|
1
|
|
|
import random as rnd |
|
2
|
|
|
import copy |
|
3
|
|
|
|
|
4
|
|
|
__all__ = ['ParticleSwarmAlgorithm'] |
|
5
|
|
|
|
|
6
|
|
|
|
|
7
|
|
|
class Particle(object): |
|
8
|
|
|
"""Defines particle for population.""" |
|
9
|
|
|
# pylint: disable=too-many-instance-attributes |
|
10
|
|
|
def __init__(self, D, LB, UB, vMin, vMax): |
|
11
|
|
|
self.D = D # dimension of the problem |
|
12
|
|
|
self.LB = LB # lower bound |
|
13
|
|
|
self.UB = UB # upper bound |
|
14
|
|
|
self.vMin = vMin # velocity min |
|
15
|
|
|
self.vMax = vMax # velocity max |
|
16
|
|
|
self.Solution = [] |
|
17
|
|
|
self.Velocity = [] |
|
18
|
|
|
|
|
19
|
|
|
self.pBestPosition = [] |
|
20
|
|
|
self.pBestSolution = [] |
|
21
|
|
|
self.bestFitness = float('inf') |
|
22
|
|
|
|
|
23
|
|
|
self.Fitness = float('inf') |
|
24
|
|
|
self.generateParticle() |
|
25
|
|
|
|
|
26
|
|
|
|
|
27
|
|
|
def generateParticle(self): |
|
28
|
|
|
self.Solution = [self.LB + (self.UB - self.LB) * rnd.random() |
|
29
|
|
|
for _i in range(self.D)] |
|
30
|
|
|
self.Velocity = [0 for _i in range(self.D)] |
|
31
|
|
|
|
|
32
|
|
|
self.pBestSolution = [0 for _i in range(self.D)] |
|
33
|
|
|
self.bestFitness = float('inf') |
|
34
|
|
|
|
|
35
|
|
|
def evaluate(self): |
|
36
|
|
|
self.Fitness = Particle.FuncEval(self.D, self.Solution) |
|
37
|
|
|
self.checkPersonalBest() |
|
38
|
|
|
|
|
39
|
|
|
def checkPersonalBest(self): |
|
40
|
|
|
if self.Fitness < self.bestFitness: |
|
41
|
|
|
self.pBestSolution = self.Solution |
|
42
|
|
|
self.bestFitness = self.Fitness |
|
43
|
|
|
|
|
|
|
|
|
|
44
|
|
|
|
|
45
|
|
|
def simpleBound(self): |
|
46
|
|
|
for i in range(self.D): |
|
47
|
|
|
if self.Solution[i] < self.LB: |
|
48
|
|
|
self.Solution[i] = self.LB |
|
49
|
|
|
if self.Solution[i] > self.UB: |
|
50
|
|
|
self.Solution[i] = self.UB |
|
51
|
|
|
if self.Velocity[i] < self.vMin: |
|
52
|
|
|
self.Velocity[i] = self.vMin |
|
53
|
|
|
if self.Velocity[i] > self.vMax: |
|
54
|
|
|
self.Velocity[i] = self.vMax |
|
55
|
|
|
|
|
56
|
|
|
def toString(self): |
|
57
|
|
|
pass |
|
58
|
|
|
|
|
59
|
|
|
def __eq__(self, other): |
|
60
|
|
|
return self.Solution == other.Solution and self.Fitness == other.Fitness |
|
61
|
|
|
|
|
62
|
|
|
|
|
63
|
|
|
class ParticleSwarmAlgorithm(object): |
|
64
|
|
|
"""Particle Swarm Optimization algorithm. |
|
65
|
|
|
|
|
66
|
|
|
Date: 12. 2. 2018 |
|
67
|
|
|
|
|
68
|
|
|
Authors : Uros Mlakar |
|
69
|
|
|
|
|
70
|
|
|
License: MIT |
|
71
|
|
|
|
|
72
|
|
|
Reference paper: Kennedy, J. and Eberhart, R. "Particle Swarm Optimization". |
|
73
|
|
|
