|
1
|
|
|
"""Firefly algorithm. |
|
2
|
|
|
|
|
3
|
|
|
Date: 2016 |
|
4
|
|
|
|
|
5
|
|
|
Authors : Iztok Fister Jr. and Iztok Fister |
|
6
|
|
|
License: MIT |
|
7
|
|
|
|
|
8
|
|
|
Reference paper: Fister, I., Fister Jr, I., Yang, X. S., & Brest, J. (2013). |
|
9
|
|
|
A comprehensive review of firefly algorithms. Swarm and Evolutionary Computation, 13, 34-46. |
|
10
|
|
|
""" |
|
11
|
|
|
|
|
12
|
|
|
import random |
|
13
|
|
|
import math |
|
14
|
|
|
from NiaPy.benchmarks.utility import Utility |
|
15
|
|
|
|
|
16
|
|
|
__all__ = ['FireflyAlgorithm'] |
|
17
|
|
|
|
|
18
|
|
|
|
|
19
|
|
|
class FireflyAlgorithm(object): |
|
20
|
|
|
"""Firefly Algorithm implementation.""" |
|
21
|
|
|
|
|
22
|
|
|
# pylint: disable=too-many-instance-attributes |
|
23
|
|
View Code Duplication |
def __init__(self, D, NP, nFES, alpha, betamin, gamma, Lower, Upper, function): |
|
|
|
|
|
|
24
|
|
|
self.D = D # dimension of the problem |
|
25
|
|
|
self.NP = NP # population size |
|
26
|
|
|
self.nFES = nFES # number of function evaluations |
|
27
|
|
|
self.alpha = alpha # alpha parameter |
|
28
|
|
|
self.betamin = betamin # beta parameter |
|
29
|
|
|
self.gamma = gamma # gamma parameter |
|
30
|
|
|
# sort of fireflies according to fitness value |
|
31
|
|
|
self.Index = [0] * self.NP |
|
32
|
|
|
self.Fireflies = [[0 for _i in range(self.D)] |
|
33
|
|
|
for _j in range(self.NP)] # firefly agents |
|
34
|
|
|
self.Fireflies_tmp = [[0 for _i in range(self.D)] for _j in range( |
|
35
|
|
|
self.NP)] # intermediate pop |
|
36
|
|
|
self.Fitness = [0.0] * self.NP # fitness values |
|
37
|
|
|
self.Intensity = [0.0] * self.NP # light intensity |
|
38
|
|
|
self.nbest = [0.0] * self.NP # the best solution found so far |
|
39
|
|
|
self.Lower = Lower # lower bound |
|
40
|
|
|
self.Upper = Upper # upper bound |
|
41
|
|
|
self.fbest = None # the best |
|
42
|
|
|
self.evaluations = 0 |
|
43
|
|
|
self.Fun = Utility.itialize_benchmark(function) |
|
44
|
|
|
|
|
45
|
|
|
def init_ffa(self): |
|
46
|
|
|
for i in range(self.NP): |
|
47
|
|
|
for j in range(self.D): |
|
48
|
|
|
self.Fireflies[i][j] = random.uniform( |
|
49
|
|
|
0, 1) * (self.Upper - self.Lower) + self.Lower |
|
50
|
|
|
self.Fitness[i] = 1.0 # initialize attractiveness |
|
51
|
|
|
self.Intensity[i] = self.Fitness[i] |
|
52
|
|
|
|
|
53
|
|
|
def alpha_new(self, a): |
|
54
|
|
|
delta = 1.0 - math.pow((math.pow(10.0, -4.0) / 0.9), 1.0 / float(a)) |
|
55
|
|
|
return (1 - delta) * self.alpha |
|
56
|
|
|
|
|
57
|
|
|
def sort_ffa(self): # implementation of bubble sort |
|
58
|
|
|
for i in range(self.NP): |
|
59
|
|
|
self.Index[i] = i |
|
60
|
|
|
|
|
61
|
|
|
for i in range(0, (self.NP - 1)): |
|
62
|
|
|
j = i + 1 |
|
63
|
|
|
for j in range(j, self.NP): |
|
64
|
|
|
if self.Intensity[i] > self.Intensity[j]: |
|
65
|
|
|
z = self.Intensity[i] # exchange attractiveness |
|
66
|
|
|
self.Intensity[i] = self.Intensity[j] |
|
67
|
|
|
self.Intensity[j] = z |
|
68
|
|
|
