forked from mmssouza/height_functions
/
optimize.py
485 lines (424 loc) · 15 KB
/
optimize.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
import sys
import cPickle
import os.path
import scipy
import math
import numpy as np
from numpy.random import seed,random_integers,rand,randn,permutation
Dim = 3
def set_dim(d):
setattr(sys.modules[__name__],"Dim",d)
class sim_ann:
arg_lim = [(35, 200),(1,15),(0.001,1.),(1,50)]
def __gera_s0(self):
l = []
l.append(random_integers(self.arg_lim[0][0],self.arg_lim[0][1]))
l.append(random_integers(self.arg_lim[1][0],self.arg_lim[1][1]))
l.append(self.arg_lim[2][0]+ (self.arg_lim[2][1] - self.arg_lim[2][0])*rand())
l.append(random_integers(self.arg_lim[3][0],self.arg_lim[3][1]))
return l
def __init__(self,f,T0,alpha,P,L):
seed()
self.f = f
if os.path.isfile("dump_sim_ann.pkl"):
dump_fd = open("dump_sim_ann.pkl",'r')
self.s = cPickle.load(dump_fd)
self.T = cPickle.load(dump_fd)
self.fit = cPickle.load(dump_fd)
self.hall_of_fame = cPickle.load(dump_fd)
else:
self.s = self.__gera_s0()
self.T = T0
self.fit = self.f(self.s)
self.hall_of_fame = []
for i in scipy.arange(15):
self.hall_of_fame.insert(0,scipy.hstack((self.fit,self.s)))
self.P = P
self.L = L
self.alpha = alpha
def Perturba(self,x,f):
for i in range(len(x)):
if scipy.rand() < 0.5:
aux = x[i]
if type(aux) == float:
x[i] = x[i] + x[i]*scipy.randn()
else:
delta = int(round(x[i]*2*rand()))
x[i] = random_integers(x[i]-delta,x[i]+delta)
if not (self.arg_lim[i][0] <= x[i] <= self.arg_lim[i][1]):
x[i] = aux
return x
def run(self):
i = 1
self.nS = 0
while (True):
si = self.Perturba(list(self.s),self.fit)
aux = self.f(si)
delta = aux - self.fit
if (delta < 0) or (math.exp(-delta/self.T) > scipy.rand()):
self.s = list(si);
self.fit = aux
self.nS = self.nS + 1
i = i + 1
if (i > self.P) or (self.nS > self.L):
k = 0
if self.nS > 0:
while (self.fit > self.hall_of_fame[k][0]):
k = k + 1
if k == 15:
break
if k < 15:
self.hall_of_fame.insert(k,scipy.hstack((self.fit,self.s)))
self.hall_of_fame.pop()
break
self.T = self.alpha*self.T
dump_fd = open("dump_sim_ann.pkl","wb")
cPickle.dump(self.s,dump_fd)
cPickle.dump(self.T,dump_fd)
cPickle.dump(self.fit,dump_fd)
cPickle.dump(self.hall_of_fame,dump_fd)
dump_fd.close()
class coevol:
arg_lim = [(5., 100.),(0.,.2),(.6,1.),(30,512)]
def __init__(self,challenge_func,ns = 10,npop1 = 20,pr = 0.3,beta = 0.85,npop2 = 20,w = 0.7,c1 = 1.5,c2 = 1.5):
# Tamanho das populacoes
seed()
self.ns = ns
self.npop1 = npop1
self.npop2 = npop2
# Parametros do DE
self.beta = beta
self.pr = pr
# Parametros do PSO
self.c1 = c1
self.c2 = c2
self.w = w
# Funcao que representa problema desafio
self.fc = challenge_func
# Respostas do problema desafio
#self.pso = pso(fitness_func = challenge_func,npop = npop2,w = w,c1 = c1,c2 = c2)
self.ans1 = scipy.zeros(self.npop1)
self.ans2 = scipy.zeros(self.npop2)
