1
|
|
|
# Author: Simon Blanke |
2
|
|
|
# Email: [email protected] |
3
|
|
|
# License: MIT License |
4
|
|
|
|
5
|
|
|
from sklearn.datasets import load_breast_cancer |
6
|
|
|
from sklearn.model_selection import cross_val_score |
7
|
|
|
from sklearn.tree import DecisionTreeClassifier |
8
|
|
|
from hyperactive import Hyperactive |
9
|
|
|
|
10
|
|
|
data = load_breast_cancer() |
11
|
|
|
X = data.data |
12
|
|
|
y = data.target |
13
|
|
|
|
14
|
|
|
random_state = 1 |
15
|
|
|
n_iter_min = 0 |
16
|
|
|
n_iter_max = 100 |
17
|
|
|
|
18
|
|
|
|
19
|
|
View Code Duplication |
def model(para, X_train, y_train): |
|
|
|
|
20
|
|
|
model = DecisionTreeClassifier( |
21
|
|
|
criterion=para["criterion"], |
22
|
|
|
max_depth=para["max_depth"], |
23
|
|
|
min_samples_split=para["min_samples_split"], |
24
|
|
|
min_samples_leaf=para["min_samples_leaf"], |
25
|
|
|
) |
26
|
|
|
scores = cross_val_score(model, X_train, y_train, cv=2) |
27
|
|
|
|
28
|
|
|
return scores.mean() |
29
|
|
|
|
30
|
|
|
|
31
|
|
|
search_config = { |
32
|
|
|
model: { |
33
|
|
|
"criterion": ["gini", "entropy"], |
34
|
|
|
"max_depth": range(1, 21), |
35
|
|
|
"min_samples_split": range(2, 21), |
36
|
|
|
"min_samples_leaf": range(1, 21), |
37
|
|
|
} |
38
|
|
|
} |
39
|
|
|
|
40
|
|
|
warm_start = {model: {"max_depth": [1]}} |
41
|
|
|
|
42
|
|
|
|
43
|
|
View Code Duplication |
def test_HillClimbing(): |
|
|
|
|
44
|
|
|
opt0 = Hyperactive( |
45
|
|
|
search_config, |
46
|
|
|
optimizer="HillClimbing", |
47
|
|
|
n_iter=n_iter_min, |
48
|
|
|
random_state=random_state, |
49
|
|
|
warm_start=warm_start, |
50
|
|
|
) |
51
|
|
|
opt0.search(X, y) |
52
|
|
|
|
53
|
|
|
opt1 = Hyperactive( |
54
|
|
|
search_config, |
55
|
|
|
optimizer="HillClimbing", |
56
|
|
|
n_iter=n_iter_max, |
57
|
|
|
random_state=random_state, |
58
|
|
|
warm_start=warm_start, |
59
|
|
|
) |
60
|
|
|
opt1.search(X, y) |
61
|
|
|
|
62
|
|
|
assert opt0._optimizer_.score_best < opt1._optimizer_.score_best |
63
|
|
|
|
64
|
|
|
|
65
|
|
View Code Duplication |
def test_StochasticHillClimbing(): |
|
|
|
|
66
|
|
|
opt0 = Hyperactive( |
67
|
|
|
search_config, |
68
|
|
|
optimizer="StochasticHillClimbing", |
69
|
|
|
n_iter=n_iter_min, |
70
|
|
|
random_state=random_state, |
71
|
|
|
warm_start=warm_start, |
72
|
|
|
) |
73
|
|
|
opt0.search(X, y) |
74
|
|
|
|
75
|
|
|
opt1 = Hyperactive( |
76
|
|
|
search_config, |
77
|
|
|
optimizer="StochasticHillClimbing", |
78
|
|
|
n_iter=n_iter_max, |
79
|
|
|
random_state=random_state, |
80
|
|
|
warm_start=warm_start, |
81
|
|
|
) |
82
|
|
|
opt1.search(X, y) |
83
|
|
|
|
84
|
|
|
assert opt0._optimizer_.score_best < opt1._optimizer_.score_best |
85
|
|
|
|
86
|
|
|
|
87
|
|
View Code Duplication |
def test_TabuOptimizer(): |
|
|
|
|
88
|
|
|
opt0 = Hyperactive( |
89
|
|
|
search_config, |
90
|
|
|
optimizer="TabuSearch", |
91
|
|
|
n_iter=n_iter_min, |
92
|
|
|
random_state=random_state, |
93
|
|
|
warm_start=warm_start, |
94
|
|
|
) |
95
|
|
|
opt0.search(X, y) |
96
|
|
|
|
97
|
|
|
opt1 = Hyperactive( |
98
|
|
|
search_config, |
99
|
|
|
optimizer="TabuSearch", |
100
|
|
|
n_iter=n_iter_max, |
101
|
|
|
random_state=random_state, |
102
|
|
|
warm_start=warm_start, |
