|
1
|
|
|
# Author: Simon Blanke |
|
2
|
|
|
# Email: [email protected] |
|
3
|
|
|
# License: MIT License |
|
4
|
|
|
|
|
5
|
|
|
|
|
6
|
|
|
from sklearn.datasets import load_iris |
|
7
|
|
|
from sklearn.model_selection import cross_val_score |
|
8
|
|
|
from sklearn.tree import DecisionTreeClassifier |
|
9
|
|
|
from hyperactive import Hyperactive |
|
10
|
|
|
|
|
11
|
|
|
data = load_iris() |
|
12
|
|
|
X = data.data |
|
13
|
|
|
y = data.target |
|
14
|
|
|
memory = False |
|
15
|
|
|
|
|
16
|
|
|
|
|
17
|
|
|
def model(para, X, y): |
|
18
|
|
|
model = DecisionTreeClassifier( |
|
19
|
|
|
max_depth=para["max_depth"], |
|
20
|
|
|
min_samples_split=para["min_samples_split"], |
|
21
|
|
|
min_samples_leaf=para["min_samples_leaf"], |
|
22
|
|
|
) |
|
23
|
|
|
scores = cross_val_score(model, X, y, cv=3) |
|
24
|
|
|
|
|
25
|
|
|
return scores.mean() |
|
26
|
|
|
|
|
27
|
|
|
|
|
28
|
|
|
search_config = { |
|
29
|
|
|
model: { |
|
30
|
|
|
"max_depth": range(1, 21), |
|
31
|
|
|
"min_samples_split": range(2, 21), |
|
32
|
|
|
"min_samples_leaf": range(1, 21), |
|
33
|
|
|
} |
|
34
|
|
|
} |
|
35
|
|
|
|
|
36
|
|
|
|
|
37
|
|
|
def test_func_return(): |
|
38
|
|
|
def model1(para, X, y): |
|
39
|
|
|
model = DecisionTreeClassifier( |
|
40
|
|
|
criterion=para["criterion"], |
|
41
|
|
|
max_depth=para["max_depth"], |
|
42
|
|
|
min_samples_split=para["min_samples_split"], |
|
43
|
|
|
min_samples_leaf=para["min_samples_leaf"], |
|
44
|
|
|
) |
|
45
|
|
|
scores = cross_val_score(model, X, y, cv=3) |
|
46
|
|
|
|
|
47
|
|
|
return scores.mean(), model |
|
48
|
|
|
|
|
49
|
|
|
search_config1 = { |
|
50
|
|
|
model1: { |
|
51
|
|
|
"criterion": ["gini", "entropy"], |
|
52
|
|
|
"max_depth": range(1, 21), |
|
53
|
|
|
"min_samples_split": range(2, 21), |
|
54
|
|
|
"min_samples_leaf": range(1, 21), |
|
55
|
|
|
} |
|
56
|
|
|
} |
|
57
|
|
|
|
|
58
|
|
|
opt = Hyperactive(X, y, memory=memory) |
|
59
|
|
|
opt.search(search_config1) |
|
60
|
|
|
|
|
61
|
|
|
|
|
62
|
|
|
def test_n_jobs_2(): |
|
63
|
|
|
opt = Hyperactive(X, y, memory=memory) |
|
64
|
|
|
opt.search(search_config, n_jobs=2) |
|
65
|
|
|
|
|
66
|
|
|
|
|
67
|
|
|
def test_n_jobs_4(): |
|
68
|
|
|
opt = Hyperactive(X, y, memory=memory) |
|
69
|
|
|
opt.search(search_config, n_jobs=4) |
|
70
|
|
|
|
|
71
|
|
|
|
|
72
|
|
|
def test_positional_args(): |
|
73
|
|
|
opt0 = Hyperactive(X, y, random_state=False, memory=memory) |
|
74
|
|
|
opt0.search(search_config) |
|
75
|
|
|
|
|
76
|
|
|
opt1 = Hyperactive(X, y, random_state=1, memory=memory) |
