -1.21311247e+00 7.92575717e-01 8.30290437e-01 4.78252321e-01 -6.02700293e-01 -5.50018191e-01 2.01588422e-01 -8.83315921e-01 -5.41303396e-01 1.00570738e+00 -6.95136428e-01 -5.21623120e-02 2.08366966e+00 5.50879419e-01 -2.78479993e-01 5.95316768e-01 1.05698419e+00 1.08072770e+00 5.30548215e-01 -2.63337977e-02 -4.17919636e-01 1.38104403e+00 -9.96817470e-01 -9.58756447e-01 -2.86364317e-01 1.02325070e+00 -7.06026196e-01 1.68820071e+00 3.83129790e-02 -8.97227287e-01 -2.80852377e-01 -5.61245680e-01 -3.01622212e-01 2.23451614e-01 -1.15087822e-01 6.81888461e-01 2.45114520e-01 -5.72514176e-01 3.83498847e-01 -1.23560369e-01 -1.55537292e-01 -1.08454153e-01 6.27470165e-02 -3.25197548e-01 3.69548090e-02 -1.79789388e+00 8.65233421e-01 3.27867031e-01 -3.42339933e-01 -3.76962721e-01 -7.71967888e-01 -1.05976140e+00 -7.31778592e-02 8.11056614e-01 -7.77704835e-01 2.07268190e+00 -8.65709603e-01 -2.05496049e+00 8.05447698e-01 -1.86183989e-01 -5.20532489e-01 7.20480919e-01 -1.72986910e-01 -2.02337831e-01 -5.26580736e-02 1.27847299e-01 5.24747193e-01 3.17601442e-01 -1.12153459e+00 -4.55655485e-01 -3.31482202e-01 6.43877983e-01 5.21957278e-01 -1.53548792e-01 -3.19139063e-02 6.84088171e-02 -5.75206101e-01 -1.74497217e-01 -8.66712749e-01 -4.18775856e-01 -4.38481033e-01 -5.90941429e-01 -5.99970043e-01 3.86921555e-01 -5.09356558e-01 1.07268715e+00 -1.78107452e+00 3.27549167e-02 -4.20127958e-01 4.49543148e-01 9.20967162e-02 -2.43225664e-01 5.01991749e-01 2.76369214e-01 7.73811415e-02 -4.49881881e-01 -3.24790150e-01 1.59725755e-01 2.20787644e-01 -6.37817979e-01 -1.62754610e-01 2.19298929e-01 8.83032799e-01 -1.20041275e+00 -3.14093113e-01 -7.72214755e-02 -3.39821041e-01 -7.89055943e-01 9.06080842e-01 -7.78963149e-01 7.47002721e-01 -5.02811432e-01 7.42474422e-02 3.73996019e-01 -3.43034625e-01 6.53798401e-01 -2.76496019e-02 2.08486170e-01 9.92604077e-01 -6.37071252e-01 1.90267491e+00 -8.64147365e-01 3.20100397e-01 3.98448557e-02 -6.15121722e-01 9.53778446e-01 4.71663386e-01 2.17454866e-01 -3.69665325e-01 9.29620117e-02 2.55248725e-01 1.97390631e-01 -5.69755077e-01 9.08260524e-01 -5.48320889e-01 -7.84094810e-01 1.14597881e+00 1.75332636e-01 -4.02486473e-01 1.20435320e-01 5.91883183e-01 9.51720357e-01 -3.60359043e-01 4.70062494e-01 5.56942932e-02 6.09569788e-01 7.91494995e-02 7.07016349e-01 7.25167871e-01 -2.78464437e-01 3.61796677e-01 8.58893991e-01 -2.71827042e-01 -9.88060772e-01 -6.05847239e-01 2.24869698e-01 1.47398436e+00 1.22740101e-02 -3.01518112e-01 -6.55729294e-01 -1.05868173e+00 -5.75684011e-02 8.63497734e-01 -3.54540557e-01 -1.24069192e-01 -7.12994456e-01 -6.00124896e-01 8.83148193e-01 1.26985401e-01 5.97738147e-01 -1.16256773e+00 -2.25190550e-01 4.35458899e-01 -1.13840580e+00 -1.11592090e+00 -6.32569194e-01 -2.88650990e-01 -6.26327097e-01 -1.30128133e+00 -5.07456779e-01 -4.55229580e-01 2.59332001e-01 3.02268445e-01 1.40137887e+00 1.64716974e-01 5.51247001e-01 -3.42387930e-02 -6.34420693e-01 -1.43121794e-01 -1.09430742e+00 -4.23856899e-02 1.17491037e-02 -3.19402032e-02 4.62740183e-01 -2.45826796e-01 -3.97521257e-01 2.57525027e-01 -7.10969925e-01 3.89158070e-01 2.06368536e-01 1.36662173e+00 -1.32332891e-01 -6.78885818e-01 1.17318058e+00 -1.47321296e+00 1.37466455e+00 -5.93161702e-01 -1.35260448e-01 2.04324841e-01 -3.17001253e-01 2.93107361e-01 1.12214458e+00 -3.78499269e-01 -1.68415856e+00 -2.58008540e-01 -2.65040100e-01 2.71696836e-01 -2.77424753e-01 3.99213195e-01 -4.87859815e-01 7.44154513e-01 9.03296649e-01 3.10803938e-04 2.38565117e-01 -6.16738319e-01 8.72982979e-01 9.00106251e-01 6.35998130e-01 -9.04175758e-01 4.00953591e-01 2.78407037e-01 -6.48998857e-01 -5.99066913e-01 -1.22950006e+00 -4.26369578e-01 -5.24623573e-01 -7.21866935e-02 4.97007281e-01 -6.95640445e-01 -7.48874962e-01 4.97056186e-01 -1.57457733e+00 -4.58639443e-01 -2.43702054e-01 2.91468531e-01 -5.27681172e-01 4.57853734e-01 -8.87599349e-01 -8.29165101e-01 -3.78453046e-01 -9.11681950e-01 6.80119693e-01 1.47119641e-01 -7.69121826e-01 -9.78778481e-01 1.58860043e-01 8.71391416e-01 8.24627727e-02 -9.40086842e-02 8.51218820e-01 -1.39333117e+00 -1.29662052e-01 6.40802160e-02 -6.94579631e-02 1.82655185e-01 2.66964704e-01 -3.95332962e-01 -9.62197900e-01 2.38998532e-02 2.10885599e-01 -1.11142504e+00 5.20034671e-01 -2.14866012e-01 9.05714989e-01 -1.16572249e+00 1.62602454e-01 -1.86478719e-02 6.18622184e-01 1.32325646e-02 1.44282356e-01 -8.78461376e-02 3.62660766e-01 1.22572327e+00 8.77367616e-01 7.01027989e-01 7.43064404e-01 5.02440751e-01 4.21018675e-02 7.53864199e-02 1.96210966e-01 -5.11620700e-01 2.04648897e-01 7.49384046e-01 5.67568779e-01 -4.23792183e-01 -5.74629307e-01 6.96478069e-01 -2.04285955e+00 -9.78009403e-01 -2.17035532e-01 1.97897649e+00 1.69162750e+00 -7.00643212e-02 3.93462598e-01 -8.86093974e-02 7.69990802e-01 2.86889702e-01 -6.65677071e-01 -4.25256640e-01 -2.81322122e-01 7.20038190e-02 -1.79655075e-01 6.52224541e-01 -6.73440158e-01 9.89058614e-01 5.72337627e+00 2.94307083e-01 -5.84893346e-01 1.89441696e-01 4.70908463e-01 -1.62077561e-01 -7.94274032e-01 1.29976958e-01 -3.79711598e-01 6.04303002e-01 8.09418559e-01 -4.90032017e-01 2.78649569e-01 6.30694687e-01 2.45798066e-01 1.18541345e-01 -1.24264646e+00 6.70879006e-01 -1.96653098e-01 -1.40990901e+00 1.77924782e-01 -2.38754719e-01 5.07703781e-01 -3.72163653e-01 -4.59715456e-01 6.41344845e-01 1.15484583e+00 -6.76762700e-01 2.28538752e-01 8.15477148e-02 6.32981122e-01 -5.38779736e-01 7.04325736e-01 7.30167270e-01 -4.52256858e-01 -4.13481817e-02 -1.30874947e-01 -5.03706098e-01 3.72580588e-01 4.33972061e-01 -1.23932290e+00 3.56324553e-01 -4.55850642e-03 3.03247780e-01 -6.14461116e-02 2.96721131e-01 -5.15376210e-01 5.41797459e-01 -1.04412593e-01 -4.64864016e-01 1.61643475e-01 -4.68094684e-02 5.30353725e-01 1.13456559e+00 -2.22413346e-01 6.70161843e-01 4.14130390e-01 7.85002947e-01 -6.57314181e-01 1.20599233e-01 -7.24837184e-01 -1.99383171e-03 9.73511338e-01 1.19014347e+00 1.18912812e-02 -6.43138647e-01 -3.39292198e-01 9.94259834e-01 5.26998878e-01 1.03105620e-01 -5.10363460e-01 -1.64741799e-01 1.02128541e+00 -3.50696683e-01 -5.82984090e-01 3.21720421e-01 -4.20068800e-01 -1.48297048e+00 -2.01838464e-03 -1.24536490e+00 8.63849640e-01 -5.83620071e-01 -1.69020724e+00 4.76569742e-01 -2.61941850e-01 -9.53545451e-01 -8.70809197e-01 -1.42316729e-01 -4.33598518e-01 9.54861820e-01 -1.15380371e+00 -9.03555512e-01 -1.11592233e-01 5.06523907e-01 1.14245057e+00 -1.08113788e-01 1.13073075e+00 -1.22210220e-01 -2.85710871e-01 6.23367906e-01 -2.49724820e-01 4.58005607e-01 1.11256528e+00 -1.48528731e+00 4.92538363e-01 5.52052140e-01 -1.91277772e-01 6.85841501e-01 7.89829373e-01 -5.10999560e-01 -9.90895867e-01 -8.40768516e-01 1.11332738e+00 -4.85367596e-01 7.57845044e-01 -3.19901705e-01 -1.40769351e+00 5.65534115e-01 5.04984498e-01 -6.46525681e-01 1.15028441e+00 4.60661471e-01 -5.35310447e-01 4.00993526e-01 -1.19916272e+00 9.23078895e-01 9.60815430e-01 -7.67870426e-01 -1.26531315e+00 5.15648484e-01 1.19877243e+00 -5.67219377e-01 -7.62264431e-01 2.06893012e-01 2.83804178e-01 -7.88171768e-01 6.15881860e-01 -1.36447275e+00 7.59083211e-01 1.43827587e-01 3.75289805e-02 -1.73207974e+00 1.03846624e-01 -1.08497834e+00 -9.37186629e-02 1.28598237e+00 5.61893344e-01 -4.91569221e-01 5.20602107e-01 1.26067388e+00 -1.13403741e-02 -4.32197273e-01 -6.19645894e-01 -3.97049487e-01 8.41957107e-02 -3.68001431e-01 1.03135395e+00 1.35910189e+00 6.75120652e-01 1.17979884e+00 -6.70222402e-01 -2.82614112e-01 2.18015194e-01 5.36695302e-01 1.26164281e+00 -1.17424440e+00 -4.02075946e-01 -3.55998188e-01 3.85296077e-01 -1.54349792e+00 6.25024021e-01 -7.25243449e-01 1.78604722e-01 -1.25909829e+00 1.06916070e-01 -4.14259195e-01 4.21130121e-01 4.03494269e-01 -7.55757332e-01 -2.77749300e-01 5.80731034e-02 1.87395841e-01 -5.67122459e-01 8.81504655e-01 1.32714593e+00 -3.34957838e-02 -3.72374028e-01 1.37424409e-01 -1.13089991e+00 8.05974662e-01 8.76493931e-01 -3.35435629e-01 -5.38976669e-01 -3.05298388e-01 1.01998828e-01 7.03439057e-01 9.02396590e-02 6.72511905e-02 -5.88891283e-02 -5.42322814e-01 -3.16506207e-01 -2.55920023e-01 3.31930578e-01 -2.53319651e-01 -4.48465168e-01 2.74094969e-01 -1.21214545e+00 -1.46496668e-01 -2.35828727e-01 7.47821569e-01 -2.77418315e-01 -2.67173350e-01 6.94580674e-01 -5.84348679e-01 -5.25408804e-01 -5.87044060e-02 -4.50682312e-01 7.75556624e-01 5.69227815e-01 1.90237865e-01 -5.54510117e-01 -7.76124537e-01 -5.38912892e-01 5.94036937e-01 4.36234146e-01 6.39868259e-01 5.41829288e-01 -1.24202275e+00 -1.05748105e+00 -2.28332520e-01 2.96029985e-01 2.97344923e-01 3.00289840e-01 2.82913774e-01 8.58070180e-02 1.88341528e-01 -1.63892936e-02 -8.49867016e-02 -1.30117834e+00 2.75563657e-01 1.75153345e-01 -5.84739685e-01 -4.26296383e-01 9.18436646e-01 3.20218414e-01 -1.24748492e+00 -7.13149160e-02 -1.51878044e-01 -3.83826315e-01 1.76199108e-01 5.62627196e-01 2.11741090e-01 -2.32396096e-01 -3.21593404e-01 1.63806438e-01 -5.06172478e-01 -1.07503913e-01 -5.08151591e-01 1.07193696e+00 -3.81507218e-01 -1.83700070e-01 5.76097369e-01 1.00641263e+00 8.20659548e-02 -1.17143083e+00 -9.02682662e-01 2.50188291e-01 -6.41989410e-01 -7.36660421e-01 -9.50561643e-01 -4.62774932e-01 5.40761471e-01 -5.41902900e-01 5.37414670e-01 7.90267825e-01 1.95095537e-03 1.18296623e+00 7.95877814e-01 2.85836667e-01 -1.21552551e+00 4.82611805e-01 9.16533470e-01 1.13782275e+00 -1.58437216e+00 -4.35806602e-01 -4.44861233e-01 -1.31143236e+00 9.17593300e-01 1.15265727e+00 2.19986111e-01 -4.15217727e-02 -8.64949077e-02 3.93382102e-01 -5.82805760e-02 -1.33257711e+00 -1.50146976e-01 7.40576675e-03 4.20353115e-01 7.85890996e-01 2.90419668e-01 -6.08545005e-01 1.03437757e+00 -6.64018273e-01 -4.60785091e-01 1.05968630e+00 4.55040216e-01 -2.00056747e-01 -1.24384141e+00 -9.09312814e-02 5.53781509e-01 -4.29136515e-01 -2.12845206e-01 -1.12854648e+00 4.87043470e-01 -6.62702560e-01 1.50114059e+00 -2.60897040e-01 -3.61587584e-01 4.41380590e-01 4.99517441e-01 1.64730430e-01 -1.00892591e+00 -8.68877411e-01 -3.50655019e-01 9.22697723e-01 -1.81916177e-01 -3.99697006e-01 -5.23023367e-01 -1.49412465e+00 -4.40054715e-01 -3.53992879e-01 5.79655230e-01 1.12391189e-01 1.15104151e+00 3.48018467e-01 2.30128095e-01 9.81158853e-01 -9.67424065e-02 -1.24631202e+00 -1.36878908e+00 -1.07213520e-01 1.10744917e+00 3.61078888e-01 -3.90933990e-01 -1.30451053e-01 -8.87595117e-02]
[12.695381164550781, 8.155467987060547]
a7ceddef-5569-44a6-8b62-6da137ed6876
a-survey-on-multimodal-large-language-models
2306.13549
null
https://arxiv.org/abs/2306.13549v1
https://arxiv.org/pdf/2306.13549v1.pdf
A Survey on Multimodal Large Language Models
Multimodal Large Language Model (MLLM) recently has been a new rising research hotspot, which uses powerful Large Language Models (LLMs) as a brain to perform multimodal tasks. The surprising emergent capabilities of MLLM, such as writing stories based on images and OCR-free math reasoning, are rare in traditional methods, suggesting a potential path to artificial general intelligence. In this paper, we aim to trace and summarize the recent progress of MLLM. First of all, we present the formulation of MLLM and delineate its related concepts. Then, we discuss the key techniques and applications, including Multimodal Instruction Tuning (M-IT), Multimodal In-Context Learning (M-ICL), Multimodal Chain of Thought (M-CoT), and LLM-Aided Visual Reasoning (LAVR). Finally, we discuss existing challenges and point out promising research directions. In light of the fact that the era of MLLM has only just begun, we will keep updating this survey and hope it can inspire more research. An associated GitHub link collecting the latest papers is available at https://github.com/BradyFU/Awesome-Multimodal-Large-Language-Models.
