|
1
|
|
|
import time |
|
2
|
|
|
import pytest |
|
3
|
|
|
import numpy as np |
|
4
|
|
|
import pandas as pd |
|
5
|
|
|
|
|
6
|
|
|
from hyperactive import Hyperactive |
|
7
|
|
|
|
|
8
|
|
|
|
|
9
|
|
|
def objective_function_0(opt): |
|
10
|
|
|
score = -opt["x1"] * opt["x1"] |
|
11
|
|
|
return score |
|
12
|
|
|
|
|
13
|
|
|
|
|
14
|
|
|
search_space_0 = { |
|
15
|
|
|
"x1": list(np.arange(-5, 6, 1)), |
|
16
|
|
|
} |
|
17
|
|
|
search_space_1 = { |
|
18
|
|
|
"x1": list(np.arange(0, 6, 1)), |
|
19
|
|
|
} |
|
20
|
|
|
search_space_2 = { |
|
21
|
|
|
"x1": list(np.arange(-5, 1, 1)), |
|
22
|
|
|
} |
|
23
|
|
|
|
|
24
|
|
|
|
|
25
|
|
|
search_space_3 = { |
|
26
|
|
|
"x1": list(np.arange(-1, 1, 0.1)), |
|
27
|
|
|
} |
|
28
|
|
|
search_space_4 = { |
|
29
|
|
|
"x1": list(np.arange(-1, 0, 0.1)), |
|
30
|
|
|
} |
|
31
|
|
|
search_space_5 = { |
|
32
|
|
|
"x1": list(np.arange(0, 1, 0.1)), |
|
33
|
|
|
} |
|
34
|
|
|
|
|
35
|
|
|
|
|
36
|
|
|
search_space_para_0 = [ |
|
37
|
|
|
(search_space_0), |
|
38
|
|
|
(search_space_1), |
|
39
|
|
|
(search_space_2), |
|
40
|
|
|
(search_space_3), |
|
41
|
|
|
(search_space_4), |
|
42
|
|
|
(search_space_5), |
|
43
|
|
|
] |
|
44
|
|
|
|
|
45
|
|
|
|
|
46
|
|
|
@pytest.mark.parametrize("search_space", search_space_para_0) |
|
47
|
|
|
def test_trafo_0(search_space): |
|
48
|
|
|
hyper = Hyperactive() |
|
49
|
|
|
hyper.add_search(objective_function_0, search_space, n_iter=25) |
|
50
|
|
|
hyper.run() |
|
51
|
|
|
|
|
52
|
|
|
for value in hyper.search_data(objective_function_0)["x1"].values: |
|
53
|
|
|
if value not in search_space["x1"]: |
|
54
|
|
|
assert False |
|
55
|
|
|
|
|
56
|
|
|
|
|
57
|
|
|
# ----------------- # Test if memory warm starts do work as intended |
|
58
|
|
|
|
|
59
|
|
|
|
|
60
|
|
|
from sklearn.datasets import load_breast_cancer |
|
61
|
|
|
from sklearn.model_selection import cross_val_score |
|
62
|
|
|
from sklearn.tree import DecisionTreeClassifier |
|
63
|
|
|
|
|
64
|
|
|
data = load_breast_cancer() |
|
65
|
|
|
X, y = data.data, data.target |
|
66
|
|
|
|
|
67
|
|
|
|
|
68
|
|
|
def objective_function_1(opt): |
|
69
|
|
|
dtc = DecisionTreeClassifier(min_samples_split=opt["min_samples_split"]) |
|
70
|
|
|
scores = cross_val_score(dtc, X, y, cv=10) |
|
71
|
|
|
time.sleep(0.1) |
|
72
|
|
|
|
|
73
|
|
|
return scores.mean() |
|
74
|
|
|
|
|
75
|
|
|
|
|
76
|
|
|
search_space_0 = { |
|
77
|
|
|
"min_samples_split": list(np.arange(2, 12)), |
|
78
|
|
|
} |
|
79
|
|
|
|
|
80
|
|
|
search_space_1 = { |
|
81
|
|
|
