File size: 21,725 Bytes
7eded29 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 |
"""A Streamlit app designed to help with data labeling with explainable
machine learning approach. It can handle data with many labels and many
classes. For each label, an Explainable Boosting Machine model is
trained on the labeled data, then it makes class predictions and
provides per-instance local explanations, which are then used to make
heatmaps, displayed in the main screen.
Model page: https://huggingface.co/Donlapark/XLabel
"""
import json
import math
import os
import pickle as pkle
from interpret.glassbox.ebm.ebm import ExplainableBoostingClassifier
import numpy as np
import pandas as pd
import altair as alt
import streamlit as st
from streamlit import session_state as _state
import streamlit.components.v1 as components
_PERSIST_STATE_KEY = f"{__name__}_PERSIST"
_CONFIGS_FILE = "configs.json"
_MODEL = "_saved_models.pickle"
_NUM_FEAT_PER_ROW = 11
st.set_page_config(layout="wide")
def main():
"""The main Streamlit app."""
if "configs" not in _state:
try:
with open(_CONFIGS_FILE, "r") as _file:
_state["configs"] = json.load(_file)
except FileNotFoundError:
create_config_file()
_state["loaded_new_file"] = True
st.sidebar.write("Current database: " + _state.configs["db_filename"])
st.sidebar.file_uploader("Upload a CSV or Excel file",
type=["csv", "xlsx", "xls"],
key="uploaded_files",
accept_multiple_files=False,
on_change=update_file)
with st.sidebar.form("sidebar"):
st.slider("Number of labels",
min_value=1,
max_value=20,
value=_state.configs["sidebar"]["num_labels"],
step=1,
key="num_labels")
st.selectbox("Include data with label/prediction mismatches?",
("Yes", "No"),
key="relabel",
index=("Yes",
"No").index(_state.configs["sidebar"]["relabel"]))
st.selectbox("Sampling mode",
("Fixed sample size", "Confidence threshold"),
key="mode",
index=("Fixed sample size", "Confidence threshold").index(
_state.configs["sidebar"]["mode"]))
st.slider("Sample size (for \"Fixed sample size\" mode)",
min_value=1,
max_value=500,
value=_state.configs["sidebar"]["n_samples"],
step=1,
key="n_samples")
st.slider("Threshold (for \"Confidence threshold\" mode)",
min_value=0.00,
max_value=1.00,
value=_state.configs["sidebar"]["threshold"],
step=0.01,
format="%.2f",
key="threshold")
form_cols = st.columns((2, 2, 2))
form_cols[1].form_submit_button("Sample", on_click=sample_and_predict)
if "pages" in _state:
page_list = list(_state.pages)
tabs = st.tabs(page_list)
for i, tab in enumerate(tabs):
with tab:
display_main_screen(page_list[i])
filename = _state.configs["db_filename"]
file_pre, file_ext = os.path.splitext(filename)
if "database" in _state:
data, mime = convert_to_downloadable(_state.database, file_ext)
st.sidebar.download_button(label="Download labeled data",
data=data,
file_name=filename,
mime=mime)
def update_file():
"""Update the state parameters after a file has been uploaded"""
if _state.uploaded_files is not None:
_state.configs["db_filename"] = _state.uploaded_files.name
_state.loaded_new_file = True
def init_state_params():
"""Initialize all state parameters.
This function will be called by sample_and_predict() when
_state.pages has not been initialized.
State parameters:
database: The pandas dataframe of the database.
configs: The saved configs of sidebar widgets.
pages: The list of all labels.
classes: The dict of all classes for each label.
class_to_num: The encoding dict of classes into integers.
num_to_class: The decoding dict of integers into classes.
class_to_num: The dict that maps each class to the
corresponding number.
previous_: The index of the previous page.
next_: The index of the next page.
next_clicked: The index of the current page.
local_results: A dict of outputs of EBM
used to write predictions and plot heatmaps on screen.
models: A dict of EBM models to predict the labels.
models_params: A dict of models' attributes, which will be
saved as a pickle file.
predictions: A pandas dataframe; each column contains EBM's
predictions of each label.
unlabeled_index: A pandas index of unlabeled rows. When new
labels are added to the database, compute_unlabeled_index()
needs to be called to track the changes.
