|
1
|
|
|
import numpy as np |
|
2
|
|
|
from gradient_free_optimizers import RandomSearchOptimizer |
|
3
|
|
|
|
|
4
|
|
|
|
|
5
|
|
|
def objective_function(pos_new): |
|
6
|
|
|
score = -pos_new[0] * pos_new[0] |
|
7
|
|
|
return score |
|
8
|
|
|
|
|
9
|
|
|
|
|
10
|
|
|
search_space = [np.arange(-100, 101, 1)] |
|
11
|
|
|
|
|
12
|
|
|
|
|
13
|
|
|
def test_initialize_warm_start(): |
|
14
|
|
|
initialize = {"warm_start": [np.array([0])]} |
|
15
|
|
|
|
|
16
|
|
|
opt = RandomSearchOptimizer(search_space) |
|
17
|
|
|
opt.search( |
|
18
|
|
|
objective_function, n_iter=0, initialize=initialize, |
|
19
|
|
|
) |
|
20
|
|
|
|
|
21
|
|
|
assert abs(opt.best_score) < 0.001 |
|
22
|
|
|
|
|
23
|
|
|
|
|
24
|
|
|
def test_initialize_vertices(): |
|
25
|
|
|
initialize = {"vertices": 2} |
|
26
|
|
|
|
|
27
|
|
|
opt = RandomSearchOptimizer(search_space) |
|
28
|
|
|
opt.search( |
|
29
|
|
|
objective_function, n_iter=0, initialize=initialize, |
|
30
|
|
|
) |
|
31
|
|
|
|
|
32
|
|
|
assert abs(opt.best_score) - 10000 < 0.001 |
|
33
|
|
|
|
|
34
|
|
|
|
|
35
|
|
|
def test_initialize_grid_0(): |
|
36
|
|
|
search_space = [np.arange(-1, 2, 1)] |
|
37
|
|
|
initialize = {"grid": 1} |
|
38
|
|
|
|
|
39
|
|
|
opt = RandomSearchOptimizer(search_space) |
|
40
|
|
|
opt.search( |
|
41
|
|
|
objective_function, n_iter=0, initialize=initialize, |
|
42
|
|
|
) |
|
43
|
|
|
|
|
44
|
|
|
assert abs(opt.best_score) < 0.001 |
|
45
|
|
|
|
|
46
|
|
|
|
|
47
|
|
|
def test_initialize_grid_1(): |
|
48
|
|
|
search_space = [np.arange(-2, 3, 1)] |
|
49
|
|
|
initialize = {"grid": 1} |
|
50
|
|
|
|
|
51
|
|
|
opt = RandomSearchOptimizer(search_space) |
|
52
|
|
|
opt.search( |
|
53
|
|
|
objective_function, n_iter=0, initialize=initialize, |
|
54
|
|
|
) |
|
55
|
|
|
|
|
56
|
|
|
assert abs(opt.best_score) - 1 < 0.001 |
|
57
|
|
|
|
|
58
|
|
|
|
SimonBlanke /
Gradient-Free-Optimizers