"min_samples_split": list(np.arange(12, 22)), |
|
82
|
|
|
} |
|
83
|
|
|
|
|
84
|
|
|
search_space_2 = { |
|
85
|
|
|
"min_samples_split": list(np.arange(22, 32)), |
|
86
|
|
|
} |
|
87
|
|
|
|
|
88
|
|
|
memory_dict = {"min_samples_split": range(2, 12), "score": range(2, 12)} |
|
89
|
|
|
memory_warm_start_0 = pd.DataFrame(memory_dict) |
|
90
|
|
|
|
|
91
|
|
|
memory_dict = {"min_samples_split": range(12, 22), "score": range(12, 22)} |
|
92
|
|
|
memory_warm_start_1 = pd.DataFrame(memory_dict) |
|
93
|
|
|
|
|
94
|
|
|
memory_dict = {"min_samples_split": range(22, 32), "score": range(22, 32)} |
|
95
|
|
|
memory_warm_start_2 = pd.DataFrame(memory_dict) |
|
96
|
|
|
|
|
97
|
|
|
search_space_para_1 = [ |
|
98
|
|
|
(search_space_0, memory_warm_start_0), |
|
99
|
|
|
(search_space_1, memory_warm_start_1), |
|
100
|
|
|
(search_space_2, memory_warm_start_2), |
|
101
|
|
|
] |
|
102
|
|
|
|
|
103
|
|
|
random_state_para_0 = [ |
|
104
|
|
|
(0), |
|
105
|
|
|
(1), |
|
106
|
|
|
(2), |
|
107
|
|
|
(3), |
|
108
|
|
|
(4), |
|
109
|
|
|
] |
|
110
|
|
|
|
|
111
|
|
|
# ----------------- # Test if wrong memory warm starts do not work as intended |
|
112
|
|
|
""" test is possible in future gfo versions |
|
113
|
|
|
@pytest.mark.parametrize("random_state", random_state_para_0) |
|
114
|
|
|
@pytest.mark.parametrize("search_space, memory_warm_start", search_space_para_1) |
|
115
|
|
|
def test_trafo_1(random_state, search_space, memory_warm_start): |
|
116
|
|
|
search_space = search_space |
|
117
|
|
|
memory_warm_start = memory_warm_start |
|
118
|
|
|
|
|
119
|
|
|
c_time_0 = time.perf_counter() |
|
120
|
|
|
hyper = Hyperactive() |
|
121
|
|
|
hyper.add_search( |
|
122
|
|
|
objective_function_1, |
|
123
|
|
|
search_space, |
|
124
|
|
|
n_iter=10, |
|
125
|
|
|
random_state=random_state, |
|
126
|
|
|
initialize={"random": 1}, |
|
127
|
|
|
) |
|
128
|
|
|
hyper.run() |
|
129
|
|
|
d_time_0 = time.perf_counter() - c_time_0 |
|
130
|
|
|
|
|
131
|
|
|
c_time_1 = time.perf_counter() |
|
132
|
|
|
hyper = Hyperactive() |
|
133
|
|
|
hyper.add_search( |
|
134
|
|
|
objective_function_1, |
|
135
|
|
|
search_space, |
|
136
|
|
|
n_iter=10, |
|
137
|
|
|
random_state=random_state, |
|
138
|
|
|
initialize={"random": 1}, |
|
139
|
|
|
memory_warm_start=memory_warm_start, |
|
140
|
|
|
) |
|
141
|
|
|
hyper.run() |
|
142
|
|
|
d_time_1 = time.perf_counter() - c_time_1 |
|
143
|
|
|
|
|
144
|
|
|
assert d_time_1 < d_time_0 * 0.5 |
|
145
|
|
|
|
|
146
|
|
|
|
|
147
|
|
|
|
|
148
|
|
|
search_space_0 = { |
|
149
