|
1
|
|
|
# Author: Simon Blanke |
|
2
|
|
|
# Email: [email protected] |
|
3
|
|
|
# License: MIT License |
|
4
|
|
|
|
|
5
|
|
|
import random |
|
6
|
|
|
import numpy as np |
|
7
|
|
|
|
|
8
|
|
|
from ._evolutionary_algorithm import EvolutionaryAlgorithmOptimizer |
|
9
|
|
|
from ._individual import Individual |
|
10
|
|
|
|
|
11
|
|
|
|
|
12
|
|
|
class DifferentialEvolutionOptimizer(EvolutionaryAlgorithmOptimizer): |
|
13
|
|
|
name = "Differential Evolution" |
|
14
|
|
|
_name_ = "differential_evolution" |
|
15
|
|
|
__name__ = "DifferentialEvolutionOptimizer" |
|
16
|
|
|
|
|
17
|
|
|
optimizer_type = "population" |
|
18
|
|
|
computationally_expensive = False |
|
19
|
|
|
|
|
20
|
|
|
def __init__( |
|
21
|
|
|
self, |
|
22
|
|
|
*args, |
|
23
|
|
|
population=10, |
|
24
|
|
|
mutation_rate=0.9, |
|
25
|
|
|
crossover_rate=0.9, |
|
26
|
|
|
**kwargs |
|
27
|
|
|
): |
|
28
|
|
|
super().__init__(*args, **kwargs) |
|
29
|
|
|
|
|
30
|
|
|
self.population = population |
|
31
|
|
|
self.mutation_rate = mutation_rate |
|
32
|
|
|
self.crossover_rate = crossover_rate |
|
33
|
|
|
|
|
34
|
|
|
self.individuals = self._create_population(Individual) |
|
35
|
|
|
self.optimizers = self.individuals |
|
36
|
|
|
|
|
37
|
|
|
self.offspring_l = [] |
|
38
|
|
|
|
|
39
|
|
|
def mutation(self, f=1): |
|
40
|
|
|
ind_selected = random.sample(self.individuals, 3) |
|
41
|
|
|
|
|
42
|
|
|
x_1, x_2, x_3 = [ind.pos_best for ind in ind_selected] |
|
43
|
|
|
return x_1 + self.mutation_rate * np.subtract(x_2, x_3) |
|
44
|
|
|
|
|
45
|
|
|
def _constraint_loop(self, position): |
|
46
|
|
|
while True: |
|
47
|
|
|
if self.conv.not_in_constraint(position): |
|
48
|
|
|
return position |
|
49
|
|
|
position = self.p_current.move_climb(position, epsilon_mod=0.3) |
|
50
|
|
|
|
|
51
|
|
|
@EvolutionaryAlgorithmOptimizer.track_new_pos |
|
52
|
|
|
def init_pos(self): |
|
53
|
|
|
nth_pop = self.nth_trial % len(self.individuals) |
|
54
|
|
|
|
|
55
|
|
|
self.p_current = self.individuals[nth_pop] |
|
56
|
|
|
return self.p_current.init_pos() |
|
57
|
|
|
|
|
58
|
|
|
@EvolutionaryAlgorithmOptimizer.track_new_pos |
|
59
|
|
|
def iterate(self): |
|
60
|
|
|
self.p_current = self.individuals[ |
|
61
|
|
|
self.nth_trial % len(self.individuals) |
|
62
|
|
|
] |
|
63
|
|
|
target_vector = self.p_current.pos_new |
|
64
|
|
|
|
|
65
|
|
|
mutant_vector = self.mutation() |
|
66
|
|
|
|
|
67
|
|
|
crossover_rates = [1 - self.crossover_rate, self.crossover_rate] |
|
68
|
|
|
pos_new = self.discrete_recombination( |
|
69
|
|
|
[target_vector, mutant_vector], |
|
70
|
|
|
crossover_rates, |
|
71
|
|
|
) |
|
72
|
|
|
pos_new = self.conv2pos(pos_new) |
|
73
|
|
|
pos_new = self._constraint_loop(pos_new) |
|
74
|
|
|
|
|
75
|
|
|
self.p_current.pos_new = self.conv2pos(pos_new) |
|
76
|
|
|
return self.p_current.pos_new |
|
77
|
|
|
|
|
78
|
|
|
@EvolutionaryAlgorithmOptimizer.track_new_score |
|
79
|
|
|
def evaluate(self, score_new): |
|
80
|
|
|
self.p_current.evaluate(score_new) # selection |
|
81
|
|
|
|
SimonBlanke /
Gradient-Free-Optimizers
