|
1
|
|
|
# Author: Simon Blanke |
|
2
|
|
|
# Email: [email protected] |
|
3
|
|
|
# License: MIT License |
|
4
|
|
|
|
|
5
|
|
|
|
|
6
|
|
|
import numpy as np |
|
7
|
|
|
|
|
8
|
|
|
from .base_population_optimizer import BasePopulationOptimizer |
|
9
|
|
|
from ...search import Search |
|
10
|
|
|
from ._particle import Particle |
|
11
|
|
|
|
|
12
|
|
|
|
|
13
|
|
|
class ParticleSwarmOptimizer(BasePopulationOptimizer, Search): |
|
14
|
|
|
def __init__( |
|
15
|
|
|
self, |
|
16
|
|
|
search_space, |
|
17
|
|
|
inertia=0.5, |
|
18
|
|
|
cognitive_weight=0.5, |
|
19
|
|
|
social_weight=0.5, |
|
20
|
|
|
temp_weight=0.2, |
|
21
|
|
|
rand_rest_p=0.03, |
|
22
|
|
|
): |
|
23
|
|
|
super().__init__(search_space) |
|
24
|
|
|
|
|
25
|
|
|
self.inertia = inertia |
|
26
|
|
|
self.cognitive_weight = cognitive_weight |
|
27
|
|
|
self.social_weight = social_weight |
|
28
|
|
|
self.temp_weight = temp_weight |
|
29
|
|
|
self.rand_rest_p = rand_rest_p |
|
30
|
|
|
|
|
31
|
|
|
self.particles = self.optimizers |
|
32
|
|
|
|
|
33
|
|
|
def _sort_best(self): |
|
34
|
|
|
scores_list = [] |
|
35
|
|
|
for _p_ in self.particles: |
|
36
|
|
|
scores_list.append(_p_.score_current) |
|
37
|
|
|
|
|
38
|
|
|
scores_np = np.array(scores_list) |
|
39
|
|
|
idx_sorted_ind = list(scores_np.argsort()[::-1]) |
|
40
|
|
|
|
|
41
|
|
|
self.p_sorted = [self.particles[i] for i in idx_sorted_ind] |
|
42
|
|
|
|
|
43
|
|
|
def init_pos(self, pos): |
|
44
|
|
|
particle = Particle( |
|
45
|
|
|
self.search_space, |
|
46
|
|
|
inertia=self.inertia, |
|
47
|
|
|
cognitive_weight=self.cognitive_weight, |
|
48
|
|
|
social_weight=self.social_weight, |
|
49
|
|
|
temp_weight=self.temp_weight, |
|
50
|
|
|
rand_rest_p=self.rand_rest_p, |
|
51
|
|
|
) |
|
52
|
|
|
self.particles.append(particle) |
|
53
|
|
|
particle.init_pos(pos) |
|
54
|
|
|
|
|
55
|
|
|
self.p_current = particle |
|
56
|
|
|
self.p_current.velo = np.zeros(len(self.max_positions)) |
|
57
|
|
|
|
|
58
|
|
|
def iterate(self): |
|
59
|
|
|
n_iter = self._iterations(self.particles) |
|
60
|
|
|
self.p_current = self.particles[n_iter % len(self.particles)] |
|
61
|
|
|
|
|
62
|
|
|
self._sort_best() |
|
63
|
|
|
self.p_current.global_pos_best = self.p_sorted[0].pos_best |
|
64
|
|
|
|
|
65
|
|
|
pos = self.p_current.iterate() |
|
66
|
|
|
|
|
67
|
|
|
return pos |
|
68
|
|
|
|
|
69
|
|
|
def evaluate(self, score_new): |
|
70
|
|
|
self.p_current.evaluate(score_new) |
|
71
|
|
|
|
|
72
|
|
|
|
SimonBlanke /
Gradient-Free-Optimizers