# Populacoes
self.pop1 = []
self.pop2 = []
# Gera pop1 e pop2 e resolve problema desafio
for i in scipy.arange(self.npop1):
self.ans1[i],aux = self.resolve_desafio(self.gera_individuo())
self.pop1.append(aux.copy())
for i in scipy.arange(self.npop2):
self.ans2[i],aux = self.resolve_desafio(self.gera_individuo())
self.pop2.append(aux.copy())
self.pop1 = scipy.array(self.pop1)
self.pop2 = scipy.array(self.pop2)
self.hall_of_fame1 = []
for i in scipy.arange(15):
self.hall_of_fame1.insert(0,scipy.hstack((self.ans1.min(),self.pop1[self.ans1.argmin()])))
self.hall_of_fame2 = []
for i in scipy.arange(15):
#self.hall_of_fame2.insert(0,scipy.hstack((self.pso.fit[0],self.pso.pop[0])))
self.hall_of_fame2.insert(0,scipy.hstack((self.ans2.min(),self.pop2[self.ans2.argmin()])))
# Funcoes fitness das populacoes
self.fit1 = scipy.zeros(self.npop1)
self.fit2 = scipy.zeros(self.npop2)
for i in scipy.arange(self.npop2):
self.fit2[i] = self.avalia_aptidao2(self.ans2[i])
for i in scipy.arange(self.npop1):
self.fit1[i] = self.avalia_aptidao1(self.ans1[i])
# inicializa velocidades iniciais do PSO
self.v = scipy.zeros(self.pop2.shape)
# guarda o melhor fitness de cada particula PSO
self.bfp = scipy.copy(self.pop2)
self.bfp_fitness = scipy.copy(self.fit2)
self.bfp_ans = scipy.copy(self.ans2)
# guarda o melhor fitness global PSO
self.bfg = self.pop2[self.bfp_fitness.argmax()].copy()
self.bfg_fitness = self.bfp_fitness.max().copy()
self.bfg_ans = self.bfp_ans[self.bfp_fitness.argmax()].copy()
def gera_individuo(self):
l = []
l.append(random_integers(self.arg_lim[0][0],self.arg_lim[0][1]))
l.append(self.arg_lim[1][0]+ (self.arg_lim[1][1] - self.arg_lim[1][0])*rand())
l.append(self.arg_lim[2][0]+ (self.arg_lim[2][1] - self.arg_lim[2][0])*rand())
l.append(random_integers(self.arg_lim[3][0],self.arg_lim[3][1]))
return np.array(l)
def resolve_desafio(self,x):
if not self.arg_lim[0][0] <= x[0] <= self.arg_lim[0][1]:
x[0] = random_integers(self.arg_lim[0][0],self.arg_lim[0][1])
if not self.arg_lim[1][0] <= x[1] <= self.arg_lim[1][1]:
x[1] = self.arg_lim[1][0]+ (self.arg_lim[1][1] - self.arg_lim[1][0])*rand()
if not self.arg_lim[2][0] <= x[2] <= self.arg_lim[2][1]:
x[2] = self.arg_lim[2][0]+ (self.arg_lim[2][1] - self.arg_lim[2][0])*rand()
if not self.arg_lim[3][0] <= x[3] <= self.arg_lim[3][1]:
x[0] = random_integers(self.arg_lim[3][0],self.arg_lim[3][1])
return (self.fc(x),x)
def avalia_aptidao2(self,x):
cnt = 0
i = permutation(self.npop1)[0:self.ns]
for a in self.ans1[i]:
if x < a:
cnt = cnt+10*(a - x)
else:
cnt = cnt + 5*(a-x)
for a in scipy.array(self.hall_of_fame1)[:,0]:
if x < a:
cnt = cnt + 20*(a - x)
else:
cnt = cnt + 5*(a-x)
return cnt
def avalia_aptidao1(self,x):
cnt = 0
i = permutation(self.npop2)[0:self.ns]
for a in self.ans2[i]:
if x<a:
cnt = cnt + 10*(a - x)
else:
cnt = cnt + 5*(a-x)
for a in scipy.array(self.hall_of_fame2)[:,0]:
