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
|
from __future__ import annotations
from collections.abc import Callable
from typing import Any
from typing import Type
import numpy as np
import pytest
import optuna
from optuna import samplers
from optuna.exceptions import ExperimentalWarning
from optuna.importance import BaseImportanceEvaluator
from optuna.importance import FanovaImportanceEvaluator
from optuna.importance import get_param_importances
from optuna.importance import MeanDecreaseImpurityImportanceEvaluator
from optuna.samplers import RandomSampler
from optuna.study import create_study
from optuna.testing.objectives import pruned_objective
from optuna.testing.storages import STORAGE_MODES
from optuna.testing.storages import StorageSupplier
from optuna.trial import Trial
evaluators: list[Type[BaseImportanceEvaluator]] = [
MeanDecreaseImpurityImportanceEvaluator,
FanovaImportanceEvaluator,
]
parametrize_evaluator = pytest.mark.parametrize("evaluator_init_func", evaluators)
@parametrize_evaluator
@pytest.mark.parametrize("storage_mode", STORAGE_MODES)
def test_get_param_importance_target_is_none_and_study_is_multi_obj(
storage_mode: str,
evaluator_init_func: Callable[[], BaseImportanceEvaluator],
) -> None:
def objective(trial: Trial) -> tuple[float, float]:
x1 = trial.suggest_float("x1", 0.1, 3)
x2 = trial.suggest_float("x2", 0.1, 3, log=True)
x3 = trial.suggest_float("x3", 0, 3, step=1)
x4 = trial.suggest_int("x4", -3, 3)
x5 = trial.suggest_int("x5", 1, 5, log=True)
x6 = trial.suggest_categorical("x6", [1.0, 1.1, 1.2])
if trial.number % 2 == 0:
# Conditional parameters are ignored unless `params` is specified and is not `None`.
x7 = trial.suggest_float("x7", 0.1, 3)
value = x1**4 + x2 + x3 - x4**2 - x5 + x6
if trial.number % 2 == 0:
value += x7
return value, 0.0
with StorageSupplier(storage_mode) as storage:
study = create_study(directions=["minimize", "minimize"], storage=storage)
study.optimize(objective, n_trials=3)
with pytest.raises(ValueError):
get_param_importances(study, evaluator=evaluator_init_func())
@parametrize_evaluator
@pytest.mark.parametrize("storage_mode", STORAGE_MODES)
@pytest.mark.parametrize("normalize", [True, False])
def test_get_param_importances(
storage_mode: str, evaluator_init_func: Callable[[], BaseImportanceEvaluator], normalize: bool
) -> None:
def objective(trial: Trial) -> float:
x1 = trial.suggest_float("x1", 0.1, 3)
x2 = trial.suggest_float("x2", 0.1, 3, log=True)
x3 = trial.suggest_float("x3", 0, 3, step=1)
x4 = trial.suggest_int("x4", -3, 3)
x5 = trial.suggest_int("x5", 1, 5, log=True)
x6 = trial.suggest_categorical("x6", [1.0, 1.1, 1.2])
if trial.number % 2 == 0:
