|
1
|
|
|
# Author: Simon Blanke |
|
2
|
|
|
# Email: [email protected] |
|
3
|
|
|
# License: MIT License |
|
4
|
|
|
|
|
5
|
|
|
import time |
|
6
|
|
|
import numpy as np |
|
7
|
|
|
import multiprocessing |
|
8
|
|
|
|
|
9
|
|
|
from .verbosity import set_verbosity |
|
10
|
|
|
|
|
11
|
|
|
from .candidate import Candidate |
|
12
|
|
|
|
|
13
|
|
|
|
|
14
|
|
|
from .optimizers import ( |
|
15
|
|
|
HillClimbingOptimizer, |
|
16
|
|
|
StochasticHillClimbingOptimizer, |
|
17
|
|
|
TabuOptimizer, |
|
18
|
|
|
RandomSearchOptimizer, |
|
19
|
|
|
RandomRestartHillClimbingOptimizer, |
|
20
|
|
|
RandomAnnealingOptimizer, |
|
21
|
|
|
SimulatedAnnealingOptimizer, |
|
22
|
|
|
StochasticTunnelingOptimizer, |
|
23
|
|
|
ParallelTemperingOptimizer, |
|
24
|
|
|
ParticleSwarmOptimizer, |
|
25
|
|
|
EvolutionStrategyOptimizer, |
|
26
|
|
|
BayesianOptimizer, |
|
27
|
|
|
TreeStructuredParzenEstimators, |
|
28
|
|
|
DecisionTreeOptimizer, |
|
29
|
|
|
) |
|
30
|
|
|
|
|
31
|
|
|
optimizer_dict = { |
|
32
|
|
|
"HillClimbing": HillClimbingOptimizer, |
|
33
|
|
|
"StochasticHillClimbing": StochasticHillClimbingOptimizer, |
|
34
|
|
|
"TabuSearch": TabuOptimizer, |
|
35
|
|
|
"RandomSearch": RandomSearchOptimizer, |
|
36
|
|
|
"RandomRestartHillClimbing": RandomRestartHillClimbingOptimizer, |
|
37
|
|
|
"RandomAnnealing": RandomAnnealingOptimizer, |
|
38
|
|
|
"SimulatedAnnealing": SimulatedAnnealingOptimizer, |
|
39
|
|
|
"StochasticTunneling": StochasticTunnelingOptimizer, |
|
40
|
|
|
"ParallelTempering": ParallelTemperingOptimizer, |
|
41
|
|
|
"ParticleSwarm": ParticleSwarmOptimizer, |
|
42
|
|
|
"EvolutionStrategy": EvolutionStrategyOptimizer, |
|
43
|
|
|
"Bayesian": BayesianOptimizer, |
|
44
|
|
|
"TPE": TreeStructuredParzenEstimators, |
|
45
|
|
|
"DecisionTree": DecisionTreeOptimizer, |
|
46
|
|
|
} |
|
47
|
|
|
|
|
48
|
|
|
|
|
49
|
|
|
class Search: |
|
50
|
|
|
def __init__(self, _main_args_, _opt_args_): |
|
51
|
|
|
self._main_args_ = _main_args_ |
|
52
|
|
|
self._opt_args_ = _opt_args_ |
|
53
|
|
|
|
|
54
|
|
|
self._info_, _pbar_ = set_verbosity(_main_args_.verbosity) |
|
55
|
|
|
self._pbar_ = _pbar_() |
|
56
|
|
|
|
|
57
|
|
|
self.optimizer = optimizer_dict[self._main_args_.optimizer](_opt_args_) |
|
58
|
|
|
|
|
59
|
|
|
self.optimizer._pbar_ = self._pbar_ |
|
60
|
|
|
|
|
61
|
|
|
def search(self, nth_process=0, rayInit=False): |
|
62
|
|
|
self.start_time = time.time() |
|
63
|
|
|
self.results = {} |
|
64
|
|
|
self.eval_times = {} |
|
65
|
|
|
self.iter_times = {} |
|
66
|
|
|
self.best_scores = {} |
|
67
|
|
|
self.pos_list = {} |
|
68
|
|
|
self.score_list = {} |
|
69
|
|
|
|
|
70
|
|
|
if rayInit: |
|
71
|
|
|
self._run_job(nth_process) |
|
72
|
|
|
elif self._main_args_.n_jobs == 1: |
|
73
|
|
|
self._run_job(nth_process) |
|
74
|
|
|
else: |
|
75
|
|
|
self._run_multiple_jobs() |
|
76
|
|
|
|
|
77
|
|
|
return ( |
|
78
|
|
|
self.results, |
|
79
|
|
|
self.pos_list, |
|
80
|
|
|
self.score_list, |
