[074d3d]: / mne / decoding / tests / test_search_light.py

Download this file

367 lines (310 with data), 12.6 kB

  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
# Authors: The MNE-Python contributors.
# License: BSD-3-Clause
# Copyright the MNE-Python contributors.
import platform
from inspect import signature
import numpy as np
import pytest
from numpy.testing import assert_array_equal, assert_equal
sklearn = pytest.importorskip("sklearn")
from sklearn.base import BaseEstimator, clone, is_classifier
from sklearn.discriminant_analysis import LinearDiscriminantAnalysis
from sklearn.ensemble import BaggingClassifier
from sklearn.linear_model import LinearRegression, LogisticRegression, Ridge
from sklearn.metrics import make_scorer, roc_auc_score
from sklearn.model_selection import cross_val_predict
from sklearn.multiclass import OneVsRestClassifier
from sklearn.pipeline import make_pipeline
from sklearn.svm import SVC
from sklearn.utils.estimator_checks import parametrize_with_checks
from mne.decoding.search_light import GeneralizingEstimator, SlidingEstimator
from mne.decoding.transformer import Vectorizer
from mne.utils import check_version, use_log_level
NEW_MULTICLASS_SAMPLE_WEIGHT = check_version("sklearn", "1.4")
def make_data():
"""Make data."""
n_epochs, n_chan, n_time = 50, 32, 10
X = np.random.rand(n_epochs, n_chan, n_time)
y = np.arange(n_epochs) % 2
for ii in range(n_time):
coef = np.random.randn(n_chan)
X[y == 0, :, ii] += coef
X[y == 1, :, ii] -= coef
return X, y
def test_search_light_basic():
"""Test SlidingEstimator."""
# https://github.com/scikit-learn/scikit-learn/issues/27711
if platform.system() == "Windows" and check_version("numpy", "2.0.0.dev0"):
pytest.skip("sklearn int_t / long long mismatch")
logreg = OneVsRestClassifier(LogisticRegression(solver="liblinear", random_state=0))
X, y = make_data()
n_epochs, _, n_time = X.shape
# init
sl = SlidingEstimator("foo")
with pytest.raises(ValueError, match="must be"):
sl.fit(X, y)
sl = SlidingEstimator(Ridge())
assert not is_classifier(sl)
sl = SlidingEstimator(LogisticRegression(solver="liblinear"))
assert is_classifier(sl.base_estimator)
assert is_classifier(sl)
# fit
assert_equal(sl.__repr__()[:18], "<SlidingEstimator(")
sl.fit(X, y)
assert_equal(sl.__repr__()[-28:], ", fitted with 10 estimators>")
pytest.raises(ValueError, sl.fit, X[1:], y)
pytest.raises(ValueError, sl.fit, X[:, :, 0], y)
sl.fit(X, y, sample_weight=np.ones_like(y))
# transforms
pytest.raises(ValueError, sl.predict, X[:, :, :2])
y_trans = sl.transform(X)
assert X.dtype == float
assert y_trans.dtype == float
y_pred = sl.predict(X)
assert y_pred.dtype == np.dtype(int)
assert_array_equal(y_pred.shape, [n_epochs, n_time])
y_proba = sl.predict_proba(X)
assert y_proba.dtype == np.dtype(float)
assert_array_equal(y_proba.shape, [n_epochs, n_time, 2])
# score
score = sl.score(X, y)
assert_array_equal(score.shape, [n_time])
assert np.sum(np.abs(score)) != 0
assert score.dtype == np.dtype(float)
sl = SlidingEstimator(logreg)
assert_equal(sl.scoring, None)
# Scoring method
for scoring in ["foo", 999]:
sl = SlidingEstimator(logreg, scoring=scoring)
sl.fit(X, y)
pytest.raises((ValueError, TypeError), sl.score, X, y)