# Conditional parameters are ignored unless `params` is specified and is not `None`.
x7 = trial.suggest_float("x7", 0.1, 3)
value = x1**4 + x2 + x3 - x4**2 - x5 + x6
if trial.number % 2 == 0:
value += x7
return value
with StorageSupplier(storage_mode) as storage:
study = create_study(storage=storage, sampler=samplers.RandomSampler())
study.optimize(objective, n_trials=3)
param_importance = get_param_importances(
study, evaluator=evaluator_init_func(), normalize=normalize
)
assert isinstance(param_importance, dict)
assert len(param_importance) == 6
assert all(
param_name in param_importance for param_name in ["x1", "x2", "x3", "x4", "x5", "x6"]
)
prev_importance = float("inf")
for param_name, importance in param_importance.items():
assert isinstance(param_name, str)
assert isinstance(importance, float)
assert importance <= prev_importance
prev_importance = importance
# Sanity check for param importances
assert all(0 <= x < float("inf") for x in param_importance.values())
if normalize:
assert np.isclose(sum(param_importance.values()), 1.0)
@parametrize_evaluator
@pytest.mark.parametrize("storage_mode", STORAGE_MODES)
@pytest.mark.parametrize("params", [[], ["x1"], ["x1", "x3"], ["x1", "x4"]])
@pytest.mark.parametrize("normalize", [True, False])
def test_get_param_importances_with_params(
storage_mode: str,
params: list[str],
evaluator_init_func: Callable[[], BaseImportanceEvaluator],
normalize: bool,
) -> None:
def objective(trial: Trial) -> float:
x1 = trial.suggest_float("x1", 0.1, 3)
x2 = trial.suggest_float("x2", 0.1, 3, log=True)
x3 = trial.suggest_float("x3", 0, 3, step=1)
if trial.number % 2 == 0:
x4 = trial.suggest_float("x4", 0.1, 3)
value = x1**4 + x2 + x3
if trial.number % 2 == 0:
value += x4
return value
with StorageSupplier(storage_mode) as storage:
study = create_study(storage=storage)
study.optimize(objective, n_trials=10)
param_importance = get_param_importances(
study, evaluator=evaluator_init_func(), params=params, normalize=normalize
)
assert isinstance(param_importance, dict)
assert len(param_importance) == len(params)
assert all(param in param_importance for param in params)
for param_name, importance in param_importance.items():
assert isinstance(param_name, str)
assert isinstance(importance, float)
# Sanity check for param importances
assert all(0 <= x < float("inf") for x in param_importance.values())
if normalize:
assert len(param_importance) == 0 or np.isclose(sum(param_importance.values()), 1.0)
def test_get_param_importances_unnormalized_experimental() -> None:
def objective(trial: Trial) -> float:
x1 = trial.suggest_float("x1", 0.1, 3)
return x1**2
study = create_study()
study.optimize(objective, n_trials=4)
with pytest.warns(ExperimentalWarning):
get_param_importances(study, normalize=False)
@parametrize_evaluator
@pytest.mark.parametrize("storage_mode", STORAGE_MODES)
@pytest.mark.parametrize("normalize", [True, False])
def test_get_param_importances_with_target(
storage_mode: str, evaluator_init_func: Callable[[], BaseImportanceEvaluator], normalize: bool
) -> None:
def objective(trial: Trial) -> float:
x1 = trial.suggest_float("x1", 0.1, 3)
x2 = trial.suggest_float("x2", 0.1, 3, log=True)
x3 = trial.suggest_float("x3", 0, 3, step=1)
if trial.number % 2 == 0:
x4 = trial.suggest_float("x4", 0.1, 3)
value = x1**4 + x2 + x3
if trial.number % 2 == 0:
value += x4
return value
with StorageSupplier(storage_mode) as storage:
study = create_study(storage=storage)
study.optimize(objective, n_trials=3)
param_importance = get_param_importances(
study,
evaluator=evaluator_init_func(),
target=lambda t: t.params["x1"] + t.params["x2"],
normalize=normalize,
)
assert isinstance(param_importance, dict)
assert len(param_importance) == 3
assert all(param_name in param_importance for param_name in ["x1", "x2", "x3"])
prev_importance = float("inf")
for param_name, importance in param_importance.items():
assert isinstance(param_name, str)
assert isinstance(importance, float)
assert importance <= prev_importance
prev_importance = importance
# Sanity check for param importances
assert all(0 <= x < float("inf") for x in param_importance.values())
if normalize:
assert np.isclose(sum(param_importance.values()), 1.0)
@parametrize_evaluator
def test_get_param_importances_invalid_empty_study(
evaluator_init_func: Callable[[], BaseImportanceEvaluator]
) -> None:
study = create_study()
with pytest.raises(ValueError):
get_param_importances(study, evaluator=evaluator_init_func())
study.optimize(pruned_objective, n_trials=3)
with pytest.raises(ValueError):
get_param_importances(study, evaluator=evaluator_init_func())
@parametrize_evaluator
def test_get_param_importances_invalid_single_trial(
evaluator_init_func: Callable[[], BaseImportanceEvaluator]
) -> None:
def objective(trial: Trial) -> float:
x1 = trial.suggest_float("x1", 0.1, 3)
return x1**2
study = create_study()
study.optimize(objective, n_trials=1)
with pytest.raises(ValueError):
get_param_importances(study, evaluator=evaluator_init_func())
@parametrize_evaluator
def test_get_param_importances_invalid_no_completed_trials_params(
evaluator_init_func: Callable[[], BaseImportanceEvaluator]
) -> None:
def objective(trial: Trial) -> float:
x1 = trial.suggest_float("x1", 0.1, 3)
if trial.number % 2 == 0:
_ = trial.suggest_float("x2", 0.1, 3, log=True)
raise optuna.TrialPruned
return x1**2
study = create_study()
study.optimize(objective, n_trials=3)
