|
1
|
|
|
import numpy as np |
|
2
|
|
|
|
|
3
|
|
|
from hyperactive.optimizers import HillClimbingOptimizer |
|
4
|
|
|
from hyperactive.experiment import BaseExperiment |
|
5
|
|
|
from hyperactive.search_config import SearchConfig |
|
6
|
|
|
|
|
7
|
|
|
|
|
8
|
|
|
class Experiment(BaseExperiment): |
|
9
|
|
|
def objective_function(self, opt): |
|
10
|
|
|
score = -(opt["x1"] * opt["x1"] + opt["x2"] * opt["x2"]) |
|
11
|
|
|
return score |
|
12
|
|
|
|
|
13
|
|
|
|
|
14
|
|
|
experiment = Experiment() |
|
15
|
|
|
|
|
16
|
|
|
|
|
17
|
|
|
search_config = SearchConfig( |
|
18
|
|
|
x1=list(np.arange(-1000, 1000, 0.1)), |
|
19
|
|
|
x2=list(np.arange(-1000, 1000, 0.1)), |
|
20
|
|
|
) |
|
21
|
|
|
|
|
22
|
|
|
|
|
23
|
|
|
err = 0.001 |
|
24
|
|
|
|
|
25
|
|
|
|
|
26
|
|
|
def test_random_state_n_jobs_0(): |
|
27
|
|
|
n_jobs = 2 |
|
28
|
|
|
|
|
29
|
|
|
hyper = HillClimbingOptimizer() |
|
30
|
|
|
hyper.add_search( |
|
31
|
|
|
experiment, |
|
32
|
|
|
search_config, |
|
33
|
|
|
n_iter=5, |
|
34
|
|
|
initialize={"random": 1}, |
|
35
|
|
|
random_state=1, |
|
36
|
|
|
n_jobs=n_jobs, |
|
37
|
|
|
) |
|
38
|
|
|
hyper.run() |
|
39
|
|
|
|
|
40
|
|
|
results = hyper.search_data(experiment) |
|
41
|
|
|
|
|
42
|
|
|
no_dup = results.drop_duplicates(subset=list(search_config.keys())) |
|
43
|
|
|
print("no_dup", no_dup) |
|
44
|
|
|
print("results", results) |
|
45
|
|
|
|
|
46
|
|
|
print(int(len(results) / n_jobs)) |
|
47
|
|
|
print(len(no_dup)) |
|
48
|
|
|
|
|
49
|
|
|
assert int(len(results) / n_jobs) != len(no_dup) |
|
50
|
|
|
|
|
51
|
|
|
|
|
52
|
|
View Code Duplication |
def test_random_state_n_jobs_1(): |
|
|
|
|
|
|
53
|
|
|
n_jobs = 3 |
|
54
|
|
|
|
|
55
|
|
|
hyper = HillClimbingOptimizer() |
|
56
|
|
|
hyper.add_search( |
|
57
|
|
|
experiment, |
|
58
|
|
|
search_config, |
|
59
|
|
|
n_iter=5, |
|
60
|
|
|
initialize={"random": 1}, |
|
61
|
|
|
random_state=1, |
|
62
|
|
|
n_jobs=n_jobs, |
|
63
|
|
|
) |
|
64
|
|
|
hyper.run() |
|
65
|
|
|
|
|
66
|
|
|
results = hyper.search_data(experiment) |
|
67
|
|
|
|
|
68
|
|
|
no_dup = results.drop_duplicates(subset=list(search_config.keys())) |
|
69
|
|
|
print("no_dup", no_dup) |
|
70
|
|
|
print("results", results) |
|
71
|
|
|
|
|
72
|
|
|
assert int(len(results) / n_jobs) != len(no_dup) |
|
73
|
|
|
|
|
74
|
|
|
|
|
75
|
|
View Code Duplication |
def test_random_state_n_jobs_2(): |
|
|
|
|
|
|
76
|
|
|
n_jobs = 4 |
|
77
|
|
|
|
|
78
|
|
|
hyper = HillClimbingOptimizer() |
|
79
|
|
|
hyper.add_search( |
|
80
|
|
|
experiment, |
|
81
|
|
|
search_config, |
|
82
|
|
|
n_iter=5, |
|
83
|
|
|
initialize={"random": 1}, |
|
84
|
|
|
random_state=1, |
|
85
|
|
|
n_jobs=n_jobs, |
|
86
|
|
|
) |
|
87
|
|
|
hyper.run() |
|
88
|
|
|
|
|
89
|
|
|
results = hyper.search_data(experiment) |
|
90
|
|
|
|
|
91
|
|
|
no_dup = results.drop_duplicates(subset=list(search_config.keys())) |
|
92
|
|
|
print("no_dup", no_dup) |
|
93
|
|
|
print("results", results) |
|
94
|
|
|
|
|
95
|
|
|
assert int(len(results) / n_jobs) != len(no_dup) |
|
96
|
|
|
|
|
97
|
|
|
|
|
98
|
|
View Code Duplication |
def test_random_state_0(): |
|
|
|
|
|
|
99
|
|
|
hyper0 = HillClimbingOptimizer() |
|
100
|
|
|
hyper0.add_search( |
|
101
|
|
|