if x<a:
cnt = cnt + 20*(a - x)
else:
cnt = cnt + 5*(a-x)
return cnt
def HF1_Updt(self,x,y):
# Hall of fame
k = 0
while (x > self.hall_of_fame1[k][0]):
k = k + 1
if k == 15:
break
if k < 15 and not (x == self.hall_of_fame1[k][0]):
self.hall_of_fame1.insert(k,scipy.hstack((x,y)))
self.hall_of_fame1.pop()
def HF2_Updt(self,x,y):
# Hall of fame
k = 0
while (x > self.hall_of_fame2[k][0]):
k = k + 1
if k == 15:
break
if k < 15 and not (x == self.hall_of_fame2[k][0]):
self.hall_of_fame2.insert(k,scipy.hstack((x,y)))
self.hall_of_fame2.pop()
def Evolve_DE(self):
for i in scipy.arange(self.npop1):
# para cada individuo da populacao
# gera trial vector usado para perturbar individuo atual (indice i)
# a partir de 3 individuos escolhidos aleatoriamente na populacao e
# cujos indices sejam distintos e diferentes de i
invalido = True
while invalido:
j = random_integers(0,self.npop1-1,3)
invalido = (i in j)
invalido = invalido or (j[0] == j[1])
invalido = invalido or (j[1] == j[2])
invalido = invalido or (j[2] == j[0])
# trial vector a partir da mutacao de um alvo
u = self.pop1[j[0]] + self.beta*(self.pop1[j[1]] - self.pop1[j[2]])
# gera por crossover solucao candidata
c = self.pop1[i].copy()
# seleciona indices para crossover
# garantindo que ocorra crossover em
# pelo menos uma vez
j = random_integers(0,self.pop1.shape[1]-1)
for k in scipy.arange(self.pop1.shape[1]):
if (scipy.rand() < self.pr) or (k == j):
c[k] = u[k]
ans,c = self.resolve_desafio(c)
c_fit = self.avalia_aptidao1(ans)
# leva para proxima geracao quem tiver melhor fitness
if (c_fit > self.fit1[i]):
self.pop1[i] = c
self.fit1[i] = c_fit
self.ans1[i] = ans
def Evolve_PSO(self):
for i in scipy.arange(self.npop2):
# Atualiza velocidade
self.v[i] = self.w*self.v[i]
self.v[i] = self.v[i] + self.c1*scipy.rand()*( self.bfp[i] - self.pop2[i])
self.v[i] = self.v[i] + self.c2*scipy.rand()*(self.bfg - self.pop2[i])
for j in range(self.v.shape[1]):
if self.v[i][j] >= self.arg_lim[j][1]/2:
self.v[i][j] = self.arg_lim[j][1]/2
elif self.v[i][j] <= -self.arg_lim[j][1]/2:
self.v[i][j] = -self.arg_lim[j][1]/2
# Atualiza posicao
self.pop2[i] = self.pop2[i] + self.v[i]
self.ans2[i],self.pop2[i] = self.resolve_desafio(self.pop2[i])
self.fit2[i] = self.avalia_aptidao2(self.ans2[i])
self.bfp_fitness[i] = self.avalia_aptidao2(self.bfp_ans[i])
self.bfg_fitness = self.avalia_aptidao2(self.bfg_ans)
# Atualiza melhor posicao da particula
if (self.fit2[i] > self.bfp_fitness[i]):
self.bfp[i] = self.pop2[i]
self.bfp_fitness[i] = self.fit2[i]
self.bfp_ans[i] = self.ans2[i]
# Atualiza melhor posicao global
if (self.bfp_fitness[i] > self.bfg_fitness):
self.bfg_fitness = self.bfp_fitness[i].copy()
self.bfg = self.bfp[i].copy()
self.bfg_ans = self.bfp_ans[i].copy()
def run(self):
for i in scipy.arange(self.npop1):
self.fit1[i] = self.avalia_aptidao1(self.ans1[i])