|
81
|
|
|
self.eval_times, |
|
82
|
|
|
self.iter_times, |
|
83
|
|
|
self.best_scores, |
|
84
|
|
|
) |
|
85
|
|
|
|
|
86
|
|
|
def _search_multiprocessing(self): |
|
87
|
|
|
"""Wrapper for the parallel search. Passes integer that corresponds to process number""" |
|
88
|
|
|
pool = multiprocessing.Pool(self._main_args_.n_jobs) |
|
89
|
|
|
_cand_list, _p_list = zip( |
|
90
|
|
|
*pool.map(self._search, self._main_args_._n_process_range) |
|
91
|
|
|
) |
|
92
|
|
|
|
|
93
|
|
|
return _cand_list, _p_list |
|
94
|
|
|
|
|
95
|
|
|
def _run_job(self, nth_process): |
|
96
|
|
|
_cand_, _p_ = self._search(nth_process) |
|
97
|
|
|
self._get_attributes(_cand_, _p_) |
|
98
|
|
|
|
|
99
|
|
|
def _get_attributes(self, _cand_, _p_): |
|
100
|
|
|
self.results[_cand_.func_] = _cand_._process_results(self._opt_args_) |
|
101
|
|
|
self.eval_times[_cand_.func_] = _cand_.eval_time |
|
102
|
|
|
self.iter_times[_cand_.func_] = _cand_.iter_times |
|
103
|
|
|
self.best_scores[_cand_.func_] = _cand_.score_best |
|
104
|
|
|
|
|
105
|
|
|
if isinstance(_p_, list): |
|
106
|
|
|
self.pos_list[_cand_.func_] = [np.array(p.pos_list) for p in _p_] |
|
107
|
|
|
self.score_list[_cand_.func_] = [np.array(p.score_list) for p in _p_] |
|
108
|
|
|
else: |
|
109
|
|
|
self.pos_list[_cand_.func_] = [np.array(_p_.pos_list)] |
|
110
|
|
|
self.score_list[_cand_.func_] = [np.array(_p_.score_list)] |
|
111
|
|
|
|
|
112
|
|
|
def _run_multiple_jobs(self): |
|
113
|
|
|
_cand_list, _p_list = self._search_multiprocessing() |
|
114
|
|
|
|
|
115
|
|
|
for _ in range(int(self._main_args_.n_jobs / 2) + 2): |
|
116
|
|
|
print("\n") |
|
117
|
|
|
|
|
118
|
|
|
for _cand_, _p_ in zip(_cand_list, _p_list): |
|
119
|
|
|
self._get_attributes(_cand_, _p_) |
|
120
|
|
|
|
|
121
|
|
|
def _search(self, nth_process): |
|
122
|
|
|
_cand_ = self._initialize_search(self._main_args_, nth_process, self._info_) |
|
123
|
|
|
|
|
124
|
|
|
for i in range(self._main_args_.n_iter): |
|
125
|
|
|
c_time = time.time() |
|
126
|
|
|
|
|
127
|
|
|
_cand_.i = i |
|
128
|
|
|
_cand_ = self.optimizer.iterate(i, _cand_) |
|
129
|
|
|
|
|
130
|
|
|
if self._time_exceeded(): |
|
131
|
|
|
break |
|
132
|
|
|
|
|
133
|
|
|
_cand_.iter_times.append(time.time() - c_time) |
|
134
|
|
|
|
|
135
|
|
|
self.optimizer._finish_search() |
|
136
|
|
|
|
|
137
|
|
|
return _cand_, self.optimizer.p_list |
|
138
|
|
|
|
|
139
|
|
|
def _time_exceeded(self): |
|
140
|
|
|
run_time = time.time() - self.start_time |
|
141
|
|
|
return self._main_args_.max_time and run_time > self._main_args_.max_time |
|
142
|
|
|
|
|
143
|
|
|
def _initialize_search(self, _main_args_, nth_process, _info_): |
|
144
|
|
|
_main_args_._set_random_seed(nth_process) |
|
145
|
|
|
|
|
146
|
|
|
_cand_ = Candidate(nth_process, _main_args_, _info_) |
|
147
|
|
|
self._pbar_.init_p_bar(nth_process, self._main_args_) |
|
148
|
|
|
|
|
149
|
|
|
return _cand_ |
|
150
|
|
|
|
SimonBlanke /
Hyperactive