|
77
|
|
|
opt1.search(search_config) |
|
78
|
|
|
|
|
79
|
|
|
opt2 = Hyperactive(X, y, random_state=1, memory=memory) |
|
80
|
|
|
opt2.search(search_config) |
|
81
|
|
|
|
|
82
|
|
|
|
|
83
|
|
|
def test_random_state(): |
|
84
|
|
|
opt0 = Hyperactive(X, y, random_state=False, memory=memory) |
|
85
|
|
|
opt0.search(search_config) |
|
86
|
|
|
|
|
87
|
|
|
opt1 = Hyperactive(X, y, random_state=0, memory=memory) |
|
88
|
|
|
opt1.search(search_config) |
|
89
|
|
|
|
|
90
|
|
|
opt2 = Hyperactive(X, y, random_state=1, memory=memory) |
|
91
|
|
|
opt2.search(search_config) |
|
92
|
|
|
|
|
93
|
|
|
|
|
94
|
|
|
def test_max_time(): |
|
95
|
|
|
opt0 = Hyperactive(X, y, memory=memory) |
|
96
|
|
|
opt0.search(search_config, max_time=0.00001) |
|
97
|
|
|
|
|
98
|
|
|
|
|
99
|
|
|
def test_memory(): |
|
100
|
|
|
opt0 = Hyperactive(X, y, memory=True) |
|
101
|
|
|
opt0.search(search_config) |
|
102
|
|
|
|
|
103
|
|
|
opt1 = Hyperactive(X, y, memory=False) |
|
104
|
|
|
opt1.search(search_config) |
|
105
|
|
|
|
|
106
|
|
|
opt2 = Hyperactive(X, y, memory="short") |
|
107
|
|
|
opt2.search(search_config) |
|
108
|
|
|
|
|
109
|
|
|
opt3 = Hyperactive(X, y, memory="long") |
|
110
|
|
|
opt3.search(search_config) |
|
111
|
|
|
|
|
112
|
|
|
opt4 = Hyperactive(X, y, memory="long") |
|
113
|
|
|
opt4.search(search_config) |
|
114
|
|
|
|
|
115
|
|
|
|
|
116
|
|
|
def test_dill(): |
|
117
|
|
|
from sklearn.gaussian_process import GaussianProcessClassifier |
|
118
|
|
|
from sklearn.gaussian_process.kernels import RBF, Matern |
|
119
|
|
|
|
|
120
|
|
|
def model(para, X, y): |
|
121
|
|
|
gpc = GaussianProcessClassifier(kernel=para["kernel"]) |
|
122
|
|
|
scores = cross_val_score(gpc, X, y, cv=2) |
|
123
|
|
|
|
|
124
|
|
|
return scores.mean() |
|
125
|
|
|
|
|
126
|
|
|
search_config = {model: {"kernel": [RBF(), Matern()]}} |
|
127
|
|
|
|
|
128
|
|
|
opt0 = Hyperactive(X, y, memory="long") |
|
129
|
|
|
opt0.search(search_config) |
|
130
|
|
|
|
|
131
|
|
|
opt1 = Hyperactive(X, y, memory="long") |
|
132
|
|
|
opt1.search(search_config) |
|
133
|
|
|
|
|
134
|
|
|
|
|
135
|
|
|
def test_verbosity0(): |
|
136
|
|
|
opt = Hyperactive(X, y, verbosity=0, memory=memory) |
|
137
|
|
|
opt.search(search_config) |
|
138
|
|
|
|
|
139
|
|
|
|
|
140
|
|
|
def test_verbosity1(): |
|
141
|
|
|
opt = Hyperactive(X, y, verbosity=0, memory=memory) |
|
142
|
|
|
opt.search(search_config, n_jobs=2) |
|
143
|
|
|
|
|
144
|
|
|
|
|
145
|
|
|
def test_verbosity2(): |
|
146
|
|
|