103
|
|
|
) |
104
|
|
|
opt1.search(X, y) |
105
|
|
|
|
106
|
|
|
assert opt0._optimizer_.score_best < opt1._optimizer_.score_best |
107
|
|
|
|
108
|
|
|
|
109
|
|
View Code Duplication |
def test_RandomSearch(): |
|
|
|
|
110
|
|
|
opt0 = Hyperactive( |
111
|
|
|
search_config, |
112
|
|
|
optimizer="RandomSearch", |
113
|
|
|
n_iter=n_iter_min, |
114
|
|
|
random_state=random_state, |
115
|
|
|
warm_start=warm_start, |
116
|
|
|
) |
117
|
|
|
opt0.search(X, y) |
118
|
|
|
|
119
|
|
|
opt1 = Hyperactive( |
120
|
|
|
search_config, |
121
|
|
|
optimizer="RandomSearch", |
122
|
|
|
n_iter=n_iter_max, |
123
|
|
|
random_state=random_state, |
124
|
|
|
warm_start=warm_start, |
125
|
|
|
) |
126
|
|
|
opt1.search(X, y) |
127
|
|
|
|
128
|
|
|
assert opt0._optimizer_.score_best < opt1._optimizer_.score_best |
129
|
|
|
|
130
|
|
|
|
131
|
|
View Code Duplication |
def test_RandomRestartHillClimbing(): |
|
|
|
|
132
|
|
|
opt0 = Hyperactive( |
133
|
|
|
search_config, |
134
|
|
|
optimizer="RandomRestartHillClimbing", |
135
|
|
|
n_iter=n_iter_min, |
136
|
|
|
random_state=random_state, |
137
|
|
|
warm_start=warm_start, |
138
|
|
|
) |
139
|
|
|
opt0.search(X, y) |
140
|
|
|
|
141
|
|
|
opt1 = Hyperactive( |
142
|
|
|
search_config, |
143
|
|
|
optimizer="RandomRestartHillClimbing", |
144
|
|
|
n_iter=n_iter_max, |
145
|
|
|
random_state=random_state, |
146
|
|
|
warm_start=warm_start, |
147
|
|
|
) |
148
|
|
|
opt1.search(X, y) |
149
|
|
|
|
150
|
|
|
assert opt0._optimizer_.score_best < opt1._optimizer_.score_best |
151
|
|
|
|
152
|
|
|
|
153
|
|
View Code Duplication |
def test_RandomAnnealing(): |
|
|
|
|
154
|
|
|
opt0 = Hyperactive( |
155
|
|
|
search_config, |
156
|
|
|
optimizer="RandomAnnealing", |
157
|
|
|
n_iter=n_iter_min, |
158
|
|
|
random_state=random_state, |
159
|
|
|
warm_start=warm_start, |
160
|
|
|
) |
161
|
|
|
opt0.search(X, y) |
162
|
|
|
|
163
|
|
|
opt1 = Hyperactive( |
164
|
|
|
search_config, |
165
|
|
|
optimizer="RandomAnnealing", |
166
|
|
|
n_iter=n_iter_max, |
167
|
|
|
random_state=random_state, |
168
|
|
|
warm_start=warm_start, |
169
|
|
|
) |
170
|
|
|
opt1.search(X, y) |
171
|
|
|
|
172
|
|
|
assert opt0._optimizer_.score_best < opt1._optimizer_.score_best |
173
|
|
|
|
174
|
|
|
|
175
|
|
View Code Duplication |
def test_SimulatedAnnealing(): |
|
|
|
|
176
|
|
|
opt0 = Hyperactive( |
177
|
|
|
search_config, |
178
|
|
|
optimizer="SimulatedAnnealing", |
179
|
|
|
n_iter=n_iter_min, |
180
|
|
|
random_state=random_state, |
181
|
|
|
warm_start=warm_start, |
182
|
|
|
) |
183
|
|
|
opt0.search(X, y) |
184
|
|
|
|
185
|
|
|
opt1 = Hyperactive( |
186
|
|
|
search_config, |
187
|
|
|
optimizer="SimulatedAnnealing", |
188
|
|
|
n_iter=n_iter_max, |
189
|
|
|
random_state=random_state, |
190
|
|
|
warm_start=warm_start, |
191
|
|
|
) |
192
|
|
|
opt1.search(X, y) |
193
|
|
|
|
194
|
|
|
assert opt0._optimizer_.score_best < opt1._optimizer_.score_best |
195
|
|
|
|
196
|
|
|
|
197
|
|
View Code Duplication |
def test_StochasticTunneling(): |
|
|
|
|
198
|
|
|
opt0 = Hyperactive( |
199
|
|
|
search_config, |
200
|
|
|
optimizer="StochasticTunneling", |
201
|
|
|
n_iter=n_iter_min, |
202
|
|
|
random_state=random_state, |