# Check sklearn's roc_auc fix: scikit-learn/scikit-learn#6874
# -- 3 class problem
sl = SlidingEstimator(logreg, scoring="roc_auc")
y = np.arange(len(X)) % 3
sl.fit(X, y)
with pytest.raises(ValueError, match="for two-class"):
sl.score(X, y)
# But check that valid ones should work with new enough sklearn
kwargs = dict()
if check_version("sklearn", "1.4"):
kwargs["response_method"] = "predict_proba"
else:
kwargs["needs_proba"] = True
if "multi_class" in signature(roc_auc_score).parameters:
scoring = make_scorer(roc_auc_score, multi_class="ovo", **kwargs)
sl = SlidingEstimator(logreg, scoring=scoring)
sl.fit(X, y)
sl.score(X, y) # smoke test
# -- 2 class problem not in [0, 1]
y = np.arange(len(X)) % 2 + 1
sl.fit(X, y)
score = sl.score(X, y)
assert_array_equal(
score,
[roc_auc_score(y - 1, _y_pred - 1) for _y_pred in sl.decision_function(X).T],
)
y = np.arange(len(X)) % 2
# Cannot pass a metric as a scoring parameter
sl1 = SlidingEstimator(logreg, scoring=roc_auc_score)
sl1.fit(X, y)
pytest.raises(ValueError, sl1.score, X, y)
# Now use string as scoring
sl1 = SlidingEstimator(logreg, scoring="roc_auc")
sl1.fit(X, y)
rng = np.random.RandomState(0)
X = rng.randn(*X.shape) # randomize X to avoid AUCs in [0, 1]
score_sl = sl1.score(X, y)
assert_array_equal(score_sl.shape, [n_time])
assert score_sl.dtype == np.dtype(float)
# Check that scoring was applied adequately
kwargs = dict()
if check_version("sklearn", "1.4"):
kwargs["response_method"] = ("decision_function", "predict_proba")
else:
kwargs["needs_threshold"] = True
scoring = make_scorer(roc_auc_score, **kwargs)
score_manual = [
scoring(est, x, y) for est, x in zip(sl1.estimators_, X.transpose(2, 0, 1))
]
assert_array_equal(score_manual, score_sl)
# n_jobs
sl = SlidingEstimator(logreg, n_jobs=None, scoring="roc_auc")
score_1job = sl.fit(X, y).score(X, y)
sl.n_jobs = 2
score_njobs = sl.fit(X, y).score(X, y)
assert_array_equal(score_1job, score_njobs)
sl.predict(X)
# n_jobs > n_estimators
sl.fit(X[..., [0]], y)
sl.predict(X[..., [0]])
# pipeline
class _LogRegTransformer(LogisticRegression):
def transform(self, X):
return super().predict_proba(X)[..., 1]
logreg_transformer = OneVsRestClassifier(_LogRegTransformer(random_state=0))
pipe = make_pipeline(SlidingEstimator(logreg_transformer), logreg)
pipe.fit(X, y)
pipe.predict(X)
# n-dimensional feature space
X = np.random.rand(10, 3, 4, 2)
y = np.arange(10) % 2
y_preds = list()
for n_jobs in [1, 2]:
pipe = SlidingEstimator(make_pipeline(Vectorizer(), logreg), n_jobs=n_jobs)
y_preds.append(pipe.fit(X, y).predict(X))
features_shape = pipe.estimators_[0].steps[0][1].features_shape_
assert_array_equal(features_shape, [3, 4])
assert_array_equal(y_preds[0], y_preds[1])
# Bagging classifiers
X = np.random.rand(10, 3, 4)
for n_jobs in (1, 2):
pipe = SlidingEstimator(BaggingClassifier(None, 2), n_jobs=n_jobs)
pipe.fit(X, y)
pipe.score(X, y)
assert isinstance(pipe.estimators_[0], BaggingClassifier)
@pytest.fixture()
def metadata_routing():
"""Temporarily enable metadata routing for new sklearn."""
if NEW_MULTICLASS_SAMPLE_WEIGHT:
sklearn.set_config(enable_metadata_routing=True)
yield
if NEW_MULTICLASS_SAMPLE_WEIGHT:
sklearn.set_config(enable_metadata_routing=False)
def test_generalization_light(metadata_routing):
"""Test GeneralizingEstimator."""