['Enhong Chen', 'Tong Xu', 'Xing Sun', 'Ke Li', 'Sirui Zhao', 'Chaoyou Fu', 'Shukang Yin']
2023-06-23
null
null
null
null
['optical-character-recognition', 'visual-reasoning', 'visual-reasoning']
['computer-vision', 'computer-vision', 'reasoning']
[-5.96845895e-02 9.67674404e-02 -2.92606384e-01 6.36399314e-02 -6.89093471e-01 -6.55045629e-01 7.77792811e-01 2.59629130e-01 -2.21057430e-01 5.74449003e-01 3.75025123e-01 -5.89565992e-01 -4.01867144e-02 -6.64829016e-01 -9.15711403e-01 -5.01126289e-01 8.37390870e-02 4.20387506e-01 -1.52082428e-01 -3.75313491e-01 6.65672898e-01 1.33016348e-01 -1.45968378e+00 7.32829452e-01 9.18117642e-01 3.46729904e-01 4.77562279e-01 1.01402581e+00 -5.54792464e-01 1.46824706e+00 -6.59178615e-01 -7.43518174e-01 -4.67631966e-01 -5.09917736e-01 -8.42122257e-01 -2.07301721e-01 5.52961648e-01 -2.29985610e-01 -4.24776286e-01 1.00920212e+00 3.90686333e-01 1.33464217e-01 5.02794802e-01 -1.31860352e+00 -1.00069559e+00 7.49119818e-01 -4.85111535e-01 -1.04913682e-01 7.06286192e-01 1.38469875e-01 6.69831157e-01 -9.95691419e-01 7.72668958e-01 1.70439065e+00 2.12985396e-01 5.87969661e-01 -8.06726277e-01 -5.11312962e-01 1.95232376e-01 5.38274288e-01 -1.24518311e+00 -2.02478543e-01 6.71762943e-01 -4.24840301e-01 1.01510525e+00 3.93451154e-01 6.83011413e-01 1.27630198e+00 5.17513812e-01 1.38877058e+00 1.23740733e+00 -8.41031969e-01 2.11203117e-02 2.79588737e-02 1.18361056e-01 1.08656144e+00 -5.00498265e-02 -2.29696766e-01 -9.99902844e-01 1.63465917e-01 7.96472549e-01 -1.00793010e-02 -1.71549153e-02 4.66754250e-02 -1.72883427e+00 7.25228906e-01 8.18719491e-02 3.60812664e-01 1.21069483e-01 2.53295749e-01 2.48003408e-01 4.77721989e-01 9.13041607e-02 2.77088016e-01 8.51554126e-02 -3.65502506e-01 -7.62758911e-01 1.20522007e-01 7.75963664e-01 8.20739865e-01 4.07236457e-01 -5.23312427e-02 4.02205735e-02 9.73364472e-01 3.24212909e-01 6.80968821e-01 4.73648190e-01 -1.13102210e+00 3.93045306e-01 5.72061300e-01 -4.22506541e-01 -9.25155282e-01 -2.83204108e-01 1.40180821e-02 -7.64203489e-01 -1.75336073e-03 3.09083641e-01 4.20593135e-02 -3.18282604e-01 1.45218444e+00 8.16841722e-02 4.10219207e-02 1.93015128e-01 7.40001917e-01 1.37850368e+00 1.04174769e+00 1.75490901e-01 4.93294261e-02 1.50881267e+00 -1.30214596e+00 -9.93316770e-01 -2.00532556e-01 6.01313293e-01 -1.03135490e+00 1.36888075e+00 5.63806772e-01 -1.37537599e+00 -4.04097259e-01 -8.31974447e-01 -3.19622010e-01 -6.74946249e-01 1.60579860e-01 8.96604836e-01 3.40221763e-01 -1.16429913e+00 6.00966327e-02 -6.61352396e-01 -7.49878287e-01 2.69499332e-01 1.33822024e-01 -2.99767733e-01 -3.13785702e-01 -1.10463023e+00 8.91640782e-01 3.06007266e-01 4.03561406e-02 -9.24459577e-01 -3.79002810e-01 -8.14627230e-01 -3.90986294e-01 5.59166014e-01 -7.28067815e-01 1.24791598e+00 -7.15499759e-01 -1.58000934e+00 1.08100295e+00 -3.32960933e-01 3.95101309e-02 3.08945626e-01 -3.28975767e-01 -3.94676298e-01 6.39620721e-01 -3.76183033e-01 1.00177896e+00 4.88203108e-01 -1.48561311e+00 -3.02165031e-01 -5.28427735e-02 2.08236635e-01 2.05729440e-01 -4.11164671e-01 1.96797684e-01 -5.40157378e-01 -7.23114908e-01 -2.67153919e-01 -7.49613404e-01 -6.89449459e-02 -2.25952163e-01 -3.36432070e-01 -3.72616261e-01 6.16281033e-01 -5.09683549e-01 1.44238377e+00 -2.03876376e+00 5.24756670e-01 -1.83040157e-01 4.50723410e-01 6.12265952e-02 -4.73758787e-01 1.22602487e+00 2.95291066e-01 6.99413717e-02 1.28488606e-02 -3.31685454e-01 2.08294094e-01 1.08457968e-01 -5.17617106e-01 1.30790710e-01 6.53331727e-02 1.63504839e+00 -9.88730967e-01 -7.60634899e-01 3.50444585e-01 5.99192441e-01 -3.02666068e-01 1.38445243e-01 -4.65479851e-01 4.75047410e-01 -2.31552958e-01 9.47401881e-01 3.64641398e-01 -5.52248955e-01 2.39936203e-01 7.78963491e-02 -3.28469306e-01 -1.66401401e-01 -6.86771929e-01 1.71239376e+00 -5.27339041e-01 1.10266459e+00 1.61647335e-01 -7.89554298e-01 7.29969621e-01 3.07299614e-01 -3.35280388e-03 -7.00066805e-01 6.52516931e-02 1.92732990e-01 -8.00747126e-02 -8.59498262e-01 4.08094883e-01 -2.10141353e-02 -1.04251526e-01 6.13525391e-01 3.61703373e-02 -3.94827247e-01 3.21890116e-01 6.87890053e-01 6.05293572e-01 2.27904826e-01 3.71505618e-01 8.25721994e-02 5.73930800e-01 -9.52254701e-03 -3.74082685e-01 7.66335964e-01 -1.11355945e-01 1.55178905e-01 5.72412729e-01 -2.61170805e-01 -6.11636817e-01 -1.11069143e+00 1.68830276e-01 1.30897760e+00 1.80212811e-01 -6.27068639e-01 -7.82971263e-01 -1.74070776e-01 -3.02739471e-01 8.10724854e-01 -3.81880105e-01 1.82922203e-02 -5.13568997e-01 -6.07381284e-01 6.65266573e-01 3.62795204e-01 4.47007895e-01 -1.30138242e+00 -4.98853356e-01 -2.84257323e-01 -4.58600849e-01 -1.18068945e+00 -3.42854448e-02 -9.51212347e-02 -8.67633343e-01 -6.50070310e-01 -8.32837880e-01 -1.17145884e+00 7.70546079e-01 2.42820621e-01 1.22545516e+00 4.30105865e-01 -3.02392036e-01 8.65327775e-01 -4.54070836e-01 -2.30470359e-01 -5.07817686e-01 -1.89943328e-01 -1.48398578e-01 -3.95002902e-01 3.22206587e-01 -4.07959491e-01 -3.29287261e-01 1.81952585e-02 -1.10630047e+00 6.96481347e-01 9.01595712e-01 6.60913765e-01 2.13385835e-01 -3.35269094e-01 6.39353693e-01 -6.91754341e-01 8.95939589e-01 -3.94057930e-01 -2.31036857e-01 5.95395267e-01 -4.62211333e-02 -1.59488186e-01 4.84116852e-01 -7.21637368e-01 -1.04303229e+00 -4.12750512e-01 -6.32955059e-02 -1.63855538e-01 -3.59506130e-01 7.11176336e-01 1.02149241e-01 -1.88226670e-01 1.34691864e-01 4.64560270e-01 -9.50196013e-02 -2.92509466e-01 6.80009842e-01 7.87732542e-01 5.92758834e-01 -9.35800433e-01 4.62044001e-01 4.28642750e-01 -2.93426178e-02 -1.17340827e+00 -6.89938188e-01 -1.40380068e-02 -6.40350580e-01 -7.15441287e-01 9.99791861e-01 -9.14072335e-01 -1.20697737e+00 3.75932187e-01 -1.18042338e+00 -8.70369494e-01 5.83178587e-02 3.61936599e-01 -6.08509123e-01 5.68182707e-01 -1.14428389e+00 -8.96865487e-01 -1.29762188e-01 -1.15574241e+00 1.05793202e+00 6.17442846e-01 -4.08677995e-01 -1.46733463e+00 -4.57682796e-02 9.74695385e-01 3.63090873e-01 4.93731610e-02 1.35109627e+00 -4.48262066e-01 -9.34193075e-01 -1.72381684e-01 -1.16571337e-01 -1.91446289e-01 -4.23999965e-01 2.31683373e-01 -8.52233410e-01 -1.76843196e-01 -2.15043172e-01 -1.00786817e+00 8.58717203e-01 1.62121624e-01 1.06239820e+00 -7.92625472e-02 -2.57661790e-01 3.18299323e-01 1.25444257e+00 2.87860334e-01 5.65843761e-01 2.91920573e-01 7.86602914e-01 6.82616711e-01 4.74767387e-01 1.91509470e-01 8.96400034e-01 6.30435199e-02 3.30384701e-01 -1.66580512e-03 -3.28780651e-01 -4.85158235e-01 7.77361393e-01 1.68329549e+00 -2.03391790e-01 -4.31140184e-01 -1.13305604e+00 1.57390684e-01 -1.83223689e+00 -9.87046182e-01 -1.48045972e-01 1.94086492e+00 7.08702743e-01 -1.67675465e-01 -1.82150170e-01 -3.51233274e-01 4.00409341e-01 2.23432004e-01 -2.65594244e-01 -5.98825514e-01 -4.63373274e-01 -2.33024303e-02 -2.52640307e-01 8.38297725e-01 -7.09526896e-01 1.19821572e+00 6.56745243e+00 1.12722182e+00 -1.03905272e+00 1.87676847e-01 6.60197020e-01 -2.18663946e-01 -5.41497350e-01 -4.95163091e-02 -6.61323845e-01 1.26224875e-01 1.00930595e+00 8.35073814e-02 7.59254098e-01 2.83421040e-01 1.89230487e-01 -5.39123714e-01 -9.76302981e-01 1.25094306e+00 5.84725857e-01 -1.45683610e+00 5.26435018e-01 -6.59628585e-02 8.39339852e-01 -1.07040063e-01 2.86620408e-01 5.85358799e-01 1.64282188e-01 -1.18935692e+00 8.08405101e-01 7.45549381e-01 7.51513064e-01 -7.52029240e-01 3.53417069e-01 6.60817087e-01 -1.24085963e+00 -1.68506652e-01 -1.25940636e-01 -1.88966036e-01 1.29250720e-01 5.04810698e-02 -1.05518252e-01 4.65276331e-01 3.75973582e-01 7.69302726e-01 -7.93196142e-01 7.13427961e-01 -3.09355527e-01 5.17347276e-01 1.75057203e-01 -5.27104735e-01 1.73132956e-01 -1.91724613e-01 5.22772789e-01 1.50691164e+00 1.37379155e-01 2.32959360e-01 2.76177794e-01 8.88609350e-01 -4.77148704e-02 3.21370125e-01 -6.28095925e-01 -8.06324303e-01 2.61024535e-01 1.28552425e+00 -8.73390853e-01 -4.95840222e-01 -7.54721522e-01 1.02079213e+00 5.12956023e-01 5.20106018e-01 -6.57045543e-01 -3.34204346e-01 2.80128252e-02 -2.65538841e-01 -2.66183048e-01 -6.40554667e-01 -2.41588503e-01 -1.49254990e+00 -3.12919021e-01 -1.19827366e+00 2.53597021e-01 -1.22391129e+00 -1.21172464e+00 3.11839342e-01 1.37442216e-01 -9.42875028e-01 7.30232596e-02 -8.35285068e-01 -6.28575146e-01 3.40258002e-01 -1.28703701e+00 -1.55168390e+00 -1.55110672e-01 3.91552895e-01 1.08794129e+00 -3.34555119e-01 8.06181550e-01 2.19750002e-01 -5.97911775e-01 4.06717509e-01 1.42883316e-01 2.82851178e-02 8.01726937e-01 -9.47210789e-01 -9.43326503e-02 4.36876088e-01 1.42817825e-01 9.04349566e-01 5.24126232e-01 -7.37999320e-01 -2.04491448e+00 -4.61163104e-01 1.08019376e+00 -7.21267819e-01 1.04350185e+00 -5.37948191e-01 -7.17741728e-01 8.06020319e-01 8.38172376e-01 -7.03026116e-01 9.62490559e-01 -2.78978974e-01 -1.52461171e-01 3.69601965e-01 -6.18787944e-01 1.15103149e+00 7.05228031e-01 -7.15805113e-01 -4.30348098e-01 5.47351480e-01 7.55325973e-01 -2.46516839e-01 -7.63443351e-01 -2.19332613e-02 6.34904623e-01 -9.16168272e-01 1.02955163e+00 -4.62327361e-01 1.01776981e+00 -1.46492317e-01 -3.99896026e-01 -9.42440510e-01 8.00407305e-02 -8.18219423e-01 -3.23674649e-01 1.16450942e+00 9.15497467e-02 -4.93128449e-01 2.00209334e-01 2.67085075e-01 -6.75520673e-02 -9.50449944e-01 -5.39361537e-01 -5.93910635e-01 4.15607423e-01 -7.98048675e-01 1.36184543e-01 1.03170526e+00 7.02876568e-01 3.11819613e-01 -3.76896918e-01 -1.12642162e-01 5.55744946e-01 2.68809736e-01 7.29866743e-01 -8.35620522e-01 -2.51947850e-01 -7.94770479e-01 -6.44237455e-03 -1.23910320e+00 3.96637946e-01 -1.18883431e+00 -3.89436930e-01 -1.85349119e+00 7.24954188e-01 4.20789778e-01 -4.88973968e-02 3.79087269e-01 -1.12809405e-01 3.28226119e-01 5.49292326e-01 1.39930576e-01 -1.25463915e+00 4.18999523e-01 1.71637011e+00 -1.24612018e-01 4.56136130e-02 -5.16865611e-01 -5.65708756e-01 8.34646046e-01 7.63551593e-01 7.55533502e-02 -2.50325710e-01 -4.33331966e-01 7.45187759e-01 2.44308591e-01 5.01552224e-01 -6.24131501e-01 4.31439281e-01 -4.23737168e-01 3.41132343e-01 -6.63308024e-01 4.51715410e-01 -5.39763868e-01 -2.34503642e-01 5.27062058e-01 -5.71102381e-01 4.07021761e-01 3.77610743e-01 2.08790958e-01 -4.19216812e-01 -2.32005268e-01 3.64137948e-01 -4.60565388e-01 -9.67367113e-01 -4.11715835e-01 -9.35327947e-01 -3.79900821e-02 9.00488138e-01 1.10859185e-01 -7.23229408e-01 -8.56404185e-01 -5.78887105e-01 4.92324859e-01 3.10723096e-01 6.54036403e-01 9.10555720e-01 -1.30221224e+00 -5.70860982e-01 -4.37164754e-02 5.52040115e-02 -4.50456023e-01 5.69244623e-01 1.12238884e+00 -7.50384808e-01 7.58666515e-01 -9.80976447e-02 -4.56075072e-01 -1.43229961e+00 6.16732597e-01 -1.10135116e-02 -1.95709039e-02 -4.53527987e-01 7.76115119e-01 2.52644479e-01 -5.91181695e-01 3.49180192e-01 -1.16887309e-01 -2.65316635e-01 -3.01707145e-02 7.07418740e-01 4.49683905e-01 -5.32743275e-01 -5.40376127e-01 -2.13209674e-01 7.76168406e-01 2.58153200e-01 -2.65151232e-01 1.16854942e+00 -3.64705235e-01 -7.12912321e-01 1.18560433e+00 8.60522747e-01 2.13929638e-01 -6.49183154e-01 2.35416694e-03 -1.03089966e-01 -1.18780345e-01 -2.82327086e-01 -9.59799409e-01 -6.19887471e-01 1.41121662e+00 2.92846113e-01 7.68095488e-05 9.99717772e-01 3.23181719e-01 6.64027691e-01 6.31341279e-01 2.50600874e-01 -9.35917795e-01 5.70257425e-01 8.64499569e-01 9.75410879e-01 -1.40379107e+00 -4.33473382e-03 -1.25382572e-01 -8.66902053e-01 1.59627020e+00 6.01739168e-01 2.82105416e-01 4.03856337e-01 4.69073534e-01 2.83211499e-01 -3.93598586e-01 -1.17559230e+00 -2.45539378e-02 2.60538667e-01 3.47224087e-01 1.00630975e+00 9.25506428e-02 -9.66227800e-02 3.65479589e-01 -3.56469840e-01 9.24352407e-02 5.12124658e-01 9.80478048e-01 -5.68402767e-01 -1.01187730e+00 -7.81104505e-01 -5.43700950e-03 -1.34210154e-01 -2.77429253e-01 -6.33514524e-01 8.25456619e-01 -3.08039635e-02 1.01461589e+00 -1.18301203e-02 -2.65610367e-01 -6.30521774e-02 3.94907385e-01 8.03479254e-01 -2.82508910e-01 -4.94194478e-01 1.66550070e-01 -1.43406197e-01 -3.85503650e-01 -3.34808975e-01 -3.78688961e-01 -1.41931379e+00 -6.91101253e-01 1.65881142e-01 -2.51703411e-01 5.57514966e-01 9.68978405e-01 3.50255072e-02 3.98211539e-01 -2.67090142e-01 -9.79030669e-01 2.74763573e-02 -6.47197664e-01 -1.69906303e-01 -5.45965284e-02 2.79414713e-01 -2.80943125e-01 -1.26600889e-02 2.96754539e-01]
[10.940729141235352, 1.7333155870437622]
e4d77c9e-d90c-4a29-a63a-d3ac39ba94d3
to-aggregate-or-not-learning-with-separate
2206.07181
null
https://arxiv.org/abs/2206.07181v2
https://arxiv.org/pdf/2206.07181v2.pdf
To Aggregate or Not? Learning with Separate Noisy Labels
The rawly collected training data often comes with separate noisy labels collected from multiple imperfect annotators (e.g., via crowdsourcing). A typical way of using these separate labels is to first aggregate them into one and apply standard training methods. The literature has also studied extensively on effective aggregation approaches. This paper revisits this choice and aims to provide an answer to the question of whether one should aggregate separate noisy labels into single ones or use them separately as given. We theoretically analyze the performance of both approaches under the empirical risk minimization framework for a number of popular loss functions, including the ones designed specifically for the problem of learning with noisy labels. Our theorems conclude that label separation is preferred over label aggregation when the noise rates are high, or the number of labelers/annotations is insufficient. Extensive empirical results validate our conclusions.