self.Evolve_DE()
self.HF1_Updt(self.ans1[self.fit1.argmax()],self.pop1[self.fit1.argmax()])
self.Evolve_PSO()
self.HF2_Updt(self.bfg_ans,self.bfg)
class de:
arg_lim = [(5., 150.),(0.,.2),(.6,1.),(32,512)]
def __init__(self,fitness_func,npop = 10,pr = 0.7,beta = 2.5,debug=False):
seed()
self.ns = npop
self.beta = beta
self.pr = pr
self.debug = debug
self.fitness_func = fitness_func
self.fit = scipy.zeros((self.ns,1))
self.pop = []
# avalia fitness de toda populacao
for i in scipy.arange(self.ns):
self.fit[i],aux = self.avalia_aptidao(self.gera_individuo())
#print i,self.fit[i]
self.pop.append(aux.copy())
self.pop = scipy.array(self.pop)
def gera_individuo(self):
l = []
l.append(random_integers(self.arg_lim[0][0],self.arg_lim[0][1]))
l.append(self.arg_lim[1][0]+ (self.arg_lim[1][1] - self.arg_lim[1][0])*rand())
l.append(self.arg_lim[2][0]+ (self.arg_lim[2][1] - self.arg_lim[2][0])*rand())
l.append(random_integers(self.arg_lim[3][0],self.arg_lim[3][1]))
return np.array(l)
def avalia_aptidao(self,x):
if not self.arg_lim[0][0] <= x[0] <= self.arg_lim[0][1]:
x[0] = random_integers(self.arg_lim[0][0],self.arg_lim[0][1])
if not self.arg_lim[1][0] <= x[1] <= self.arg_lim[1][1]:
x[1] = self.arg_lim[1][0]+ (self.arg_lim[1][1] - self.arg_lim[1][0])*rand()
if not self.arg_lim[2][0] <= x[2] <= self.arg_lim[2][1]:
x[2] = self.arg_lim[2][0]+ (self.arg_lim[2][1] - self.arg_lim[2][0])*rand()
if not self.arg_lim[3][0] <= x[3] <= self.arg_lim[3][1]:
x[3] = random_integers(self.arg_lim[3][0],self.arg_lim[3][1])
return (self.fitness_func(x),x)
def run(self):
#prox_geracao = []
for i in scipy.arange(self.ns):
if self.debug: print "i = {0}".format(i)
# para cada individuo da populacao
# gera trial vector usado para perturbar individuo atual (indice i)
# a partir de 3 individuos escolhidos aleatoriamente na populacao e
# cujos indices sejam distintos e diferentes de i
invalido = True
while invalido:
j = random_integers(0,self.ns-1,3)
invalido = (i in j)
invalido = invalido or (j[0] == j[1])
invalido = invalido or (j[1] == j[2])
invalido = invalido or (j[2] == j[0])
if self.debug: print "j (mutacao) = {0}".format(j)
if self.debug: print "invalido flag = {0}".format(invalido)
if self.debug: print "j (mutacao) = {0}".format(j)
# trial vector a partir da mutacao de um alvo
u = self.pop[j[0]] + self.beta*(self.pop[j[1]] - self.pop[j[2]])
if self.debug: print "u (target vector) = {0}".format(u)
# gera por crossover solucao candidata
c = self.pop[i].copy()
# seleciona indices para crossover
# garantindo que ocorra crossover em
# pelo menos uma vez
j = random_integers(0,self.pop.shape[1]-1)
for k in scipy.arange(self.pop.shape[1]):
if (scipy.rand() < self.pr) or (k == j):
c[k] = u[k]
c_fit,c = self.avalia_aptidao(c)
if self.debug: print "atual = {0}, fitness = {1}".format(self.pop[i],self.fit[i])