if NEW_MULTICLASS_SAMPLE_WEIGHT:
clf = LogisticRegression(random_state=0)
clf.set_fit_request(sample_weight=True)
logreg = OneVsRestClassifier(clf)
else:
logreg = LogisticRegression(
solver="liblinear",
random_state=0,
multi_class="ovr",
)
X, y = make_data()
n_epochs, _, n_time = X.shape
# fit
gl = GeneralizingEstimator(logreg)
assert_equal(repr(gl)[:23], "<GeneralizingEstimator(")
gl.fit(X, y)
gl.fit(X, y, sample_weight=np.ones_like(y))
assert_equal(gl.__repr__()[-28:], ", fitted with 10 estimators>")
# transforms
y_pred = gl.predict(X)
assert_array_equal(y_pred.shape, [n_epochs, n_time, n_time])
assert y_pred.dtype == np.dtype(int)
y_proba = gl.predict_proba(X)
assert y_proba.dtype == np.dtype(float)
assert_array_equal(y_proba.shape, [n_epochs, n_time, n_time, 2])
# transform to different datasize
y_pred = gl.predict(X[:, :, :2])
assert_array_equal(y_pred.shape, [n_epochs, n_time, 2])
# score
score = gl.score(X[:, :, :3], y)
assert_array_equal(score.shape, [n_time, 3])
assert np.sum(np.abs(score)) != 0
assert score.dtype == np.dtype(float)
gl = GeneralizingEstimator(logreg, scoring="roc_auc")
gl.fit(X, y)
score = gl.score(X, y)
auc = roc_auc_score(y, gl.estimators_[0].predict_proba(X[..., 0])[..., 1])
assert_equal(score[0, 0], auc)
for scoring in ["foo", 999]:
gl = GeneralizingEstimator(logreg, scoring=scoring)
gl.fit(X, y)
pytest.raises((ValueError, TypeError), gl.score, X, y)
# Check sklearn's roc_auc fix: scikit-learn/scikit-learn#6874
# -- 3 class problem
gl = GeneralizingEstimator(logreg, scoring="roc_auc")
y = np.arange(len(X)) % 3
gl.fit(X, y)
pytest.raises(ValueError, gl.score, X, y)
# -- 2 class problem not in [0, 1]
y = np.arange(len(X)) % 2 + 1
gl.fit(X, y)
score = gl.score(X, y)
manual_score = [
[roc_auc_score(y - 1, _y_pred) for _y_pred in _y_preds]
for _y_preds in gl.decision_function(X).transpose(1, 2, 0)
]
assert_array_equal(score, manual_score)
# n_jobs
gl = GeneralizingEstimator(logreg, n_jobs=2)
gl.fit(X, y)
y_pred = gl.predict(X)
assert_array_equal(y_pred.shape, [n_epochs, n_time, n_time])
score = gl.score(X, y)
assert_array_equal(score.shape, [n_time, n_time])
# n_jobs > n_estimators
gl.fit(X[..., [0]], y)
gl.predict(X[..., [0]])
# n-dimensional feature space
X = np.random.rand(10, 3, 4, 2)
y = np.arange(10) % 2
y_preds = list()
for n_jobs in [1, 2]:
pipe = GeneralizingEstimator(make_pipeline(Vectorizer(), logreg), n_jobs=n_jobs)
y_preds.append(pipe.fit(X, y).predict(X))
features_shape = pipe.estimators_[0].steps[0][1].features_shape_
assert_array_equal(features_shape, [3, 4])
assert_array_equal(y_preds[0], y_preds[1])
@pytest.mark.parametrize(
"n_jobs, verbose", [(1, False), (2, False), (1, True), (2, "info")]
)
def test_verbose_arg(capsys, n_jobs, verbose):
"""Test controlling output with the ``verbose`` argument."""
X, y = make_data()
clf = SVC()
# shows progress bar and prints other messages to the console
with use_log_level(True):
for estimator_object in [SlidingEstimator, GeneralizingEstimator]:
estimator = estimator_object(clf, n_jobs=n_jobs, verbose=verbose)
estimator = estimator.fit(X, y)
estimator.score(X, y)
estimator.predict(X)
stdout, stderr = capsys.readouterr()
if isinstance(verbose, bool) and not verbose:
assert all(channel == "" for channel in (stdout, stderr))
else:
assert any(len(channel) > 0 for channel in (stdout, stderr))
def test_cross_val_predict():
"""Test cross_val_predict with predict_proba."""
rng = np.random.RandomState(42)
X = rng.randn(10, 1, 3)
y = rng.randint(0, 2, 10)
estimator = SlidingEstimator(LinearRegression())
cross_val_predict(estimator, X, y, cv=2)
class Classifier(BaseEstimator):
"""Moch class that does not have classes_ attribute."""
def __init__(self):
self.base_estimator = LinearDiscriminantAnalysis()
def fit(self, X, y):
self.estimator_ = clone(self.base_estimator).fit(X, y)
return self
def predict_proba(self, X):
return self.estimator_.predict_proba(X)
with pytest.raises(AttributeError, match="classes_ attribute"):
estimator = SlidingEstimator(Classifier())
cross_val_predict(estimator, X, y, method="predict_proba", cv=2)
estimator = SlidingEstimator(LinearDiscriminantAnalysis())
cross_val_predict(estimator, X, y, method="predict_proba", cv=2)
@pytest.mark.slowtest
@parametrize_with_checks(
[
SlidingEstimator(LogisticRegression(), allow_2d=True),
GeneralizingEstimator(LogisticRegression(), allow_2d=True),
]
)
def test_sklearn_compliance(estimator, check):
"""Test LinearModel compliance with sklearn."""
ignores = (
# TODO: we don't handle singleton right (probably)
"check_classifiers_one_label_sample_weights",
"check_classifiers_classes",
"check_classifiers_train",
)
if any(ignore in str(check) for ignore in ignores):
return
check(estimator)