['Yang Liu', 'Abhishek Kumar', 'Ehsan Amid', 'Tianyi Luo', 'Zhaowei Zhu', 'Jiaheng Wei']
2022-06-14
null
null
null
null
['learning-with-noisy-labels', 'learning-with-noisy-labels']
['computer-vision', 'natural-language-processing']
[ 2.61030883e-01 2.95630962e-01 -1.96603283e-01 -5.15331805e-01 -1.42628193e+00 -8.68830740e-01 2.81169057e-01 4.38214958e-01 -7.15333223e-01 1.09290504e+00 -2.50893176e-01 -1.82148710e-01 1.30097181e-01 -3.72293860e-01 -4.94460255e-01 -9.91554320e-01 4.03874546e-01 6.86763644e-01 1.30451927e-02 2.41189584e-01 -6.38733134e-02 1.40930489e-01 -1.47902989e+00 -2.34692320e-01 1.01733112e+00 1.05207777e+00 -1.29387066e-01 4.50485229e-01 -3.08170736e-01 6.70916021e-01 -8.24215055e-01 -1.00664663e+00 5.62391460e-01 -4.20235753e-01 -1.09764493e+00 5.37395597e-01 3.00723433e-01 -2.17606723e-01 3.70827407e-01 1.43402541e+00 4.68425781e-01 2.10781619e-01 4.60462838e-01 -1.26878679e+00 -4.09310341e-01 7.52541542e-01 -2.59385318e-01 -2.28060856e-01 1.51600718e-01 -1.52063459e-01 1.17865682e+00 -8.16435337e-01 2.71994770e-01 9.34117675e-01 7.83710539e-01 7.65269518e-01 -1.35758913e+00 -4.02474582e-01 9.33172330e-02 -3.48705977e-01 -1.45069289e+00 -1.95392996e-01 3.78307492e-01 -5.55704892e-01 1.49377659e-01 3.93695891e-01 -4.23810519e-02 9.28390205e-01 -5.27025282e-01 8.03100586e-01 1.50423145e+00 -6.46287560e-01 4.59689111e-01 4.81867373e-01 5.04749477e-01 5.67157686e-01 6.52870655e-01 -2.68374205e-01 -4.04118747e-01 -8.04391980e-01 1.66067883e-01 -2.13856488e-01 -2.50986546e-01 -1.00188375e-01 -7.19983101e-01 9.77944255e-01 -8.13614130e-02 2.30491504e-01 -3.12380344e-02 1.53164640e-02 4.43352252e-01 3.37009847e-01 1.04416597e+00 4.84814674e-01 -4.74803090e-01 2.21834213e-01 -9.76621032e-01 3.03958684e-01 1.01700878e+00 9.97668028e-01 9.53213871e-01 -2.14987651e-01 -2.20830113e-01 9.65174735e-01 2.71404713e-01 5.32892108e-01 1.91248357e-01 -1.16494310e+00 4.22478408e-01 3.97641212e-01 8.44084501e-01 -6.81500673e-01 -2.50157773e-01 -1.36493340e-01 -4.83586013e-01 7.13731647e-02 9.29648161e-01 -5.21825194e-01 -6.27044380e-01 1.72183359e+00 4.74007607e-01 -4.08506282e-02 -2.59612799e-01 1.04392326e+00 4.53618199e-01 1.29922181e-01 3.25708210e-01 -4.36131716e-01 1.23501825e+00 -8.46672297e-01 -9.10320401e-01 -1.06307089e-01 1.13253820e+00 -7.16756165e-01 1.00296724e+00 5.63389003e-01 -9.32444453e-01 -1.75214276e-01 -7.89641082e-01 -2.04818949e-01 -3.50246906e-01 4.33914363e-01 3.26731205e-01 1.13890839e+00 -7.08813548e-01 7.85677373e-01 -3.90284419e-01 -9.02369097e-02 2.86488146e-01 4.09419328e-01 -2.43469864e-01 -1.46612510e-01 -1.25138390e+00 7.52075195e-01 1.66415557e-01 7.24189132e-02 -6.49582803e-01 -1.13257095e-01 -6.67888105e-01 -3.26040626e-01 8.02298307e-01 -2.80929714e-01 1.61779428e+00 -1.06011641e+00 -1.19409633e+00 1.16329563e+00 -2.09979355e-01 -3.14169079e-01 8.61727178e-01 -4.06322807e-01 6.94882050e-02 -1.66093796e-01 4.60132837e-01 1.22061543e-01 4.59121436e-01 -1.68018150e+00 -7.85397708e-01 -3.35440516e-01 3.18655044e-01 1.38194039e-01 -2.61700690e-01 2.76622474e-01 -4.61047217e-02 -5.30858159e-01 1.20505169e-01 -1.15478110e+00 -4.01897013e-01 -6.18045740e-02 -5.90977788e-01 -6.47150636e-01 3.63107085e-01 -4.24190640e-01 1.12928951e+00 -2.04252100e+00 -1.28109097e-01 1.55894697e-01 3.06727141e-01 3.03440988e-01 2.35458910e-01 1.69487715e-01 1.26121238e-01 7.14205742e-01 -5.77840745e-01 -1.07528627e+00 1.63818613e-01 5.46243131e-01 -1.74630925e-01 7.25945711e-01 8.10202211e-02 5.95179558e-01 -1.25563097e+00 -6.32794023e-01 -2.30033800e-01 -2.26141158e-02 1.59733176e-01 2.04391167e-01 -2.70429671e-01 5.16928136e-01 -4.82105911e-01 4.88417447e-01 6.45842314e-01 -2.82843769e-01 2.67361879e-01 1.82408094e-01 2.26965815e-01 3.23815197e-01 -1.55419922e+00 1.11990464e+00 -2.13690996e-01 1.97089851e-01 3.73252630e-01 -8.34973097e-01 6.42701983e-01 6.34642839e-01 4.58105743e-01 -1.67994387e-02 4.09929216e-01 6.14864469e-01 -6.00617468e-01 -5.03526747e-01 2.83624381e-01 -5.69413602e-01 -2.86014766e-01 7.65622675e-01 5.99112222e-03 2.83343922e-02 2.85266817e-01 5.46595175e-03 9.65465307e-01 -1.33687034e-01 3.73356611e-01 -2.09156662e-01 3.32282960e-01 -1.18078604e-01 8.48652422e-01 9.80421245e-01 -3.68237913e-01 7.49762297e-01 5.56537867e-01 -2.95329124e-01 -9.82522726e-01 -5.52120924e-01 -2.37131178e-01 1.18743837e+00 1.92168936e-01 -2.19777286e-01 -9.13779199e-01 -1.24592853e+00 -1.24619216e-01 4.43430871e-01 -6.43711329e-01 2.54863739e-01 -3.44454020e-01 -8.27559412e-01 6.62814796e-01 4.52512473e-01 2.67277360e-01 -6.10871494e-01 -2.53830582e-01 1.73463583e-01 -6.06342494e-01 -1.17781126e+00 -3.48365754e-01 5.16588330e-01 -6.18349314e-01 -1.10984778e+00 -6.20633483e-01 -5.12504280e-01 8.40731263e-01 2.98677474e-01 1.13319528e+00 3.09991539e-01 3.39984477e-01 3.50489736e-01 -6.64039910e-01 -5.03426433e-01 -3.96610767e-01 1.39004946e-01 2.13257656e-01 3.07012558e-01 5.33681989e-01 -1.54759958e-01 2.35636644e-02 5.40696621e-01 -1.09293199e+00 -5.68016171e-01 1.01753935e-01 7.83656001e-01 5.22782743e-01 -5.97216599e-02 6.89485967e-01 -1.44566011e+00 7.84445226e-01 -6.86161935e-01 -6.41888022e-01 4.57129002e-01 -6.30495310e-01 4.48016785e-02 6.03823662e-01 -4.13829118e-01 -8.70554924e-01 2.68424571e-01 8.96340534e-02 -1.24903053e-01 -3.92497033e-01 1.56758815e-01 -3.49065095e-01 -1.71614796e-01 7.66537070e-01 -3.27496260e-01 -3.08031291e-01 -5.75873375e-01 3.03302705e-01 8.74660671e-01 1.18599512e-01 -8.84005785e-01 6.51112378e-01 5.19723177e-01 -1.94152564e-01 -3.64111573e-01 -1.44838500e+00 -7.13802993e-01 -5.83496988e-01 -1.30809337e-01 7.78519928e-01 -7.37729311e-01 -4.69722867e-01 3.05670738e-01 -1.31579089e+00 -2.58703142e-01 -4.68725353e-01 2.45207638e-01 -3.85969341e-01 7.22550631e-01 -7.13803172e-01 -1.43422067e+00 -4.02630754e-02 -1.14992261e+00 1.31537902e+00 2.01213494e-01 -1.53000042e-01 -9.06581521e-01 4.33519185e-02 4.73166019e-01 -4.61499998e-03 1.16155006e-01 3.84532869e-01 -1.22113991e+00 -2.68093377e-01 -6.16037250e-01 -9.22428370e-02 7.92873025e-01 8.48442018e-02 -3.73008251e-01 -1.24491227e+00 1.80845130e-02 3.42168838e-01 -6.98468566e-01 5.85304081e-01 6.57962263e-02 9.47016597e-01 -3.94727856e-01 -6.33330494e-02 -6.03425615e-02 1.49931240e+00 -1.98093340e-01 8.51726606e-02 4.06402275e-02 6.63486004e-01 9.19435024e-01 6.26320899e-01 4.73705798e-01 2.52561808e-01 5.64411104e-01 2.51885712e-01 1.34665772e-01 4.89270449e-01 1.35217056e-01 6.05829209e-02 6.16076231e-01 5.48303127e-02 -5.16987741e-01 -7.10060775e-01 5.39016068e-01 -2.05197906e+00 -4.52417016e-01 -3.14062685e-01 2.38548660e+00 1.02262855e+00 -1.07209332e-01 2.82722980e-01 3.21068138e-01 9.78914261e-01 -1.27801687e-01 -1.12469472e-01 -1.55022591e-01 -2.79652894e-01 -7.32769519e-02 8.19921553e-01 