"""
if _state.uploaded_files is not None:
data_file = _state.uploaded_files
_state.configs["db_filename"] = data_file.name
else:
data_file = _state.configs["db_filename"]
filename = _state.configs["db_filename"]
if filename == "None":
return
file_pre, file_ext = os.path.splitext(filename)
if file_ext == ".csv":
_state.database = pd.read_csv(data_file, index_col=0)
elif (file_ext == ".xlsx") or (file_ext == ".xls"):
_state.database = pd.read_excel(data_file, index_col=0)
create_pages()
def create_pages():
"""Add or change state parameters that are related to labeling pages
These parameters assign the labels to multiple pages, with
one label per page.
"""
_state["pages"] = _state.database.columns[-_state.num_labels:]
_state["classes"] = {
label: sorted(list(_state.database[label].dropna().unique()))
for label in _state.pages
}
_state["num_to_class"] = {
label: dict(enumerate(_state.classes[label]))
for label in _state.pages
}
_state["class_to_num"] = {
label: {c: i
for i, c in enumerate(_state.classes[label])}
for label in _state.pages
}
_state.update({
'local_results': {}
})
_state["predictions"] = pd.DataFrame(index=_state.database.index,
columns=_state.pages)
file_pre, file_ext = os.path.splitext(_state.configs["db_filename"])
try:
with open(file_pre + str(_state.num_labels) + _MODEL, 'rb') as _file:
_state["models_params"] = pkle.load(_file)
_state["models"] = {}
for label in _state.pages:
_state.models[label] = ExplainableBoostingClassifier()
_state.models[label].__dict__.update(
_state.models_params[label])
except FileNotFoundError:
_state["models"], _state["models_params"] = initialize_models()
compute_unlabeled_index()
def initialize_models():
"""initialize and train EBMs for all labels.
If a pickle file of EBM models (stored in _MODEL) is not found
in the directory, this function will be called by init_state_params()
to initialize the models.
"""
models = {}
models_params = {}
for label in _state.pages:
y = _state.database[label].dropna().map(_state.class_to_num[label])
X = subset_features(_state.database, label)
X = X.loc[y.index, :]
models[label] = ExplainableBoostingClassifier().fit(X, y)
models_params[label] = models[label].__dict__
return models, models_params
def subset_features(X, label):
"""Returns a subset of features specified in state's input_features parameters
Args:
X: a Pandas DataFrame.
label: The column name of the labels.
Returns
A Pandas DataFrame consisting of a subset of features in X.
"""
input_features = _state.configs["input_features"]
if label not in input_features.keys():
X = X.iloc[:, :-_state.num_labels]
else:
X = X.loc[:, input_features[label]]
return X
def compute_unlabeled_index(new_labeled_index=None, label=None):
"""Track the indices of unlabeled data after introducing new labels.
Args:
new_labeled_index: A pandas index of newly labeled data.
label: The column name of the new labels.
"""
if new_labeled_index is not None:
_state.unlabeled_index[label] = _state.unlabeled_index[
label].difference(new_labeled_index)
else:
all_index = _state.database.index
_state.unlabeled_index = {
label: all_index[_state.database[label].isna()]
for label in _state.pages
}
def create_config_file():
"""Create a new config file"""
_state["configs"] = {
"db_filename": "None",
"sidebar": {
"num_labels": 1,
"relabel": "Yes",
"mode": "Fixed sample size",
"n_samples": 50,
"threshold": 0.95
},
"input_features": {}
}
with open(_CONFIGS_FILE, "w") as _file:
json.dump(_state.configs, _file, indent=4)
@st.experimental_memo
def convert_to_downloadable(data, file_type):
"""Convert a dataframe to a downloadable format
Args:
data: A Pandas DataFrame.
file_type: A target format for the conversion: ".csv", ".xls" or ".xlsx".
Returns:
converted_data: Data converted to the specified format.
"""
if file_type == ".csv":
converted_data = data.to_csv().encode('utf-8')
mime = "text/csv"
elif (file_type == ".xlsx") or (file_type == ".xls"):
converted_data = data.to_excel().encode('utf-8')
mime = "application/vnd.ms-excel"
else:
raise ValueError("file_type must be \"csv\" or \"excel\"")
return converted_data, mime
def display_main_screen(label):
"""Display predictions and heatmaps on the main screen.
This function is called after EBM has been trained on the labeled data.
The predictions and explanations (displayed as heatmaps) will be shown
on the main screen.
Args:
label: the column name of the predictions.
