|
1
|
|
|
import pytest |
|
2
|
|
|
import time |
|
3
|
|
|
import numpy as np |
|
4
|
|
|
import pandas as pd |
|
5
|
|
|
|
|
6
|
|
|
|
|
7
|
|
|
from ._parametrize import optimizers_non_deterministic as optimizers |
|
8
|
|
|
from gradient_free_optimizers import DirectAlgorithm |
|
9
|
|
|
|
|
10
|
|
|
|
|
11
|
|
|
def objective_function(para): |
|
12
|
|
|
score = -(para["x0"] * para["x0"] + para["x1"] * para["x1"]) |
|
13
|
|
|
return score |
|
14
|
|
|
|
|
15
|
|
|
|
|
16
|
|
|
search_space = { |
|
17
|
|
|
"x0": np.arange(-75, 100, 1), |
|
18
|
|
|
"x1": np.arange(-100, 75, 1), |
|
19
|
|
|
} |
|
20
|
|
|
|
|
21
|
|
|
err = 0.000001 |
|
22
|
|
|
|
|
23
|
|
|
n_iter = 10 |
|
24
|
|
|
n_random = 2 |
|
25
|
|
|
|
|
26
|
|
|
n_last = n_iter - n_random |
|
27
|
|
|
|
|
28
|
|
|
|
|
29
|
|
View Code Duplication |
@pytest.mark.parametrize(*optimizers) |
|
|
|
|
|
|
30
|
|
|
def test_random_state_0(Optimizer): |
|
31
|
|
|
opt0 = Optimizer(search_space, initialize={"random": n_random}, random_state=1) |
|
32
|
|
|
opt0.search( |
|
33
|
|
|
objective_function, |
|
34
|
|
|
n_iter=n_iter, |
|
35
|
|
|
) |
|
36
|
|
|
|
|
37
|
|
|
opt1 = Optimizer(search_space, initialize={"random": n_random}, random_state=1) |
|
38
|
|
|
opt1.search( |
|
39
|
|
|
objective_function, |
|
40
|
|
|
n_iter=n_iter, |
|
41
|
|
|
) |
|
42
|
|
|
|
|
43
|
|
|
print("\n opt0.search_data \n", opt0.search_data) |
|
44
|
|
|
print("\n opt1.search_data \n", opt1.search_data) |
|
45
|
|
|
|
|
46
|
|
|
n_last_scores0 = list(opt0.search_data["score"].values)[-n_last:] |
|
47
|
|
|
n_last_scores1 = list(opt1.search_data["score"].values)[-n_last:] |
|
48
|
|
|
|
|
49
|
|
|
assert abs(np.sum(n_last_scores0) - np.sum(n_last_scores1)) < err |
|
50
|
|
|
|
|
51
|
|
|
|
|
52
|
|
|
@pytest.mark.parametrize(*optimizers) |
|
53
|
|
|
def test_random_state_1(Optimizer): |
|
54
|
|
|
opt0 = Optimizer(search_space, initialize={"random": n_random}, random_state=10) |
|
55
|
|
|
opt0.search( |
|
56
|
|
|
objective_function, |
|
57
|
|
|
n_iter=n_iter, |
|
58
|
|
|
) |
|
59
|
|
|
|
|
60
|
|
|
opt1 = Optimizer(search_space, initialize={"random": n_random}, random_state=10) |
|
61
|
|
|
opt1.search( |
|
62
|
|
|
objective_function, |
|
63
|
|
|
n_iter=n_iter, |
|
64
|
|
|
) |
|
65
|
|
|
|
|
66
|
|
|
n_last_scores0 = list(opt0.search_data["score"].values)[-n_last:] |
|
67
|
|
|
n_last_scores1 = list(opt1.search_data["score"].values)[-n_last:] |
|
68
|
|
|
|
|
69
|
|
|
assert abs(np.sum(n_last_scores0) - np.sum(n_last_scores1)) < err |
|
70
|
|
|
|
|
71
|
|
|
|
|
72
|
|
View Code Duplication |
@pytest.mark.parametrize(*optimizers) |
|
|
|
|
|
|
73
|
|
|
def test_random_state_2(Optimizer): |
|
74
|
|
|