6.46725833e-01 -1.27898669e+00 6.29199266e-01 6.18742609e+00 1.01392174e+00 -4.32199955e-01 5.53831100e-01 8.86988342e-01 -3.85950655e-02 -1.55215949e-01 9.61072817e-02 -8.69674563e-01 6.30293071e-01 8.57288241e-01 7.30178654e-02 2.18110487e-01 7.04188943e-01 8.80072340e-02 -5.13580918e-01 -1.10805261e+00 7.06349850e-01 -2.39906814e-02 -6.94928765e-01 -4.72797185e-01 1.22931510e-01 9.90952373e-01 -2.45846093e-01 -3.30127567e-01 2.39097327e-02 8.75256777e-01 -8.90425801e-01 7.28004336e-01 5.10754108e-01 4.58754241e-01 -4.42098677e-01 1.27772534e+00 6.32134318e-01 -8.58153462e-01 -1.57145768e-01 -2.85901129e-01 -1.73166305e-01 3.80448729e-01 1.20636189e+00 -3.31531972e-01 6.74044311e-01 4.64060932e-01 -2.21948307e-02 -4.54171658e-01 1.15178454e+00 -3.54017854e-01 8.96275163e-01 -7.05295503e-01 1.43936444e-02 1.79350331e-01 -2.42496267e-01 4.19654429e-01 1.04006469e+00 -6.20574225e-03 9.50875953e-02 5.93141735e-01 4.83410984e-01 -3.10406625e-01 2.71448791e-01 -5.85198641e-01 2.58834541e-01 6.07585907e-01 1.31298518e+00 -8.55455220e-01 -5.08855402e-01 -4.77542490e-01 7.36840487e-01 5.31474352e-01 4.21347082e-01 -6.28122866e-01 -3.38483639e-02 3.07230264e-01 3.90988067e-02 -1.21774405e-01 -1.04573518e-01 -6.23283684e-01 -9.11163807e-01 4.55862403e-01 -7.37021029e-01 2.87615478e-01 -4.29795951e-01 -1.66791022e+00 5.69885373e-01 1.63365400e-03 -1.05007458e+00 1.36060029e-01 -5.43812573e-01 -2.38068447e-01 8.67558658e-01 -1.23268008e+00 -6.39691055e-01 -1.33136973e-01 1.28415182e-01 3.41794819e-01 4.09507245e-01 7.12199628e-01 4.22019690e-01 -7.07636118e-01 5.87317765e-01 2.03922153e-01 2.91218698e-01 8.15002084e-01 -1.56574011e+00 -6.19138218e-02 7.91137159e-01 1.69986427e-01 3.21048409e-01 7.48794258e-01 -6.12921774e-01 -4.63338047e-01 -1.04104674e+00 1.20339119e+00 -8.16409886e-01 5.17008424e-01 -4.84794438e-01 -1.06380510e+00 3.81870776e-01 -4.89788279e-02 1.83139011e-01 1.06230962e+00 -8.72615054e-02 -3.44516188e-01 1.23170942e-01 -1.39956355e+00 2.99726099e-01 7.94350624e-01 -5.21018028e-01 -2.35026568e-01 7.40812600e-01 5.73769093e-01 -2.15217233e-01 -8.74786556e-01 3.12769204e-01 8.29167962e-02 -8.61739933e-01 3.68927449e-01 -5.64822197e-01 7.52063394e-02 -3.27425212e-01 -3.16933572e-01 -1.21663439e+00 1.53549641e-01 -6.68365538e-01 2.79116482e-01 1.53193748e+00 4.91878390e-01 -6.37772441e-01 6.93206251e-01 1.17215955e+00 2.95279384e-01 -7.41616488e-01 -1.04163218e+00 -8.64287078e-01 6.19755574e-02 -4.02144492e-01 4.93221849e-01 9.53907788e-01 -8.32332671e-02 2.91285008e-01 -5.69295228e-01 1.25046596e-01 6.72288954e-01 -3.75027984e-01 4.85153288e-01 -1.46265602e+00 -7.37797469e-02 -8.43109563e-02 -3.27126347e-02 -7.67414987e-01 4.20782745e-01 -5.47484338e-01 5.58478892e-01 -1.27696526e+00 2.44305399e-03 -8.53534758e-01 4.07392755e-02 5.82979262e-01 -5.57476044e-01 4.35343355e-01 2.03592554e-01 3.33135366e-01 -8.87912095e-01 3.00725132e-01 7.92147458e-01 1.23402722e-01 1.62484631e-01 3.96754056e-01 -7.49750555e-01 8.21861148e-01 9.01160598e-01 -8.81412864e-01 -1.52748421e-01 -3.76148015e-01 5.09891331e-01 -2.30290458e-01 1.90963611e-01 -8.15333068e-01 2.12191343e-01 -7.70446509e-02 -2.93116868e-01 1.15037501e-01 8.06243066e-03 -1.02254438e+00 -1.77602902e-01 -2.26489455e-01 -6.87454700e-01 -1.41444221e-01 -2.76032001e-01 9.27728951e-01 -1.30535871e-01 -1.04926157e+00 8.04723084e-01 -3.44550103e-01 -4.61611226e-02 4.22078744e-02 -2.98782259e-01 3.55211437e-01 9.14056897e-01 -9.56519321e-02 -1.56646475e-01 -3.62143993e-01 -1.09235561e+00 3.23395193e-01 3.84970665e-01 -1.35084063e-01 -1.63273856e-01 -1.31378734e+00 -7.23744988e-01 -4.63612199e-01 1.09898947e-01 1.30561665e-01 -2.06971675e-01 8.69022548e-01 -6.36350513e-02 -1.38678420e-02 6.64566338e-01 -3.58116955e-01 -1.20272100e+00 5.61953962e-01 4.16249245e-01 -5.40753543e-01 -9.69674587e-02 8.82402599e-01 -9.23779756e-02 -5.28296411e-01 5.26659906e-01 -1.43488958e-01 -1.95328835e-02 3.09956402e-01 3.51275891e-01 7.33942807e-01 2.86072999e-01 -7.35940278e-01 -1.09923773e-01 3.96511465e-01 2.73898572e-01 -1.90425530e-01 7.98783183e-01 -5.18769503e-01 -2.36450151e-01 7.98477829e-01 1.13781500e+00 1.60934746e-01 -1.01228976e+00 -6.83817744e-01 4.46761280e-01 -3.95115882e-01 -4.34778303e-01 -5.00709414e-01 -8.38238418e-01 6.01333261e-01 4.03211683e-01 8.77561688e-01 9.39152181e-01 6.16042838e-02 6.18254960e-01 4.40399349e-01 6.67362452e-01 -1.37634873e+00 -1.27665162e-01 3.08650911e-01 3.06943059e-01 -1.53394282e+00 -6.61916137e-02 -6.86149359e-01 -8.03866804e-01 6.31559134e-01 4.58791018e-01 -1.87126189e-01 5.79308391e-01 2.12161988e-01 2.60180235e-01 -4.46846522e-02 -2.43210375e-01 -5.16298175e-01 1.19907044e-01 3.94521326e-01 3.95855635e-01 2.92879164e-01 -7.39529014e-01 8.34742427e-01 1.41233951e-01 -1.67266130e-01 6.32803679e-01 9.50530529e-01 -5.20962179e-01 -1.43145823e+00 -6.45324290e-01 6.33341789e-01 -9.16274726e-01 5.85126951e-02 -6.83029950e-01 1.91339374e-01 5.89983344e-01 1.48558772e+00 -4.75203753e-01 -1.16999924e-01 2.86577582e-01 5.85599184e-01 1.39891535e-01 -8.59013855e-01 -7.46273398e-01 -4.52751480e-02 2.83104569e-01 -2.75256217e-01 -9.40360904e-01 -5.46508431e-01 -8.43847573e-01 2.22838186e-02 -1.01558888e+00 6.37599230e-01 4.83405560e-01 1.43965030e+00 -9.23419446e-02 -7.64723197e-02 6.47968709e-01 -7.20119774e-01 -9.35409546e-01 -9.04619396e-01 -8.58531713e-01 5.74718952e-01 3.55544269e-01 -8.17888141e-01 -8.11327040e-01 8.06560814e-02]
[9.579669952392578, 4.563762664794922]
024d81aa-bf20-4926-b860-d26043fe608a
stereo-r-cnn-based-3d-object-detection-for
1902.09738
null
http://arxiv.org/abs/1902.09738v2
http://arxiv.org/pdf/1902.09738v2.pdf
Stereo R-CNN based 3D Object Detection for Autonomous Driving
We propose a 3D object detection method for autonomous driving by fully exploiting the sparse and dense, semantic and geometry information in stereo imagery. Our method, called Stereo R-CNN, extends Faster R-CNN for stereo inputs to simultaneously detect and associate object in left and right images. We add extra branches after stereo Region Proposal Network (RPN) to predict sparse keypoints, viewpoints, and object dimensions, which are combined with 2D left-right boxes to calculate a coarse 3D object bounding box. We then recover the accurate 3D bounding box by a region-based photometric alignment using left and right RoIs. Our method does not require depth input and 3D position supervision, however, outperforms all existing fully supervised image-based methods. Experiments on the challenging KITTI dataset show that our method outperforms the state-of-the-art stereo-based method by around 30% AP on both 3D detection and 3D localization tasks. Code has been released at https://github.com/HKUST-Aerial-Robotics/Stereo-RCNN.