if self.debug: print "candidato = {0}, fitness = {1}".format(c,c_fit)
# leva para proxima geracao quem tiver melhor fitness
if c_fit < self.fit[i]:
self.pop[i] = c
self.fit[i] = c_fit
# avalia fitness da nova populacao
# for i in scipy.arange(self.ns):
# self.fit[i],self.pop[i] = self.avalia_aptidao(self.pop[i])
class pso:
def __init__(self,fitness_func,npop = 20,w = 0.5,c1 = 2.01,c2 = 2.02,debug = False):
seed()
self.debug = debug
self.c1 = c1
self.c2 = c2
self.w = w
self.ns = npop
# gera pop inicial
# centroides dos Kmax agrupamentos
self.pop = scipy.array([self.gera_individuo() for i in scipy.arange(self.ns)])
self.fitness_func = fitness_func
self.fit = scipy.zeros(self.ns)
# avalia fitness de toda populacao
for i in scipy.arange(self.ns):
self.fit[i],self.pop[i] = self.avalia_aptidao(self.pop[i])
# inicializa velocidades iniciais
self.v = scipy.zeros((self.ns,Dim))
# guarda a melhor posicao de cada particula
self.bfp = scipy.copy(self.pop)
self.bfp_fitness = scipy.copy(self.fit)
# guarda a melhor posicao global
self.bfg = self.pop[self.bfp_fitness.argmin()].copy()
self.bfg_fitness = self.bfp_fitness.min().copy()
def gera_individuo(self):
l = []
l.append(random_integers(3,200))
l.append(0.1+ 0.2*rand())
l.append(0.6+0.4*rand())
return np.array(l)
def avalia_aptidao(self,x):
if not 3 <= x[0] <= 200:
x[0] = random_integers(3,200)
if not 0. <= x[1] <= .3:
x[1] = 0.1+0.2*rand()
if not 0.6 <= x[2] <= 1.:
x[2] = 0.6+0.4*rand()
return (self.fitness_func(x),x)
def run(self):
for i in scipy.arange(self.ns):
# Atualiza velocidade
self.v[i] = self.w*self.v[i]
self.v[i] = self.v[i] + self.c1*scipy.rand()*( self.bfp[i] - self.pop[i])
self.v[i] = self.v[i] + self.c2*scipy.rand()*(self.bfg - self.pop[i])
for j in range(Dim):
if self.v[i][j] >= 52.6:
self.v[i][j] = 52.6
elif self.v[i][j] <= -52.6:
self.v[i][j] = -52.6
# Atualiza posicao
self.pop[i] = self.pop[i] + self.v[i]
self.fit[i],self.pop[i] = self.avalia_aptidao(self.pop[i])
# Atualiza melhor posicao da particula
if self.fit[i] < self.bfp_fitness[i]:
self.bfp[i] = self.pop[i]
self.bfp_fitness[i] = self.fit[i]
if self.debug:
print "self.fit[{0}] = {1} bfp_fitness = {2}".format(i,self.fit[i],self.bfp_fitness[i])
# Atualiza melhor posicao global
if self.bfp_fitness[i] < self.bfg_fitness:
self.bfg_fitness = self.bfp_fitness[i].copy()
self.bfg = self.bfp[i].copy()
#######################
## Some Benchmark functions #
#######################
def f1(x):
aux = 0
for i in scipy.arange(x.shape[0]-1):
aux = aux + 100*(x[i+1] - x[i]**2)**2+(x[i]-1)**2
return aux
def f2(x):
xx = x[0]**2+x[1]**2
aux = 1 + scipy.cos(12*scipy.sqrt(xx))
aux = aux / (.5*xx+2)
return -aux
def f3(x):
k1 = 0
k2 = 0
k3 = 1/float(x.shape[0])
for i in range(x.shape[0]):
k1 = k1 + x[i]**2
k2 = k2 + scipy.cos(2*scipy.pi*x[i])
return -20*scipy.exp(-0.2*scipy.sqrt(k3*k1))-scipy.exp(k3*k2)+20+scipy.exp(1)