Proceedings of IEEE International Conference on Neural Networks. IV. pp. 1942--1948, 1995. |
|
74
|
|
|
|
|
75
|
|
|
EDITED: TODO: Tests and validation! Bug in code. |
|
76
|
|
|
""" |
|
77
|
|
|
|
|
78
|
|
View Code Duplication |
def __init__(self, Np, D, nFES, C1, C2, w, vMin, vMax, Lower, Upper, function): |
|
|
|
|
|
|
79
|
|
|
"""Constructor.""" |
|
80
|
|
|
self.Np = Np |
|
81
|
|
|
self.D = D |
|
82
|
|
|
self.C1 = C1 |
|
83
|
|
|
self.C2 = C2 |
|
84
|
|
|
self.w = w |
|
85
|
|
|
self.vMin = vMin |
|
86
|
|
|
self.vMax = vMax |
|
87
|
|
|
self.Lower = Lower |
|
88
|
|
|
self.Upper = Upper |
|
89
|
|
|
self.Swarm = [] |
|
90
|
|
|
self.nFES = nFES |
|
91
|
|
|
self.FEs = 0 |
|
92
|
|
|
self.Done = False |
|
93
|
|
|
Particle.FuncEval = staticmethod(function) |
|
94
|
|
|
|
|
95
|
|
|
self.gBest = Particle(self.D, self.Lower, self.Upper, self.vMin, self.vMax) |
|
96
|
|
|
|
|
97
|
|
|
def evalSwarm(self): |
|
98
|
|
|
for p in self.Swarm: |
|
99
|
|
|
p.evaluate() |
|
100
|
|
|
if p.Fitness < self.gBest.Fitness: |
|
101
|
|
|
self.gBest = copy.deepcopy(p) |
|
102
|
|
|
|
|
103
|
|
|
def initSwarm(self): |
|
104
|
|
|
for _i in range(self.Np): |
|
105
|
|
|
self.Swarm.append(Particle(self.D, self.Lower, self.Upper, self.vMin, self.vMax)) |
|
106
|
|
|
|
|
107
|
|
|
def tryEval(self,p): |
|
|
|
|
|
|
108
|
|
|
if self.FEs <= self.nFES: |
|
109
|
|
|
p.evaluate() |
|
110
|
|
|
self.FEs+=1 |
|
|
|
|
|
|
111
|
|
|
else: |
|
112
|
|
|
self.Done = True |
|
113
|
|
|
|
|
114
|
|
|
def moveSwarm(self, Swarm): |
|
115
|
|
|
MovedSwarm = [] |
|
116
|
|
|
for p in Swarm: |
|
117
|
|
|
|
|
118
|
|
|
part1 = ([(a - b) * rnd.random() * self.C1 for a, |
|
119
|
|
|
b in zip(p.pBestSolution, p.Solution)]) |
|
120
|
|
|
part2 = ([(a - b) * rnd.random() * self.C2 for a, |
|
121
|
|
|
b in zip(self.gBest.Solution, p.Solution)]) |
|
122
|
|
|
|
|
123
|
|
|
p.Velocity = ([self.w * a + b + c for a, b, |
|
124
|
|
|
c in zip(p.Velocity, part1, part2)]) |
|
125
|
|
|
p.Solution = ([a + b for a, b in zip(p.Solution, p.Velocity)]) |
|
126
|
|
|
|
|
127
|
|
|
p.simpleBound() |
|
128
|
|
|
self.tryEval(p) |
|
129
|
|
|
if p.Fitness < self.gBest.Fitness: |
|
130
|
|
|
self.gBest = copy.deepcopy(p) |
|
131
|
|
|
|
|
132
|
|
|
MovedSwarm.append(p) |
|
133
|
|
|
return MovedSwarm |
|
134
|
|
|
|
|
135
|
|
|
def run(self): |
|
136
|
|
|
self.initSwarm() |
|
137
|
|
|
self.evalSwarm() |
|
138
|
|
|
self.FEs += self.Np |
|
139
|
|
|
while not self.Done: |
|
140
|
|
|
MovedSwarm = self.moveSwarm(self.Swarm) |
|
141
|
|
|
self.Swarm = MovedSwarm |
|
142
|
|
|
|
|
143
|
|
|
return self.gBest.Fitness |
|
144
|
|
|
|
NiaOrg /
NiaPy
Loading history...