['Shaojie Shen', 'Xiaozhi Chen', 'Peiliang Li']
2019-02-26
stereo-r-cnn-based-3d-object-detection-for-1
http://openaccess.thecvf.com/content_CVPR_2019/html/Li_Stereo_R-CNN_Based_3D_Object_Detection_for_Autonomous_Driving_CVPR_2019_paper.html
http://openaccess.thecvf.com/content_CVPR_2019/papers/Li_Stereo_R-CNN_Based_3D_Object_Detection_for_Autonomous_Driving_CVPR_2019_paper.pdf
cvpr-2019-6
['3d-object-detection-from-stereo-images']
['computer-vision']
[-1.36594459e-01 7.01423213e-02 -6.13542609e-02 -4.56393182e-01 -6.17484212e-01 -7.86634445e-01 5.47438085e-01 -3.03063393e-01 -4.99504447e-01 2.42560819e-01 -3.66329215e-02 -1.71754614e-01 3.21249962e-01 -6.12849057e-01 -1.19277489e+00 -3.19691896e-01 3.28905731e-01 6.57940567e-01 7.44328916e-01 -2.85106242e-01 5.15708089e-01 8.64934742e-01 -1.90869379e+00 1.54720843e-02 5.01931727e-01 1.19856310e+00 6.23499870e-01 4.08272386e-01 2.85465270e-01 5.02304971e-01 1.02486499e-01 2.24090979e-01 1.03134632e+00 1.67004809e-01 -5.30032456e-01 1.38043255e-01 9.21239078e-01 -8.73266101e-01 -6.20155156e-01 9.55086589e-01 3.22723627e-01 1.31235003e-01 5.33360839e-01 -1.04830146e+00 -1.77800596e-01 8.77072215e-02 -1.00793159e+00 9.33600068e-02 3.54226887e-01 3.11978847e-01 7.82051206e-01 -1.18316174e+00 7.32410491e-01 1.37421250e+00 7.01638758e-01 3.11385483e-01 -1.22918093e+00 -8.33798409e-01 1.89057633e-01 -1.99749377e-02 -1.64752686e+00 -2.75541812e-01 7.67849147e-01 -5.72453260e-01 1.18756580e+00 -1.67752564e-01 8.07874918e-01 7.62251914e-01 -6.17062971e-02 7.35002577e-01 1.13195384e+00 -7.84184858e-02 1.16253120e-03 -5.18108830e-02 -2.87408987e-03 9.93864834e-01 1.47620142e-01 6.01764023e-01 -4.43304062e-01 1.19591847e-01 1.25007224e+00 4.12244707e-01 -9.24613848e-02 -1.00054586e+00 -1.50014877e+00 7.87319899e-01 9.85306323e-01 -3.88999164e-01 -2.51901597e-01 9.40103158e-02 5.41161746e-02 -1.82030916e-01 5.79558730e-01 3.20585251e-01 -5.27828813e-01 3.54326934e-01 -6.84701025e-01 4.72716659e-01 3.09911758e-01 1.40510941e+00 1.32072794e+00 -1.67524576e-01 2.32178971e-01 6.86665058e-01 2.99499393e-01 9.64086294e-01 6.82852715e-02 -1.63049996e+00 5.70283890e-01 8.74245822e-01 2.39992201e-01 -7.96194673e-01 -4.45946485e-01 -1.74891934e-01 -6.04336977e-01 5.71015894e-01 2.42185950e-01 1.84181675e-01 -1.10955310e+00 1.13938904e+00 6.03683949e-01 2.53056884e-02 -7.13678896e-02 1.24928582e+00 1.08745587e+00 4.82408136e-01 -7.15429842e-01 5.94325185e-01 1.11480999e+00 -1.07610130e+00 1.11752041e-01 -8.86203706e-01 4.00390446e-01 -5.54032803e-01 6.14669859e-01 8.16973299e-02 -9.17503238e-01 -7.07172334e-01 -8.06964815e-01 -6.02591813e-01 -3.08233142e-01 4.18595076e-01 3.82905602e-01 -8.09651390e-02 -1.07307112e+00 2.28264436e-01 -5.80734491e-01 -4.00319040e-01 5.08660793e-01 2.51213670e-01 -7.31337309e-01 -2.81734705e-01 -6.04483128e-01 8.39864016e-01 4.61866140e-01 2.20775139e-02 -1.36930633e+00 -7.23500848e-01 -1.42331922e+00 -4.20052201e-01 4.22352940e-01 -8.71500909e-01 1.17772019e+00 -3.98629367e-01 -1.13753104e+00 1.48866570e+00 -3.10370356e-01 -6.00202620e-01 4.94733185e-01 -4.37370121e-01 3.56205255e-01 9.49183851e-02 5.31861365e-01 1.57425976e+00 7.82260597e-01 -1.45202839e+00 -9.62847829e-01 -7.85977960e-01 1.48611605e-01 5.23818314e-01 6.62826777e-01 -3.22074354e-01 -6.11101151e-01 -8.40999708e-02 7.97254920e-01 -1.02288330e+00 -4.57320899e-01 2.20429838e-01 -4.89070922e-01 -9.03361887e-02 7.20625341e-01 -3.96471471e-01 1.33661330e-01 -2.05976391e+00 7.54021555e-02 -9.82864574e-02 2.96668679e-01 -1.35198355e-01 -5.70651032e-02 -7.97866508e-02 3.06661036e-02 -4.30935234e-01 -1.91389620e-01 -4.05791193e-01 -1.94894776e-01 1.17967755e-01 -4.41389501e-01 8.28108907e-01 2.68738210e-01 8.67140651e-01 -8.72540593e-01 -2.83032864e-01 8.10795248e-01 4.50300306e-01 -6.53392076e-01 2.21667245e-01 -1.98698744e-01 6.82063818e-01 -4.11603898e-01 8.23796511e-01 9.75933850e-01 -9.80861336e-02 -4.37345475e-01 -1.60978228e-01 -5.53144634e-01 4.37736720e-01 -1.03327930e+00 1.93743026e+00 -3.57409298e-01 7.55869806e-01 1.79794595e-01 -7.81118095e-01 1.30243647e+00 -3.82375479e-01 1.30699381e-01 -7.79483378e-01 2.01292053e-01 4.06110704e-01 -4.85437661e-01 -9.91551429e-02 6.57669187e-01 3.41410547e-01 -1.00418232e-01 -6.46165609e-02 -1.17891459e-02 -7.60520399e-01 -1.50040478e-01 9.37104225e-02 8.79835725e-01 5.94283760e-01 3.42549890e-01 -2.28121042e-01 5.40901065e-01 4.06134188e-01 5.99831522e-01 5.90903282e-01 -1.50018901e-01 9.78693783e-01 1.92062855e-01 -6.93120778e-01 -1.41949558e+00 -9.54687715e-01 -2.03649685e-01 6.08601391e-01 8.02368402e-01 -2.84671914e-02 -3.03884089e-01 -5.02216339e-01 5.94475925e-01 4.20258999e-01 -5.66968799e-01 2.15730608e-01 -5.92641234e-01 1.92461982e-01 5.81297316e-02 6.46838605e-01 5.65301299e-01 -8.62491429e-01 -8.69239509e-01 -1.56110600e-01 -2.14123815e-01 -1.42753005e+00 -3.81037921e-01 5.40183306e-01 -9.81742501e-01 -1.01551402e+00 -6.57593310e-01 -7.32991159e-01 8.28610599e-01 1.02278173e+00 9.53442335e-01 -4.04653490e-01 -3.12533259e-01 1.83490455e-01 -1.46645591e-01 -3.77551168e-01 -3.28335539e-02 3.08523495e-02 3.84992361e-02 -4.63656753e-01 3.82496715e-01 -4.50095296e-01 -8.15940738e-01 7.41276324e-01 -3.23520303e-01 4.44430947e-01 6.35724187e-01 5.76597214e-01 9.90974367e-01 -5.66119611e-01 -3.86253715e-01 -3.94584119e-01 -4.71034676e-01 -2.95593679e-01 -1.36597216e+00 -4.23099428e-01 -2.14019015e-01 -1.12064453e-02 2.16787517e-01 7.49335960e-02 -6.62885249e-01 7.72874534e-01 3.03232167e-02 -9.09190595e-01 -5.56228876e-01 -2.10058942e-01 1.89963162e-01 -2.10215658e-01 7.68933117e-01 3.20907950e-01 1.39320269e-01 -3.71979684e-01 4.00050491e-01 6.16085470e-01 7.31678545e-01 -9.58591476e-02 1.00977004e+00 1.13528585e+00 1.64086431e-01 -4.34991658e-01 -1.30026686e+00 -1.08167374e+00 -1.16440737e+00 -2.77714878e-01 9.48260844e-01 -1.73020363e+00 -6.74058914e-01 4.01994765e-01 -1.39885128e+00 -5.64900994e-01 -2.10282430e-01 5.79500616e-01 -8.62608850e-01 1.73486948e-01 -2.70837963e-01 -7.03392506e-01 -2.07168281e-01 -1.30789518e+00 1.83294678e+00 2.01960392e-02 1.97838739e-01 -3.68383348e-01 -9.89005119e-02 7.69880474e-01 -1.42773734e-02 1.88528702e-01 -9.20975115e-03 -4.44519892e-02 -1.25502205e+00 -1.87484220e-01 -6.51148498e-01 