|
|
|
"min_samples_split": list(np.arange(2, 12)), |
|
150
|
|
|
} |
|
151
|
|
|
|
|
152
|
|
|
search_space_1 = { |
|
153
|
|
|
"min_samples_split": list(np.arange(12, 22)), |
|
154
|
|
|
} |
|
155
|
|
|
|
|
156
|
|
|
search_space_2 = { |
|
157
|
|
|
"min_samples_split": list(np.arange(22, 32)), |
|
158
|
|
|
} |
|
159
|
|
|
|
|
160
|
|
|
memory_dict = {"min_samples_split": range(12, 22), "score": range(2, 12)} |
|
161
|
|
|
memory_warm_start_0 = pd.DataFrame(memory_dict) |
|
162
|
|
|
|
|
163
|
|
|
memory_dict = {"min_samples_split": range(22, 32), "score": range(12, 22)} |
|
164
|
|
|
memory_warm_start_1 = pd.DataFrame(memory_dict) |
|
165
|
|
|
|
|
166
|
|
|
memory_dict = {"min_samples_split": range(2, 12), "score": range(22, 32)} |
|
167
|
|
|
memory_warm_start_2 = pd.DataFrame(memory_dict) |
|
168
|
|
|
|
|
169
|
|
|
search_space_para_2 = [ |
|
170
|
|
|
(search_space_0, memory_warm_start_0), |
|
171
|
|
|
(search_space_1, memory_warm_start_1), |
|
172
|
|
|
(search_space_2, memory_warm_start_2), |
|
173
|
|
|
] |
|
174
|
|
|
|
|
175
|
|
|
random_state_para_0 = [ |
|
176
|
|
|
(0), |
|
177
|
|
|
(1), |
|
178
|
|
|
(2), |
|
179
|
|
|
(3), |
|
180
|
|
|
(4), |
|
181
|
|
|
] |
|
182
|
|
|
|
|
183
|
|
|
|
|
184
|
|
|
@pytest.mark.parametrize("random_state", random_state_para_0) |
|
185
|
|
|
@pytest.mark.parametrize("search_space, memory_warm_start", search_space_para_2) |
|
186
|
|
|
def test_trafo_2(random_state, search_space, memory_warm_start): |
|
187
|
|
|
search_space = search_space |
|
188
|
|
|
memory_warm_start = memory_warm_start |
|
189
|
|
|
|
|
190
|
|
|
c_time_0 = time.perf_counter() |
|
191
|
|
|
hyper = Hyperactive() |
|
192
|
|
|
hyper.add_search( |
|
193
|
|
|
objective_function_1, |
|
194
|
|
|
search_space, |
|
195
|
|
|
n_iter=25, |
|
196
|
|
|
random_state=random_state, |
|
197
|
|
|
initialize={"random": 1}, |
|
198
|
|
|
) |
|
199
|
|
|
hyper.run() |
|
200
|
|
|
d_time_0 = time.perf_counter() - c_time_0 |
|
201
|
|
|
|
|
202
|
|
|
c_time_1 = time.perf_counter() |
|
203
|
|
|
hyper = Hyperactive() |
|
204
|
|
|
hyper.add_search( |
|
205
|
|
|
objective_function_1, |
|
206
|
|
|
search_space, |
|
207
|
|
|
n_iter=25, |
|
208
|
|
|
random_state=random_state, |
|
209
|
|
|
initialize={"random": 1}, |
|
210
|
|
|
memory_warm_start=memory_warm_start, |
|
211
|
|
|
) |
|
212
|
|
|
hyper.run() |
|
213
|
|
|
d_time_1 = time.perf_counter() - c_time_1 |
|
214
|
|
|
|
|
215
|
|
|
assert not (d_time_1 < d_time_0 * 0.8) |
|
216
|
|
|
""" |
|
217
|
|
|
|
SimonBlanke /
Hyperactive
