|
1
|
|
|
import numpy as np |
|
2
|
|
|
from gradient_free_optimizers import ( |
|
3
|
|
|
HillClimbingOptimizer, |
|
4
|
|
|
StochasticHillClimbingOptimizer, |
|
5
|
|
|
RepulsingHillClimbingOptimizer, |
|
6
|
|
|
RandomSearchOptimizer, |
|
7
|
|
|
RandomRestartHillClimbingOptimizer, |
|
8
|
|
|
RandomAnnealingOptimizer, |
|
9
|
|
|
SimulatedAnnealingOptimizer, |
|
10
|
|
|
ParallelTemperingOptimizer, |
|
11
|
|
|
ParticleSwarmOptimizer, |
|
12
|
|
|
EvolutionStrategyOptimizer, |
|
13
|
|
|
BayesianOptimizer, |
|
14
|
|
|
TreeStructuredParzenEstimators, |
|
15
|
|
|
ForestOptimizer, |
|
16
|
|
|
EnsembleOptimizer, |
|
17
|
|
|
) |
|
18
|
|
|
|
|
19
|
|
|
# check if there are any debug-prints left in code |
|
20
|
|
|
|
|
21
|
|
|
|
|
22
|
|
|
optimizer_list = [ |
|
23
|
|
|
HillClimbingOptimizer, |
|
24
|
|
|
StochasticHillClimbingOptimizer, |
|
25
|
|
|
RepulsingHillClimbingOptimizer, |
|
26
|
|
|
RandomSearchOptimizer, |
|
27
|
|
|
RandomRestartHillClimbingOptimizer, |
|
28
|
|
|
RandomAnnealingOptimizer, |
|
29
|
|
|
SimulatedAnnealingOptimizer, |
|
30
|
|
|
ParallelTemperingOptimizer, |
|
31
|
|
|
ParticleSwarmOptimizer, |
|
32
|
|
|
EvolutionStrategyOptimizer, |
|
33
|
|
|
BayesianOptimizer, |
|
34
|
|
|
TreeStructuredParzenEstimators, |
|
35
|
|
|
ForestOptimizer, |
|
36
|
|
|
EnsembleOptimizer, |
|
37
|
|
|
] |
|
38
|
|
|
|
|
39
|
|
|
|
|
40
|
|
|
def objective_function(para): |
|
41
|
|
|
score = -para["x1"] * para["x1"] |
|
42
|
|
|
return score |
|
43
|
|
|
|
|
44
|
|
|
|
|
45
|
|
|
search_space = { |
|
46
|
|
|
"x1": np.arange(0, 5, 1), |
|
47
|
|
|
} |
|
48
|
|
|
|
|
49
|
|
|
|
|
50
|
|
|
for optimizer in optimizer_list: |
|
51
|
|
|
opt0 = optimizer(search_space) |
|
52
|
|
|
opt0.search(objective_function, n_iter=15, verbosity=False, memory=False) |
|
53
|
|
|
|
|
54
|
|
|
opt1 = optimizer(search_space) |
|
55
|
|
|
opt1.search(objective_function, n_iter=15, verbosity=False) |
|
56
|
|
|
|
|
57
|
|
|
opt2 = optimizer(search_space) |
|
58
|
|
|
opt2.search( |
|
59
|
|
|
objective_function, |
|
60
|
|
|
n_iter=15, |
|
61
|
|
|
verbosity=False, |
|
62
|
|
|
memory_warm_start=opt1.search_data, |
|
63
|
|
|
) |
|
64
|
|
|
|
|
65
|
|
|
opt3 = optimizer(search_space, initialize={"warm_start": [{"x1": 1}]}) |
|
66
|
|
|
opt3.search( |
|
67
|
|
|
objective_function, |
|
68
|
|
|
n_iter=15, |
|
69
|
|
|
verbosity=False, |
|
70
|
|
|
) |
|
71
|
|
|
|
SimonBlanke /
Gradient-Free-Optimizers
