|
1
|
|
|
import pytest |
|
2
|
|
|
import random |
|
3
|
|
|
import numpy as np |
|
4
|
|
|
|
|
5
|
|
|
from ._parametrize import optimizers |
|
6
|
|
|
|
|
7
|
|
|
|
|
8
|
|
|
def objective_function_nan(para): |
|
9
|
|
|
rand = random.randint(0, 1) |
|
10
|
|
|
|
|
11
|
|
|
if rand == 0: |
|
12
|
|
|
return 1 |
|
13
|
|
|
else: |
|
14
|
|
|
return np.nan |
|
15
|
|
|
|
|
16
|
|
|
|
|
17
|
|
|
def objective_function_m_inf(para): |
|
18
|
|
|
rand = random.randint(0, 1) |
|
19
|
|
|
|
|
20
|
|
|
if rand == 0: |
|
21
|
|
|
return 1 |
|
22
|
|
|
else: |
|
23
|
|
|
return -np.inf |
|
24
|
|
|
|
|
25
|
|
|
|
|
26
|
|
|
def objective_function_inf(para): |
|
27
|
|
|
rand = random.randint(0, 1) |
|
28
|
|
|
|
|
29
|
|
|
if rand == 0: |
|
30
|
|
|
return 1 |
|
31
|
|
|
else: |
|
32
|
|
|
return np.inf |
|
33
|
|
|
|
|
34
|
|
|
|
|
35
|
|
|
search_space = {"x1": np.arange(0, 20, 1)} |
|
36
|
|
|
|
|
37
|
|
|
|
|
38
|
|
|
objective_para = ( |
|
39
|
|
|
"objective", |
|
40
|
|
|
[ |
|
41
|
|
|
(objective_function_nan), |
|
42
|
|
|
(objective_function_m_inf), |
|
43
|
|
|
(objective_function_inf), |
|
44
|
|
|
], |
|
45
|
|
|
) |
|
46
|
|
|
|
|
47
|
|
|
|
|
48
|
|
|
@pytest.mark.parametrize(*objective_para) |
|
49
|
|
|
@pytest.mark.parametrize(*optimizers) |
|
50
|
|
|
def test_inf_nan_0(Optimizer, objective): |
|
51
|
|
|
objective_function = objective |
|
52
|
|
|
initialize = {"random": 20} |
|
53
|
|
|
|
|
54
|
|
|
opt = Optimizer(search_space, initialize=initialize) |
|
55
|
|
|
opt.search( |
|
56
|
|
|
objective_function, |
|
57
|
|
|
n_iter=80, |
|
58
|
|
|
verbosity={"print_results": False, "progress_bar": False}, |
|
59
|
|
|
) |
|
60
|
|
|
|
|
61
|
|
|
|
|
62
|
|
|
@pytest.mark.parametrize(*objective_para) |
|
63
|
|
|
@pytest.mark.parametrize(*optimizers) |
|
64
|
|
|
def test_inf_nan_1(Optimizer, objective): |
|
65
|
|
|
objective_function = objective |
|
66
|
|
|
initialize = {"random": 20} |
|
67
|
|
|
|
|
68
|
|
|
opt = Optimizer(search_space, initialize=initialize) |
|
69
|
|
|
opt.search( |
|
70
|
|
|
objective_function, |
|
71
|
|
|
n_iter=50, |
|
72
|
|
|
memory=False, |
|
73
|
|
|
verbosity={"print_results": False, "progress_bar": False}, |
|
74
|
|
|
) |
|
75
|
|
|
|
|
76
|
|
|
search_data = opt.search_data |
|
77
|
|
|
print("\n search_data \n", search_data) |
|
78
|
|
|
|
|
79
|
|
|
non_inf_mask = ~np.isinf(search_data["score"].values) |
|
80
|
|
|
non_nan_mask = ~np.isnan(search_data["score"].values) |
|
81
|
|
|
|
|
82
|
|
|
non_inf_nan = np.sum(non_inf_mask * non_nan_mask) |
|
83
|
|
|
|
|
84
|
|
|
assert 10 < non_inf_nan < 40 |
|
85
|
|
|
|
SimonBlanke /
Gradient-Free-Optimizers