opt0 = Optimizer(search_space, initialize={"random": n_random}, random_state=1) |
|
75
|
|
|
opt0.search( |
|
76
|
|
|
objective_function, |
|
77
|
|
|
n_iter=n_iter, |
|
78
|
|
|
) |
|
79
|
|
|
|
|
80
|
|
|
opt1 = Optimizer(search_space, initialize={"random": n_random}, random_state=10) |
|
81
|
|
|
opt1.search( |
|
82
|
|
|
objective_function, |
|
83
|
|
|
n_iter=n_iter, |
|
84
|
|
|
) |
|
85
|
|
|
|
|
86
|
|
|
print("\n opt0.search_data \n", opt0.search_data) |
|
87
|
|
|
print("\n opt1.search_data \n", opt1.search_data) |
|
88
|
|
|
|
|
89
|
|
|
n_last_scores0 = list(opt0.search_data["score"].values)[-n_last:] |
|
90
|
|
|
n_last_scores1 = list(opt1.search_data["score"].values)[-n_last:] |
|
91
|
|
|
|
|
92
|
|
|
assert abs(np.sum(n_last_scores0) - np.sum(n_last_scores1)) > err |
|
93
|
|
|
|
|
94
|
|
|
|
|
95
|
|
View Code Duplication |
def test_random_state_direct(): |
|
|
|
|
|
|
96
|
|
|
opt0 = DirectAlgorithm( |
|
97
|
|
|
search_space, initialize={"random": n_random}, random_state=1 |
|
98
|
|
|
) |
|
99
|
|
|
opt0.search( |
|
100
|
|
|
objective_function, |
|
101
|
|
|
n_iter=n_iter, |
|
102
|
|
|
) |
|
103
|
|
|
|
|
104
|
|
|
opt1 = DirectAlgorithm( |
|
105
|
|
|
search_space, initialize={"random": n_random}, random_state=10 |
|
106
|
|
|
) |
|
107
|
|
|
opt1.search( |
|
108
|
|
|
objective_function, |
|
109
|
|
|
n_iter=n_iter, |
|
110
|
|
|
) |
|
111
|
|
|
|
|
112
|
|
|
print("\n opt0.search_data \n", opt0.search_data) |
|
113
|
|
|
print("\n opt1.search_data \n", opt1.search_data) |
|
114
|
|
|
|
|
115
|
|
|
n_last_scores0 = list(opt0.search_data["score"].values)[-n_last:] |
|
116
|
|
|
n_last_scores1 = list(opt1.search_data["score"].values)[-n_last:] |
|
117
|
|
|
|
|
118
|
|
|
assert abs(np.sum(n_last_scores0) - np.sum(n_last_scores1)) < err |
|
119
|
|
|
|
|
120
|
|
|
|
|
121
|
|
|
@pytest.mark.parametrize(*optimizers) |
|
122
|
|
|
def test_no_random_state_0(Optimizer): |
|
123
|
|
|
opt0 = Optimizer(search_space, initialize={"random": n_random}) |
|
124
|
|
|
opt0.search(objective_function, n_iter=n_iter) |
|
125
|
|
|
|
|
126
|
|
|
opt1 = Optimizer(search_space, initialize={"random": n_random}) |
|
127
|
|
|
opt1.search(objective_function, n_iter=n_iter) |
|
128
|
|
|
|
|
129
|
|
|
print("\n opt0.search_data \n", opt0.search_data) |
|
130
|
|
|
print("\n opt1.search_data \n", opt1.search_data) |
|
131
|
|
|
|
|
132
|
|
|
n_last_scores0 = list(opt0.search_data["score"].values)[-n_last:] |
|
133
|
|
|
n_last_scores1 = list(opt1.search_data["score"].values)[-n_last:] |
|
134
|
|
|
|
|
135
|
|
|
assert abs(np.sum(n_last_scores0) - np.sum(n_last_scores1)) > err |
|
136
|
|
|
|
SimonBlanke /
Gradient-Free-Optimizers
Loading history...