z = self.Fitness[i] # exchange fitness |
|
69
|
|
|
self.Fitness[i] = self.Fitness[j] |
|
70
|
|
|
self.Fitness[j] = z |
|
71
|
|
|
z = self.Index[i] # exchange indexes |
|
72
|
|
|
self.Index[i] = self.Index[j] |
|
73
|
|
|
self.Index[j] = z |
|
74
|
|
|
|
|
75
|
|
|
def replace_ffa(self): # replace the old population according to the new Index values |
|
76
|
|
|
# copy original population to a temporary area |
|
77
|
|
|
for i in range(self.NP): |
|
78
|
|
|
for j in range(self.D): |
|
79
|
|
|
self.Fireflies_tmp[i][j] = self.Fireflies[i][j] |
|
80
|
|
|
|
|
81
|
|
|
# generational selection in the sense of an EA |
|
82
|
|
|
for i in range(self.NP): |
|
83
|
|
|
for j in range(self.D): |
|
84
|
|
|
self.Fireflies[i][j] = self.Fireflies_tmp[self.Index[i]][j] |
|
85
|
|
|
|
|
86
|
|
|
def FindLimits(self, k): |
|
87
|
|
|
for i in range(self.D): |
|
88
|
|
|
if self.Fireflies[k][i] < self.Lower: |
|
89
|
|
|
self.Fireflies[k][i] = self.Lower |
|
90
|
|
|
if self.Fireflies[k][i] > self.Upper: |
|
91
|
|
|
self.Fireflies[k][i] = self.Upper |
|
92
|
|
|
|
|
93
|
|
|
def move_ffa(self): |
|
94
|
|
|
for i in range(self.NP): |
|
95
|
|
|
scale = abs(self.Upper - self.Lower) |
|
96
|
|
|
for j in range(self.NP): |
|
97
|
|
|
r = 0.0 |
|
98
|
|
|
for k in range(self.D): |
|
99
|
|
|
r += (self.Fireflies[i][k] - self.Fireflies[j][k]) * \ |
|
100
|
|
|
(self.Fireflies[i][k] - self.Fireflies[j][k]) |
|
101
|
|
|
r = math.sqrt(r) |
|
102
|
|
|
if self.Intensity[i] > self.Intensity[j]: # brighter and more attractive |
|
103
|
|
|
beta0 = 1.0 |
|
104
|
|
|
beta = (beta0 - self.betamin) * \ |
|
105
|
|
|
math.exp(-self.gamma * math.pow(r, 2.0)) + self.betamin |
|
106
|
|
|
for k in range(self.D): |
|
107
|
|
|
r = random.uniform(0, 1) |
|
108
|
|
|
tmpf = self.alpha * (r - 0.5) * scale |
|
109
|
|
|
self.Fireflies[i][k] = self.Fireflies[i][ |
|
110
|
|
|
k] * (1.0 - beta) + self.Fireflies_tmp[j][k] * beta + tmpf |
|
111
|
|
|
self.FindLimits(i) |
|
112
|
|
|
|
|
113
|
|
|
def run(self): |
|
114
|
|
|
self.init_ffa() |
|
115
|
|
|
|
|
116
|
|
|
while True: |
|
117
|
|
|
if self.evaluations == self.nFES: |
|
118
|
|
|
break |
|
119
|
|
|
|
|
120
|
|
|
# optional reducing of alpha |
|
121
|
|
|
self.alpha = self.alpha_new(self.nFES / self.NP) |
|
122
|
|
|
|
|
123
|
|
|
# evaluate new solutions |
|
124
|
|
|
for i in range(self.NP): |
|
125
|
|
|
self.Fitness[i] = self.Fun(self.D, self.Fireflies[i]) |
|
126
|
|
|
self.evaluations = self.evaluations + 1 |
|
127
|
|
|
self.Intensity[i] = self.Fitness[i] |
|
128
|
|
|
|
|
129
|
|
|
# ranking fireflies by their light intensity |
|
130
|
|
|
self.sort_ffa() |
|
131
|
|
|
# replace old population |
|
132
|
|
|
self.replace_ffa() |
|
133
|
|
|
# find the current best |
|
134
|
|
|
self.fbest = self.Intensity[0] |
|
135
|
|
|
# move all fireflies to the better locations |
|
136
|
|
|
self.move_ffa() |
|
137
|
|
|
|
|
138
|
|
|
return self.fbest |
|
139
|
|
|
|
NiaOrg /
NiaPy
Loading history...