|
1
|
|
|
import numpy as np |
|
2
|
|
|
from sklearn.datasets import load_iris |
|
3
|
|
|
from sklearn.neighbors import KNeighborsClassifier |
|
4
|
|
|
from sklearn.model_selection import cross_val_score |
|
5
|
|
|
|
|
6
|
|
|
from gradient_free_optimizers import RandomSearchOptimizer |
|
7
|
|
|
|
|
8
|
|
|
|
|
9
|
|
|
def test_function(): |
|
10
|
|
|
def objective_function(para): |
|
11
|
|
|
score = -para["x1"] * para["x1"] |
|
12
|
|
|
return score |
|
13
|
|
|
|
|
14
|
|
|
search_space = { |
|
15
|
|
|
"x1": np.arange(-100, 101, 1), |
|
16
|
|
|
} |
|
17
|
|
|
|
|
18
|
|
|
opt = RandomSearchOptimizer(search_space) |
|
19
|
|
|
opt.search(objective_function, n_iter=30) |
|
20
|
|
|
|
|
21
|
|
|
|
|
22
|
|
|
def test_sklearn(): |
|
23
|
|
|
data = load_iris() |
|
24
|
|
|
X, y = data.data, data.target |
|
25
|
|
|
|
|
26
|
|
|
def model(para): |
|
27
|
|
|
knr = KNeighborsClassifier(n_neighbors=para["n_neighbors"]) |
|
28
|
|
|
scores = cross_val_score(knr, X, y, cv=5) |
|
29
|
|
|
score = scores.mean() |
|
30
|
|
|
|
|
31
|
|
|
return score |
|
32
|
|
|
|
|
33
|
|
|
search_space = { |
|
34
|
|
|
"n_neighbors": np.arange(1, 51, 1), |
|
35
|
|
|
} |
|
36
|
|
|
|
|
37
|
|
|
opt = RandomSearchOptimizer(search_space) |
|
38
|
|
|
opt.search(model, n_iter=30) |
|
39
|
|
|
|
|
40
|
|
|
|
|
41
|
|
View Code Duplication |
def test_obj_func_return_dictionary_0(): |
|
|
|
|
|
|
42
|
|
|
def objective_function(para): |
|
43
|
|
|
score = -para["x1"] * para["x1"] |
|
44
|
|
|
return score, {"_x1_": para["x1"]} |
|
45
|
|
|
|
|
46
|
|
|
search_space = { |
|
47
|
|
|
"x1": np.arange(-100, 101, 1), |
|
48
|
|
|
} |
|
49
|
|
|
|
|
50
|
|
|
opt = RandomSearchOptimizer(search_space) |
|
51
|
|
|
opt.search(objective_function, n_iter=30) |
|
52
|
|
|
|
|
53
|
|
|
assert "_x1_" in list(opt.search_data.columns) |
|
54
|
|
|
|
|
55
|
|
|
|
|
56
|
|
View Code Duplication |
def test_obj_func_return_dictionary_1(): |
|
|
|
|
|
|
57
|
|
|
def objective_function(para): |
|
58
|
|
|
score = -para["x1"] * para["x1"] |
|
59
|
|
|
return score, {"_x1_": para["x1"], "_x1_*2": para["x1"] * 2} |
|
60
|
|
|
|
|
61
|
|
|
search_space = { |
|
62
|
|
|
"x1": np.arange(-100, 101, 1), |
|
63
|
|
|
} |
|
64
|
|
|
|
|
65
|
|
|
opt = RandomSearchOptimizer(search_space) |
|
66
|
|
|
opt.search(objective_function, n_iter=30) |
|
67
|
|
|
|
|
68
|
|
|
assert "_x1_" in list(opt.search_data.columns) |
|
69
|
|
|
assert "_x1_*2" in list(opt.search_data.columns) |
|
70
|
|
|
|
SimonBlanke /
Gradient-Free-Optimizers
Loading history...