# None of the trials with `x2` are completed.
with pytest.raises(ValueError):
get_param_importances(study, evaluator=evaluator_init_func(), params=["x2"])
# None of the trials with `x2` are completed. Adding "x1" should not matter.
with pytest.raises(ValueError):
get_param_importances(study, evaluator=evaluator_init_func(), params=["x1", "x2"])
# None of the trials contain `x3`.
with pytest.raises(ValueError):
get_param_importances(study, evaluator=evaluator_init_func(), params=["x3"])
@parametrize_evaluator
def test_get_param_importances_invalid_dynamic_search_space_params(
evaluator_init_func: Callable[[], BaseImportanceEvaluator]
) -> None:
def objective(trial: Trial) -> float:
x1 = trial.suggest_float("x1", 0.1, trial.number + 0.1)
return x1**2
study = create_study()
study.optimize(objective, n_trials=3)
with pytest.raises(ValueError):
get_param_importances(study, evaluator=evaluator_init_func(), params=["x1"])
@parametrize_evaluator
def test_get_param_importances_empty_search_space(
evaluator_init_func: Callable[[], BaseImportanceEvaluator]
) -> None:
def objective(trial: Trial) -> float:
x = trial.suggest_float("x", 0, 5)
y = trial.suggest_float("y", 1, 1)
return 4 * x**2 + 4 * y**2
study = create_study()
study.optimize(objective, n_trials=3)
param_importance = get_param_importances(study, evaluator=evaluator_init_func())
assert len(param_importance) == 2
assert all([param in param_importance for param in ["x", "y"]])
assert param_importance["x"] > 0.0
assert param_importance["y"] == 0.0
@parametrize_evaluator
def test_importance_evaluator_seed(evaluator_init_func: Any) -> None:
def objective(trial: Trial) -> float:
x1 = trial.suggest_float("x1", 0.1, 3)
x2 = trial.suggest_float("x2", 0.1, 3, log=True)
x3 = trial.suggest_float("x3", 2, 4, log=True)
return x1 + x2 * x3
study = create_study(sampler=RandomSampler(seed=0))
study.optimize(objective, n_trials=3)
evaluator = evaluator_init_func(seed=2)
param_importance = evaluator.evaluate(study)
evaluator = evaluator_init_func(seed=2)
param_importance_same_seed = evaluator.evaluate(study)
assert param_importance == param_importance_same_seed
evaluator = evaluator_init_func(seed=3)
param_importance_different_seed = evaluator.evaluate(study)
assert param_importance != param_importance_different_seed
@parametrize_evaluator
def test_importance_evaluator_with_target(evaluator_init_func: Any) -> None:
def objective(trial: Trial) -> float:
x1 = trial.suggest_float("x1", 0.1, 3)
x2 = trial.suggest_float("x2", 0.1, 3, log=True)
x3 = trial.suggest_float("x3", 2, 4, log=True)
return x1 + x2 * x3
# Assumes that `seed` can be fixed to reproduce identical results.
study = create_study(sampler=RandomSampler(seed=0))
study.optimize(objective, n_trials=3)
evaluator = evaluator_init_func(seed=0)
param_importance = evaluator.evaluate(study)
param_importance_with_target = evaluator.evaluate(
study,
target=lambda t: t.params["x1"] + t.params["x2"],
)
assert param_importance != param_importance_with_target
|