experiment, |
|
102
|
|
|
search_config, |
|
103
|
|
|
n_iter=10, |
|
104
|
|
|
initialize={"random": 1}, |
|
105
|
|
|
random_state=1, |
|
106
|
|
|
) |
|
107
|
|
|
hyper0.run() |
|
108
|
|
|
|
|
109
|
|
|
hyper1 = HillClimbingOptimizer() |
|
110
|
|
|
hyper1.add_search( |
|
111
|
|
|
experiment, |
|
112
|
|
|
search_config, |
|
113
|
|
|
n_iter=10, |
|
114
|
|
|
initialize={"random": 1}, |
|
115
|
|
|
random_state=1, |
|
116
|
|
|
) |
|
117
|
|
|
hyper1.run() |
|
118
|
|
|
|
|
119
|
|
|
best_score0 = hyper0.best_score(experiment) |
|
120
|
|
|
best_score1 = hyper1.best_score(experiment) |
|
121
|
|
|
|
|
122
|
|
|
assert abs(best_score0 - best_score1) < err |
|
123
|
|
|
|
|
124
|
|
|
|
|
125
|
|
View Code Duplication |
def test_random_state_1(): |
|
|
|
|
|
|
126
|
|
|
hyper0 = HillClimbingOptimizer() |
|
127
|
|
|
hyper0.add_search( |
|
128
|
|
|
experiment, |
|
129
|
|
|
search_config, |
|
130
|
|
|
n_iter=10, |
|
131
|
|
|
initialize={"random": 1}, |
|
132
|
|
|
random_state=10, |
|
133
|
|
|
) |
|
134
|
|
|
hyper0.run() |
|
135
|
|
|
|
|
136
|
|
|
hyper1 = HillClimbingOptimizer() |
|
137
|
|
|
hyper1.add_search( |
|
138
|
|
|
experiment, |
|
139
|
|
|
search_config, |
|
140
|
|
|
n_iter=10, |
|
141
|
|
|
initialize={"random": 1}, |
|
142
|
|
|
random_state=10, |
|
143
|
|
|
) |
|
144
|
|
|
hyper1.run() |
|
145
|
|
|
|
|
146
|
|
|
best_score0 = hyper0.best_score(experiment) |
|
147
|
|
|
best_score1 = hyper1.best_score(experiment) |
|
148
|
|
|
|
|
149
|
|
|
assert abs(best_score0 - best_score1) < err |
|
150
|
|
|
|
|
151
|
|
|
|
|
152
|
|
View Code Duplication |
def test_random_state_2(): |
|
|
|
|
|
|
153
|
|
|
hyper0 = HillClimbingOptimizer() |
|
154
|
|
|
hyper0.add_search( |
|
155
|
|
|
experiment, |
|
156
|
|
|
search_config, |
|
157
|
|
|
n_iter=10, |
|
158
|
|
|
initialize={"random": 1}, |
|
159
|
|
|
random_state=1, |
|
160
|
|
|
) |
|
161
|
|
|
hyper0.run() |
|
162
|
|
|
|
|
163
|
|
|
hyper1 = HillClimbingOptimizer() |
|
164
|
|
|
hyper1.add_search( |
|
165
|
|
|
experiment, |
|
166
|
|
|
search_config, |
|
167
|
|
|
n_iter=10, |
|
168
|
|
|
initialize={"random": 1}, |
|
169
|
|
|
random_state=10, |
|
170
|
|
|
) |
|
171
|
|
|
hyper1.run() |
|
172
|
|
|
|
|
173
|
|
|
best_score0 = hyper0.best_score(experiment) |
|
174
|
|
|
best_score1 = hyper1.best_score(experiment) |
|
175
|
|
|
|
|
176
|
|
|
assert abs(best_score0 - best_score1) > err |
|
177
|
|
|
|
|
178
|
|
|
|
|
179
|
|
|
def test_no_random_state_0(): |
|
180
|
|
|
hyper0 = HillClimbingOptimizer() |
|
181
|
|
|
hyper0.add_search( |
|
182
|
|
|
experiment, |
|
183
|
|
|
search_config, |
|
184
|
|
|
n_iter=10, |
|
185
|
|
|
initialize={"random": 1}, |
|
186
|
|
|
) |
|
187
|
|
|
hyper0.run() |
|
188
|
|
|
|
|
189
|
|
|
hyper1 = HillClimbingOptimizer() |
|
190
|
|
|
hyper1.add_search( |
|
191
|
|
|
experiment, |
|
192
|
|
|
search_config, |
|
193
|
|
|
n_iter=10, |
|
194
|
|
|
initialize={"random": 1}, |
|
195
|
|
|
) |
|
196
|
|
|
hyper1.run() |
|
197
|
|
|
|
|
198
|
|
|
best_score0 = hyper0.best_score(experiment) |
|
199
|
|
|
best_score1 = hyper1.best_score(experiment) |
|
200
|
|
|
|
|
201
|
|
|
assert abs(best_score0 - best_score1) > err |
|
202
|
|
|
|
SimonBlanke /
Hyperactive
Loading history...