2.48603299e-01 -2.82230616e-01 -3.70363235e-01 -1.07191086e+00 -1.43100142e-01 -2.85613090e-01 -5.07873297e-01 1.34931564e+00 5.62231660e-01 1.04490817e+00 1.86699897e-01 -5.03986120e-01 1.05055869e+00 1.37601244e+00 -2.17723787e-01 3.04714531e-01 5.15779674e-01 9.55804765e-01 6.56615496e-01 1.07090783e+00 4.73511517e-01 6.63297236e-01 6.37376487e-01 1.05235982e+00 -2.06088051e-01 -6.02076948e-02 -5.39437830e-01 3.76719147e-01 4.34978381e-02 -1.32517055e-01 2.37028584e-01 -1.01509297e+00 6.52981222e-01 -1.89075387e+00 -7.99562395e-01 -2.24372268e-01 2.16706848e+00 4.30808246e-01 1.06568135e-01 1.34686232e-01 -3.31341088e-01 6.65606260e-01 1.91684142e-01 -8.56788754e-01 1.83674142e-01 -7.78216049e-02 -3.52768004e-01 1.32705843e+00 5.62822640e-01 -1.23238051e+00 1.29934704e+00 5.27335310e+00 4.60527480e-01 -8.96098614e-01 -9.91172940e-02 5.36131978e-01 -6.80033937e-02 -1.18086606e-01 9.15983692e-02 -1.43024361e+00 -3.62259373e-02 2.70640016e-01 2.76096046e-01 4.76418495e-01 1.30194819e+00 2.87633032e-01 -3.46016139e-01 -1.02553439e+00 1.23120499e+00 1.72799423e-01 -1.48411059e+00 -1.84804261e-01 3.63999844e-01 9.99789178e-01 1.02711153e+00 -1.90393090e-01 8.69686231e-02 5.72917879e-01 -7.38363326e-01 1.26915050e+00 2.52571285e-01 8.77278149e-01 -5.97131550e-01 7.39247262e-01 6.39845908e-01 -1.32634342e+00 -2.39330992e-01 -7.47941792e-01 -1.86752871e-01 -1.76618248e-02 4.68227714e-01 -9.23168123e-01 3.24481308e-01 1.22526181e+00 1.28415883e+00 -5.86979449e-01 9.64294016e-01 -4.10776943e-01 -7.46625438e-02 -5.74914873e-01 2.90402144e-01 3.58418733e-01 -2.40472645e-01 7.35941112e-01 7.06024051e-01 3.67394835e-01 -2.62754541e-02 2.81602919e-01 1.24391842e+00 -1.82235260e-02 -2.86855698e-01 -1.16340601e+00 6.88532531e-01 5.65047801e-01 1.41935396e+00 -6.28565133e-01 -2.73604989e-01 -3.51905286e-01 8.29698145e-01 4.34663981e-01 3.05514093e-02 -7.15577424e-01 -1.23399913e-01 7.96033561e-01 3.42848390e-01 7.39387393e-01 -4.93861198e-01 -2.38481298e-01 -1.05396950e+00 1.49310559e-01 -3.99566710e-01 -2.05564015e-02 -1.27811682e+00 -9.91971910e-01 3.64641100e-01 3.18671167e-02 -1.56821644e+00 -1.11924797e-01 -8.23338389e-01 -1.42113507e-01 9.52855408e-01 -1.82906306e+00 -1.14538956e+00 -8.62035275e-01 5.37473559e-01 6.37537956e-01 -4.41899337e-02 3.32098812e-01 -1.84858084e-01 -1.46921858e-01 -1.57489181e-01 -9.08490196e-02 1.44475549e-01 6.51102304e-01 -1.16608691e+00 8.28396976e-01 7.49357998e-01 -1.09375179e-01 1.96319029e-01 3.28045964e-01 -6.04667962e-01 -1.36560392e+00 -1.46446407e+00 6.19739175e-01 -8.03547323e-01 4.24152792e-01 -6.43788338e-01 -4.72865999e-01 7.74043977e-01 -1.35998949e-01 4.52670723e-01 -2.32359886e-01 -2.86704659e-01 -6.23928249e-01 -3.66254836e-01 -1.02745104e+00 4.66909796e-01 1.51455462e+00 -5.45023918e-01 -5.67993701e-01 4.38701540e-01 9.52689528e-01 -9.23537552e-01 -4.84197795e-01 6.86429918e-01 3.60155463e-01 -1.27698767e+00 1.28867137e+00 1.54600009e-01 4.86756980e-01 -8.15554976e-01 -2.97920793e-01 -9.57274854e-01 -3.56351823e-01 -2.19089866e-01 2.49814510e-01 4.90085512e-01 3.36549938e-01 -6.31432712e-01 8.12940657e-01 -9.15518962e-03 -5.12045085e-01 -4.25688297e-01 -9.26302731e-01 -7.93092787e-01 -3.06330591e-01 -6.12318218e-01 3.76113892e-01 4.67820048e-01 -4.33142781e-01 3.04432631e-01 -1.50944903e-01 5.94724476e-01 8.76964092e-01 5.97024381e-01 1.28254187e+00 -1.29316425e+00 5.07549196e-02 -3.49966168e-01 -7.20513701e-01 -1.69332337e+00 2.03333348e-01 -9.63404655e-01 4.22869891e-01 -1.51063609e+00 1.58461526e-01 -2.88658589e-01 1.87018350e-01 5.55799127e-01 2.49793351e-01 6.16906941e-01 -1.12593621e-01 3.39820623e-01 -7.14053035e-01 6.47873461e-01 1.22942722e+00 -3.54572274e-02 -1.22344792e-01 -1.47364065e-01 -2.43074551e-01 8.36970568e-01 8.27786326e-01 -4.03352350e-01 1.79504580e-03 -5.11652768e-01 1.03093654e-01 7.31077492e-02 9.87286747e-01 -1.20800626e+00 2.19828889e-01 -7.12132975e-02 6.29802525e-01 -1.51661944e+00 6.64582014e-01 -8.66816759e-01 -1.94501698e-01 3.31359059e-01 -1.97386429e-01 -4.50057276e-02 2.78031319e-01 6.31553531e-01 -9.31352843e-03 1.06499054e-01 9.72095907e-01 -4.55197036e-01 -1.10520601e+00 5.44267237e-01 2.43791696e-02 -4.04227562e-02 1.17822695e+00 -4.28664058e-01 -2.07204655e-01 -1.08687721e-01 -3.27198744e-01 4.98254448e-01 1.01673925e+00 5.94956756e-01 9.48177636e-01 -1.03975284e+00 -5.97669840e-01 6.27972364e-01 5.65496385e-01 9.17970181e-01 1.41104549e-01 8.22786510e-01 -8.19063365e-01 7.05574870e-01 -2.82202214e-01 -1.26443875e+00 -1.27576244e+00 5.42437911e-01 4.37802434e-01 4.19892013e-01 -9.70802903e-01 8.87353361e-01 6.90949500e-01 -1.04557467e+00 2.36481905e-01 -6.14198744e-01 1.01060241e-01 -4.42361325e-01 3.02528620e-01 1.60656452e-01 -1.47676691e-01 -8.51962268e-01 -4.43685025e-01 9.93877053e-01 1.20042525e-01 9.76991840e-03 1.45674574e+00 -3.55261922e-01 -1.23957396e-01 3.35116625e-01 1.09631360e+00 -3.95948976e-01 -1.83532321e+00 -4.68510330e-01 -4.49580878e-01 -6.54934227e-01 3.96495074e-01 -3.64081174e-01 -9.95169580e-01 9.31497395e-01 5.63953221e-01 -3.81624013e-01 5.99197865e-01 5.08899689e-01 4.97603118e-01 5.52574873e-01 6.60678327e-01 -8.58559191e-01 -1.60867590e-02 9.58212912e-01 1.04817271e+00 -1.77516711e+00 1.12288535e-01 -5.20515680e-01 -4.85775262e-01 1.07831502e+00 9.21770632e-01 -4.89299357e-01 5.15887439e-01 7.87186548e-02 -2.48495787e-02 -3.08233202e-01 -4.70835745e-01 -6.40444100e-01 4.38823581e-01 6.91513836e-01 -8.05906579e-02 -2.96019968e-02 5.71987152e-01 -1.15552768e-01 -2.74227172e-01 -2.50543207e-01 2.94746757e-01 7.32949317e-01 -6.78321123e-01 -4.73839909e-01 -5.32307208e-01 1.08281270e-01 2.36286953e-01 -1.03237242e-01 -4.26667839e-01 8.30985427e-01 3.29576373e-01 6.18688464e-01 4.99613315e-01 -4.28656757e-01 3.88058513e-01 -4.06186700e-01 5.28442442e-01 -8.85050714e-01 3.60411704e-02 -2.64298283e-02 -1.89786643e-01 -1.08159196e+00 -3.75079870e-01 -7.55484164e-01 -1.25494170e+00 -3.59206721e-02 -2.32261524e-01 -4.97774601e-01 8.33231509e-01 6.53715491e-01 5.82375288e-01 -8.16685408e-02 7.64904082e-01 -1.71577561e+00 -3.01731229e-01 -7.96624541e-01 -5.19269049e-01 -5.54278586e-03 5.96826553e-01 -8.35197330e-01 -5.19263387e-01 -1.38723493e-01]
[7.821316242218018, -2.5995378494262695]