|
1
|
|
|
# Author: Simon Blanke |
|
2
|
|
|
# Email: [email protected] |
|
3
|
|
|
# License: MIT License |
|
4
|
|
|
|
|
5
|
|
|
import numpy as np |
|
6
|
|
|
|
|
7
|
|
|
from sklearn.datasets import load_iris |
|
8
|
|
|
from sklearn.model_selection import cross_val_score |
|
9
|
|
|
from sklearn.tree import DecisionTreeClassifier |
|
10
|
|
|
|
|
11
|
|
|
from hyperactive import Hyperactive |
|
12
|
|
|
|
|
13
|
|
|
data = load_iris() |
|
14
|
|
|
X, y = data.data, data.target |
|
15
|
|
|
|
|
16
|
|
|
|
|
17
|
|
|
def model(para, X, y): |
|
18
|
|
|
dtc = DecisionTreeClassifier( |
|
19
|
|
|
max_depth=para["max_depth"], min_samples_split=para["min_samples_split"], |
|
20
|
|
|
) |
|
21
|
|
|
scores = cross_val_score(dtc, X, y, cv=2) |
|
22
|
|
|
|
|
23
|
|
|
return scores.mean() |
|
24
|
|
|
|
|
25
|
|
|
|
|
26
|
|
|
search_space = { |
|
27
|
|
|
"max_depth": range(1, 21), |
|
28
|
|
|
"min_samples_split": range(2, 21), |
|
29
|
|
|
} |
|
30
|
|
|
|
|
31
|
|
|
|
|
32
|
|
|
def _base_test(search, opt_args={}, time=None): |
|
33
|
|
|
opt = Hyperactive(X, y, **opt_args) |
|
34
|
|
|
opt.add_search(**search) |
|
35
|
|
|
opt.run(time) |
|
36
|
|
|
|
|
37
|
|
|
|
|
38
|
|
|
def test_HillClimbingOptimizer(): |
|
39
|
|
|
search = { |
|
40
|
|
|
"model": model, |
|
41
|
|
|
"function_parameter": {"features": X, "target": y}, |
|
42
|
|
|
"search_space": search_space, |
|
43
|
|
|
"n_iter": 15, |
|
44
|
|
|
"optimizer": "HillClimbing", |
|
45
|
|
|
} |
|
46
|
|
|
|
|
47
|
|
|
_base_test(search) |
|
48
|
|
|
|
|
49
|
|
|
|
|
50
|
|
|
def test_StochasticHillClimbingOptimizer(): |
|
51
|
|
|
search = { |
|
52
|
|
|
"model": model, |
|
53
|
|
|
"function_parameter": {"features": X, "target": y}, |
|
54
|
|
|
"search_space": search_space, |
|
55
|
|
|
"n_iter": 15, |
|
56
|
|
|
"optimizer": "StochasticHillClimbing", |
|
57
|
|
|
} |
|
58
|
|
|
|
|
59
|
|
|
_base_test(search) |
|
60
|
|
|
|
|
61
|
|
|
|
|
62
|
|
|
def test_TabuOptimizer(): |
|
63
|
|
|
search = { |
|
64
|
|
|
"model": model, |
|
65
|
|
|
"function_parameter": {"features": X, "target": y}, |
|
66
|
|
|
"search_space": search_space, |
|
67
|
|
|
"n_iter": 15, |
|
68
|
|
|
"optimizer": "TabuSearch", |
|
69
|
|
|
} |
|
70
|
|
|
|
|
71
|
|
|
_base_test(search) |
|
72
|
|
|
|
|
73
|
|
|
|
|
74
|
|
|
def test_RandomSearchOptimizer(): |
|
75
|
|
|
search = { |
|
76
|
|
|
"model": model, |
|
77
|
|
|
"function_parameter": {"features": X, "target": y}, |
|
78
|
|
|
"search_space": search_space, |
|
79
|
|
|
"n_iter": 15, |
|
80
|
|
|
"optimizer": "RandomSearch", |
|
81
|
|
|
} |
|
82
|
|
|
|
|
83
|
|
|
_base_test(search) |
|
84
|
|
|
|
|
85
|
|
|
|
|
86
|
|
|
def test_RandomRestartHillClimbingOptimizer(): |
|
87
|
|
|
search = { |
|
88
|
|
|
"model": model, |
|
89
|
|
|
"function_parameter": {"features": X, "target": y}, |
|
90
|
|
|
"search_space": search_space, |
|
91
|
|
|
"n_iter": 15, |
|
92
|
|
|
"optimizer": "RandomRestartHillClimbing", |
|
93
|
|
|
} |
|
94
|
|
|
|
|
95
|
|
|
_base_test(search) |
|
96
|
|
|
|
|
97
|
|
|
|
|
98
|
|
|
def test_RandomAnnealingOptimizer(): |
|
99
|
|
|
search = { |
|
100
|
|
|
"model": model, |
|
101
|
|
|
"function_parameter": {"features": X, "target": y}, |
|
102
|
|
|