opt = Hyperactive(X, y, verbosity=1, memory=memory) |
|
147
|
|
|
opt.search(search_config, n_jobs=2) |
|
148
|
|
|
|
|
149
|
|
|
|
|
150
|
|
|
def test_verbosity3(): |
|
151
|
|
|
opt = Hyperactive(X, y, verbosity=1, memory=memory) |
|
152
|
|
|
opt.search(search_config) |
|
153
|
|
|
|
|
154
|
|
|
|
|
155
|
|
|
def test_verbosity4(): |
|
156
|
|
|
opt = Hyperactive(X, y, verbosity=2, memory=memory) |
|
157
|
|
|
opt.search(search_config) |
|
158
|
|
|
|
|
159
|
|
|
|
|
160
|
|
|
def test_verbosity5(): |
|
161
|
|
|
opt = Hyperactive(X, y, verbosity=2, memory=memory) |
|
162
|
|
|
opt.search(search_config, n_jobs=2) |
|
163
|
|
|
|
|
164
|
|
|
|
|
165
|
|
|
def test_scatter_init(): |
|
166
|
|
|
init_config = {model: {"scatter_init": 10}} |
|
167
|
|
|
opt = Hyperactive(X, y, memory=memory) |
|
168
|
|
|
opt.search(search_config, init_config=init_config) |
|
169
|
|
|
|
|
170
|
|
|
|
|
171
|
|
|
def test_warm_start(): |
|
172
|
|
|
init_config = { |
|
173
|
|
|
model: {"max_depth": 10, "min_samples_split": 2, "min_samples_leaf": 5} |
|
174
|
|
|
} |
|
175
|
|
|
opt = Hyperactive(X, y, memory=memory) |
|
176
|
|
|
opt.search(search_config, n_iter=0, init_config=init_config) |
|
177
|
|
|
|
|
178
|
|
|
assert opt.results[model] == init_config[model] |
|
179
|
|
|
|
|
180
|
|
|
|
|
181
|
|
|
def test_partial_warm_start(): |
|
182
|
|
|
init_config = {model: {"min_samples_split": 2, "min_samples_leaf": 5}} |
|
183
|
|
|
opt = Hyperactive(X, y, memory=memory) |
|
184
|
|
|
opt.search(search_config, n_iter=0, init_config=init_config) |
|
185
|
|
|
|
|
186
|
|
|
|
|
187
|
|
|
def test_optimizer_args(): |
|
188
|
|
|
opt = Hyperactive(X, y, memory=memory) |
|
189
|
|
|
opt.search(search_config, optimizer={"HillClimbing": {"epsilon": 0.1}}) |
|
190
|
|
|
|
|
191
|
|
|
|
|
192
|
|
|
def test_get_search_path(): |
|
193
|
|
|
opt = Hyperactive(X, y, verbosity=10, memory=memory) |
|
194
|
|
|
opt.search(search_config) |
|
195
|
|
|
|
|
196
|
|
|
opt = Hyperactive(X, y, verbosity=10, memory=memory) |
|
197
|
|
|
opt.search(search_config, optimizer="ParticleSwarm") |
|
198
|
|
|
|
|
199
|
|
|
|
|
200
|
|
|
""" |
|
201
|
|
|
def test_ray_1(): |
|
202
|
|
|
ray.init() |
|
203
|
|
|
opt = Hyperactive(X, y, memory=memory) |
|
204
|
|
|
opt.search(search_config, n_jobs=1) |
|
205
|
|
|
|
|
206
|
|
|
|
|
207
|
|
|
def test_ray_2(): |
|
208
|
|
|
ray.init() |
|
209
|
|
|
opt = Hyperactive(X, y, memory=memory) |
|
210
|
|
|
opt.search(search_config, n_jobs=2) |
|
211
|
|
|
""" |
|
212
|
|
|
|
SimonBlanke /
Hyperactive