"""
main_cols = st.columns((4, 4, 4))
if _state.unlabeled_index[label].empty:
main_cols[1].write("All " + label + " data are labeled.")
else:
with st.form(label + "Label form"):
if _state.local_results[label] == {}:
main_cols[1].write("""There are some unlabeled data left. \n \
This means that the confidences of the remaining data are \
above the threshold. \n You can either let the model label \
these data automatically \n or change the sampling mode to \
\"Fixed sample size\".""")
else:
input_features = _state.configs["input_features"]
if label not in input_features.keys():
num_features = _state.database.shape[1] - _state.num_labels
else:
num_features = len(input_features[label])
num_heatmap_rows = math.ceil(num_features / _NUM_FEAT_PER_ROW)
for page in _state.local_results[label]:
current_plot = plot_all_features(
_state.local_results[label][page]['data'],
title=str(page),
height=50,
num_rows=num_heatmap_rows)
cols = st.columns((6, 1))
#with cols[0]:
# if _state.text1 is not None:
# st.write(_state.data[_state.text1][page])
# if _state.text2 is not None:
# st.write(_state.data[_state.text2][page])
cols[0].altair_chart(current_plot,
use_container_width=True)
prediction = _state.local_results[label][page][
'prediction']
cols[1].radio("Label",
options=_state.classes[label],
key=label + str(page),
index=int(prediction))
results = report_results(page, label)
for result in results:
cols[1].write(result)
st.markdown("""---""")
label_from_cols = st.columns((4, 4, 4))
label_from_cols[1].radio("Automatically label the remaining data?",
("Yes", "No"),
index=1,
key=label+"_auto")
label_from_cols[1].form_submit_button("Submit Labels",
on_click=update_and_save,
args=(label, ))
@st.experimental_memo
def plot_all_features(data, title, height, num_rows):
"""Plot all rows of the heatmap of EBM's per-instance explanation.
Args:
data: Per-instance local explanations from EBM.
title: The plot's title.
height: The height of the plot.
num_rows: The number of rows of the heatmap.
Returns:
obj: An Altair plot object.
"""
plot_list = [None] * num_rows
if num_rows == 1:
plot_list[0] = plot(data, title, height)
else:
plot_list[0] = plot(data.iloc[0:_NUM_FEAT_PER_ROW], title, height)
for i in range(1, num_rows - 1):
plot_list[i] = plot(
data.iloc[_NUM_FEAT_PER_ROW * i:_NUM_FEAT_PER_ROW * (i + 1)],
"", height)
plot_list[-1] = plot(data.iloc[_NUM_FEAT_PER_ROW * (num_rows - 1):],
"", height)
obj = alt.vconcat(*plot_list).configure_axis(
labelFontSize=13, titleFontSize=16, labelAngle=0,
title=None).configure_title(fontSize=16)
return obj
def plot(data, title, height):
"""Plot each row of the heatmap of EBM's per-instance explanation.
Args:
data: Per-instance local explanations from EBM.
title: The plot's title.
height: The height of the plot.
Returns:
obj: An Altair plot object.
"""
base = alt.Chart(data).encode(x=alt.X('features', sort=None))
heatmap = base.mark_rect().encode(color=alt.Color(
'scores:Q',
scale=alt.Scale(scheme='redblue', reverse=True, domain=[0, 1]),
legend=alt.Legend(direction='vertical')))
# Configure text
text = base.mark_text(baseline='middle', fontSize=14).encode(
text='values:N',
color=alt.condition(
(alt.datum.scores > 0.8) | (alt.datum.scores < 0.2),
alt.value('white'), alt.value('black')))
obj = (heatmap + text).properties(height=height, width=550, title=title)
return obj
@st.experimental_memo
def report_results(idx, col_name):
"""Create a list that contains current label (if exists) and confidence score.
Args:
idx: A row's index in the database.
col_name: A column's name in the database.
Returns:
results: A list of current label (if exists) and confidence score.
"""
results = []
current_label = _state.database[col_name][idx]
if not pd.isna(current_label):
results.append(f"Current label: {current_label}")
confidence = _state.local_results[col_name][idx]['confidence']
results.append(f"Confidence: {confidence:.2f}")
return results
def sample_and_predict():
"""Sample data and make a dict of predictions and explanations.
This function calls EBM to predict the labels and give per-instance
local explanations. This function calls generate_explanation() to store
the predictions and explanations in a dictionary.