"search_space": search_space, |
|
103
|
|
|
"n_iter": 15, |
|
104
|
|
|
"optimizer": "RandomAnnealing", |
|
105
|
|
|
} |
|
106
|
|
|
|
|
107
|
|
|
_base_test(search) |
|
108
|
|
|
|
|
109
|
|
|
|
|
110
|
|
|
def test_SimulatedAnnealingOptimizer(): |
|
111
|
|
|
search = { |
|
112
|
|
|
"model": model, |
|
113
|
|
|
"function_parameter": {"features": X, "target": y}, |
|
114
|
|
|
"search_space": search_space, |
|
115
|
|
|
"n_iter": 15, |
|
116
|
|
|
"optimizer": "SimulatedAnnealing", |
|
117
|
|
|
} |
|
118
|
|
|
|
|
119
|
|
|
_base_test(search) |
|
120
|
|
|
|
|
121
|
|
|
|
|
122
|
|
|
def test_StochasticTunnelingOptimizer(): |
|
123
|
|
|
search = { |
|
124
|
|
|
"model": model, |
|
125
|
|
|
"function_parameter": {"features": X, "target": y}, |
|
126
|
|
|
"search_space": search_space, |
|
127
|
|
|
"n_iter": 15, |
|
128
|
|
|
"optimizer": "StochasticTunneling", |
|
129
|
|
|
} |
|
130
|
|
|
|
|
131
|
|
|
_base_test(search) |
|
132
|
|
|
|
|
133
|
|
|
|
|
134
|
|
|
def test_ParallelTemperingOptimizer(): |
|
135
|
|
|
search = { |
|
136
|
|
|
"model": model, |
|
137
|
|
|
"function_parameter": {"features": X, "target": y}, |
|
138
|
|
|
"search_space": search_space, |
|
139
|
|
|
"n_iter": 15, |
|
140
|
|
|
"optimizer": "ParallelTempering", |
|
141
|
|
|
} |
|
142
|
|
|
|
|
143
|
|
|
_base_test(search) |
|
144
|
|
|
|
|
145
|
|
|
|
|
146
|
|
|
def test_ParticleSwarmOptimizer(): |
|
147
|
|
|
search = { |
|
148
|
|
|
"model": model, |
|
149
|
|
|
"function_parameter": {"features": X, "target": y}, |
|
150
|
|
|
"search_space": search_space, |
|
151
|
|
|
"n_iter": 15, |
|
152
|
|
|
"optimizer": "ParticleSwarm", |
|
153
|
|
|
} |
|
154
|
|
|
|
|
155
|
|
|
_base_test(search) |
|
156
|
|
|
|
|
157
|
|
|
|
|
158
|
|
|
def test_EvolutionStrategyOptimizer(): |
|
159
|
|
|
search = { |
|
160
|
|
|
"model": model, |
|
161
|
|
|
"function_parameter": {"features": X, "target": y}, |
|
162
|
|
|
"search_space": search_space, |
|
163
|
|
|
"n_iter": 15, |
|
164
|
|
|
"optimizer": "EvolutionStrategy", |
|
165
|
|
|
} |
|
166
|
|
|
|
|
167
|
|
|
_base_test(search) |
|
168
|
|
|
|
|
169
|
|
|
|
|
170
|
|
|
def test_BayesianOptimizer(): |
|
171
|
|
|
search = { |
|
172
|
|
|
"model": model, |
|
173
|
|
|
"function_parameter": {"features": X, "target": y}, |
|
174
|
|
|
"search_space": search_space, |
|
175
|
|
|
"n_iter": 15, |
|
176
|
|
|
"optimizer": "Bayesian", |
|
177
|
|
|
} |
|
178
|
|
|
|
|
179
|
|
|
_base_test(search) |
|
180
|
|
|
|
|
181
|
|
|
|
|
182
|
|
|
def test_TPE(): |
|
183
|
|
|
search = { |
|
184
|
|
|
"model": model, |
|
185
|
|
|
"function_parameter": {"features": X, "target": y}, |
|
186
|
|
|
"search_space": search_space, |
|
187
|
|
|
"n_iter": 15, |
|
188
|
|
|
"optimizer": "TPE", |
|
189
|
|
|
} |
|
190
|
|
|
|
|
191
|
|
|
_base_test(search) |
|
192
|
|
|
|
|
193
|
|
|
|
|
194
|
|
|
def test_DecisionTreeOptimizer(): |
|
195
|
|
|
search = { |
|
196
|
|
|
"model": model, |
|
197
|
|
|
"function_parameter": {"features": X, "target": y}, |
|
198
|
|
|
"search_space": search_space, |
|
199
|
|
|
"n_iter": 15, |
|
200
|
|
|
"optimizer": "DecisionTree", |
|
201
|
|
|
} |
|
202
|
|
|
|
|
203
|
|
|
_base_test(search) |
|
204
|
|
|
|
SimonBlanke /
Hyperactive
