|
1
|
|
|
import time |
|
2
|
|
|
from tqdm import tqdm |
|
3
|
|
|
import numpy as np |
|
4
|
|
|
import pandas as pd |
|
5
|
|
|
|
|
6
|
|
|
from gradient_free_optimizers import ( |
|
7
|
|
|
HillClimbingOptimizer, |
|
8
|
|
|
StochasticHillClimbingOptimizer, |
|
9
|
|
|
TabuOptimizer, |
|
10
|
|
|
RandomSearchOptimizer, |
|
11
|
|
|
RandomRestartHillClimbingOptimizer, |
|
12
|
|
|
RandomAnnealingOptimizer, |
|
13
|
|
|
SimulatedAnnealingOptimizer, |
|
14
|
|
|
ParallelTemperingOptimizer, |
|
15
|
|
|
ParticleSwarmOptimizer, |
|
16
|
|
|
EvolutionStrategyOptimizer, |
|
17
|
|
|
BayesianOptimizer, |
|
18
|
|
|
TreeStructuredParzenEstimators, |
|
19
|
|
|
DecisionTreeOptimizer, |
|
20
|
|
|
EnsembleOptimizer, |
|
21
|
|
|
) |
|
22
|
|
|
|
|
23
|
|
|
|
|
24
|
|
|
n_inits = 4 |
|
25
|
|
|
|
|
26
|
|
|
optimizer_dict = { |
|
27
|
|
|
"Hill climbing": HillClimbingOptimizer, |
|
28
|
|
|
"Stochastic hill climbing": StochasticHillClimbingOptimizer, |
|
29
|
|
|
"Tabu search": TabuOptimizer, |
|
30
|
|
|
"Random search": RandomSearchOptimizer, |
|
31
|
|
|
"Random restart hill climbing": RandomRestartHillClimbingOptimizer, |
|
32
|
|
|
"Random annealing": RandomAnnealingOptimizer, |
|
33
|
|
|
"Simulated annealing": SimulatedAnnealingOptimizer, |
|
34
|
|
|
"Parallel tempering": ParallelTemperingOptimizer, |
|
35
|
|
|
"Particle swarm optimizer": ParticleSwarmOptimizer, |
|
36
|
|
|
"Evolution strategy": EvolutionStrategyOptimizer, |
|
37
|
|
|
# "Bayesian optimizer": BayesianOptimizer, |
|
38
|
|
|
# "Tree structured parzen estimators": TreeStructuredParzenEstimators, |
|
39
|
|
|
# "Decision tree optimizer": DecisionTreeOptimizer, |
|
40
|
|
|
# "Ensemble optimizer": EnsembleOptimizer, |
|
41
|
|
|
} |
|
42
|
|
|
|
|
43
|
|
|
|
|
44
|
|
|
def objective_function(pos_new): |
|
45
|
|
|
score = -(pos_new["x1"] * pos_new["x1"] + pos_new["x2"] * pos_new["x2"]) |
|
46
|
|
|
return score |
|
47
|
|
|
|
|
48
|
|
|
|
|
49
|
|
|
search_space = {"x1": np.arange(-10, 11, 0.1), "x2": np.arange(-10, 11, 0.1)} |
|
50
|
|
|
|
|
51
|
|
|
runs = 30 |
|
52
|
|
|
|
|
53
|
|
|
|
|
54
|
|
|
def create_performance_data( |
|
55
|
|
|
study_name, objective_function, search_space, n_iter |
|
56
|
|
|
): |
|
57
|
|
|
results = [] |
|
58
|
|
|
|
|
59
|
|
|
for opt_name in tqdm(optimizer_dict.keys()): |
|
60
|
|
|
total_time_list = [] |
|
61
|
|
|
eval_time_list = [] |
|
62
|
|
|
iter_time_list = [] |
|
63
|
|
|
|
|
64
|
|
|
for random_state in tqdm(range(runs)): |
|
65
|
|
|
|
|
66
|
|
|
c_time = time.time() |
|
67
|
|
|
opt = optimizer_dict[opt_name](search_space) |
|
68
|
|
|
opt.search( |
|
69
|
|
|
objective_function, |
|
70
|
|
|
n_iter=n_iter, |
|
71
|
|
|
verbosity=False, |
|
72
|
|
|
random_state=random_state, |
|
73
|
|
|
) |
|
74
|
|
|
|
|
75
|
|
|
total_time = time.time() - c_time |
|
76
|
|
|
eval_time = np.array(opt.eval_times).sum() |
|
77
|
|
|
iter_time = np.array(opt.iter_times).sum() |
|
78
|
|
|
|
|
79
|
|
|
total_time_list.append(total_time) |
|
80
|
|
|
eval_time_list.append(eval_time) |
|
81
|
|
|
iter_time_list.append(iter_time) |
|
82
|
|
|
|
|
83
|
|
|
total_time_mean = np.array(total_time_list).mean() |
|
84
|
|
|
eval_time_mean = np.array(eval_time_list).mean() |
|
85
|
|
|
iter_time_mean = np.array(iter_time_list).mean() |
|
86
|
|
|
|
|
87
|
|
|
total_time_std = np.array(total_time_list).std() |
|
88
|
|
|
eval_time_std = np.array(eval_time_list).std() |
|
89
|
|
|
iter_time_std = np.array(iter_time_list).std() |
|
90
|
|
|
|
|
91
|
|
|
results.append( |
|
92
|
|
|
[ |
|
93
|
|
|
total_time_mean, |
|
94
|
|
|
total_time_std, |
|
95
|
|
|
eval_time_mean, |
|
96
|
|
|
eval_time_std, |
|
97
|
|
|
iter_time_mean, |
|
98
|
|
|
iter_time_std, |
|
99
|
|
|
] |
|
100
|
|
|
) |
|
101
|
|
|
|
|
102
|
|
|
index = [ |
|
103
|
|
|
"total_time_mean", |
|
104
|
|
|
"total_time_std", |
|
105
|
|
|
"eval_time_mean", |
|
106
|
|
|
"eval_time_std", |
|
107
|
|
|
"iter_time_mean", |
|
108
|
|
|
"iter_time_std", |
|
109
|
|
|
] |
|
110
|
|
|
columns = list(optimizer_dict.keys()) |
|
111
|
|
|
|
|
112
|
|
|
results = np.array(results).T |
|
113
|
|
|
results = pd.DataFrame(results, columns=columns, index=index) |
|
114
|
|
|
results.to_csv("./_data/" + study_name + ".csv") |
|
115
|
|
|
|
|
116
|
|
|
|
|
117
|
|
|
create_performance_data( |
|
118
|
|
|
"simple function", objective_function, search_space, n_iter=50 |
|
119
|
|
|
) |
|
120
|
|
|
|
|
121
|
|
|
|
SimonBlanke /
Gradient-Free-Optimizers