203
|
|
|
warm_start=warm_start, |
204
|
|
|
) |
205
|
|
|
opt0.search(X, y) |
206
|
|
|
|
207
|
|
|
opt1 = Hyperactive( |
208
|
|
|
search_config, |
209
|
|
|
optimizer="StochasticTunneling", |
210
|
|
|
n_iter=n_iter_max, |
211
|
|
|
random_state=random_state, |
212
|
|
|
warm_start=warm_start, |
213
|
|
|
) |
214
|
|
|
opt1.search(X, y) |
215
|
|
|
|
216
|
|
|
assert opt0._optimizer_.score_best < opt1._optimizer_.score_best |
217
|
|
|
|
218
|
|
|
|
219
|
|
View Code Duplication |
def test_ParallelTempering(): |
|
|
|
|
220
|
|
|
opt0 = Hyperactive( |
221
|
|
|
search_config, |
222
|
|
|
optimizer="ParallelTempering", |
223
|
|
|
n_iter=n_iter_min, |
224
|
|
|
random_state=random_state, |
225
|
|
|
warm_start=warm_start, |
226
|
|
|
) |
227
|
|
|
opt0.search(X, y) |
228
|
|
|
|
229
|
|
|
opt1 = Hyperactive( |
230
|
|
|
search_config, |
231
|
|
|
optimizer="ParallelTempering", |
232
|
|
|
n_iter=n_iter_max, |
233
|
|
|
random_state=random_state, |
234
|
|
|
warm_start=warm_start, |
235
|
|
|
) |
236
|
|
|
opt1.search(X, y) |
237
|
|
|
|
238
|
|
|
assert opt0._optimizer_.score_best < opt1._optimizer_.score_best |
239
|
|
|
|
240
|
|
|
|
241
|
|
View Code Duplication |
def test_ParticleSwarm(): |
|
|
|
|
242
|
|
|
opt0 = Hyperactive( |
243
|
|
|
search_config, |
244
|
|
|
optimizer="ParticleSwarm", |
245
|
|
|
n_iter=n_iter_min, |
246
|
|
|
random_state=random_state, |
247
|
|
|
warm_start=warm_start, |
248
|
|
|
) |
249
|
|
|
opt0.search(X, y) |
250
|
|
|
|
251
|
|
|
opt1 = Hyperactive( |
252
|
|
|
search_config, |
253
|
|
|
optimizer="ParticleSwarm", |
254
|
|
|
n_iter=n_iter_max, |
255
|
|
|
random_state=random_state, |
256
|
|
|
warm_start=warm_start, |
257
|
|
|
) |
258
|
|
|
opt1.search(X, y) |
259
|
|
|
|
260
|
|
|
assert opt0._optimizer_.score_best < opt1._optimizer_.score_best |
261
|
|
|
|
262
|
|
|
|
263
|
|
View Code Duplication |
def test_EvolutionStrategy(): |
|
|
|
|
264
|
|
|
opt0 = Hyperactive( |
265
|
|
|
search_config, |
266
|
|
|
optimizer="EvolutionStrategy", |
267
|
|
|
n_iter=n_iter_min, |
268
|
|
|
random_state=random_state, |
269
|
|
|
warm_start=warm_start, |
270
|
|
|
) |
271
|
|
|
opt0.search(X, y) |
272
|
|
|
|
273
|
|
|
opt1 = Hyperactive( |
274
|
|
|
search_config, |
275
|
|
|
optimizer="EvolutionStrategy", |
276
|
|
|
n_iter=n_iter_max, |
277
|
|
|
random_state=random_state, |
278
|
|
|
warm_start=warm_start, |
279
|
|
|
) |
280
|
|
|
opt1.search(X, y) |
281
|
|
|
|
282
|
|
|
assert opt0._optimizer_.score_best < opt1._optimizer_.score_best |
283
|
|
|
|
284
|
|
|
|
285
|
|
View Code Duplication |
def test_Bayesian(): |
|
|
|
|
286
|
|
|
opt0 = Hyperactive( |
287
|
|
|
search_config, |
288
|
|
|
optimizer="Bayesian", |
289
|
|
|
n_iter=n_iter_min, |
290
|
|
|
random_state=random_state, |
291
|
|
|
warm_start=warm_start, |
292
|
|
|
) |
293
|
|
|
opt0.search(X, y) |
294
|
|
|
|
295
|
|
|
opt1 = Hyperactive( |
296
|
|
|
search_config, |
297
|
|
|
optimizer="Bayesian", |
298
|
|
|
n_iter=n_iter_max, |
299
|
|
|
random_state=random_state, |
300
|
|
|
warm_start=warm_start, |
301
|
|
|
) |
302
|
|
|
opt1.search(X, y) |
303
|
|
|
|
304
|
|
|
assert opt0._optimizer_.score_best < opt1._optimizer_.score_best |
305
|
|
|
|