"""
st.experimental_memo.clear()
if _state.loaded_new_file:
init_state_params()
_state.loaded_new_file = False
else:
if "database" not in _state:
st.error("No database has been uploaded.")
return
if _state.configs["sidebar"]["num_labels"] != _state.num_labels:
create_pages()
_state.local_results = dict.fromkeys(_state.pages)
for label in _state.pages:
X = subset_features(_state.database, label)
if _state.relabel == "No":
X_unlabeled = X.loc[_state.unlabeled_index[label], :]
else:
X_unlabeled = X
_state.local_results[label] = {}
model = _state.models[label]
generate_explanation(X_unlabeled, label, model)
for k in _state.configs["sidebar"].keys():
_state.configs["sidebar"][k] = _state[k]
with open(_CONFIGS_FILE, "w") as _file:
json.dump(_state.configs, _file, indent=4)
def update_and_save(label):
"""Update the labels, then retrain and save the models.
Store the user's labels in the database, which is then saved to
a local disk. EBM is then retrained on the database with addition
labels, after which, a new list of predictions and explanations
will be shown on the main screen. This function calls
generate_explanation() to store the predictions and explanations
in a dictionary.
Args:
label: the column name of the label.
"""
new_labeled_index = list(_state.local_results[label].keys())
_state.database.loc[new_labeled_index, label] = [
_state[label + str(ix)] for ix in new_labeled_index
]
compute_unlabeled_index(new_labeled_index, label)
if _state[label + "_auto"] == "Yes":
unlabeled_idx = _state.unlabeled_index[label]
class_pred = _state.predictions.loc[unlabeled_idx, label]
_state.database.loc[unlabeled_idx, label] = class_pred
_state.unlabeled_index[label] = pd.Index([])
labeled_index = _state.database.index
else:
labeled_index = _state.database.index.difference(
_state.unlabeled_index[label])
X = subset_features(_state.database, label)
X_train = X.loc[labeled_index, :]
ytrain = _state.database.loc[labeled_index, label]
ebm = ExplainableBoostingClassifier()
ebm.fit(X_train, ytrain.map(_state.class_to_num[label]))
_state.models[label] = ebm
_state.models_params[label] = ebm.__dict__
filename = _state.configs["db_filename"]
file_pre, file_ext = os.path.splitext(filename)
with open(file_pre + str(_state.num_labels) + _MODEL, 'wb') as _file:
pkle.dump(_state.models_params, _file, protocol=pkle.HIGHEST_PROTOCOL)
_state.local_results[label] = {}
if _state[label + "_auto"] == "No":
X = X.loc[_state.unlabeled_index[label], :]
generate_explanation(X, label, ebm)
def generate_explanation(X, label, model):
"""Create a dict of predictions and explanations of a sample.
Make label predictions and per-instance local explanations,
which are then stored as a dict in _state.local_results.
Args:
X: A set of unlabeled data.
label: The column name of a label.
model: A model to predict labels and provide explanations.
"""
n_samples = X.shape[0]
n_features = X.shape[1]
localx = model.explain_local(X)._internal_obj['specific']
ypred = np.array([
_state.num_to_class[label][localx[j]['perf']['predicted']]
for j in range(n_samples)
])
_state.predictions.loc[X.index, label] = ypred
y = _state.database.loc[X.index, label]
p = np.array(
[localx[j]['perf']['predicted_score'] for j in range(n_samples)])
scores = np.minimum(p, (pd.isnull(y) | (ypred == y)))
if _state.mode == "Confidence threshold":
top_ind = np.where(scores <= _state.threshold)[0]
else:
n_samples = np.minimum(_state.n_samples, scores.shape[0] - 1)
top_ind = np.argpartition(scores, n_samples)[:n_samples]
X_ = X.iloc[top_ind, :].copy()
ypred = ypred[top_ind]
id_idx_pair = dict(zip(X_.index, top_ind))
try:
data_by_class = [X_[ypred == c] for c in _state.classes[label]]
except KeyError:
return
feature_names = X.columns
for sgn_data in data_by_class:
current_dict = _state.local_results[label]
for j in sgn_data.index:
localxi = localx[id_idx_pair[j]]
if len(_state.classes[label]) == 2:
feature_contrib = localxi['scores'][:n_features]
else:
feature_contrib = [
localxi['scores'][k][localxi['perf']['predicted']]
for k in range(n_features)
]
heatmap_data = pd.DataFrame({
'features':
feature_names,
'values':
localxi['values'][:n_features],
'scores':
1 / (1 + 1 / np.exp(feature_contrib))
})
heatmap_data = heatmap_data.astype({
'features': str,
'values': str,
'scores': float
})
current_dict[j] = {
'actual': localxi['perf']['actual'],
'prediction': localxi['perf']['predicted'],
'confidence': localxi['perf']['predicted_score'],
'data': heatmap_data
}
if __name__ == "__main__":
main()
|