forked from atulsinha007/Domain-adaptation
-
Notifications
You must be signed in to change notification settings - Fork 0
/
chromosome.py
742 lines (631 loc) · 27.8 KB
/
chromosome.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
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
# chromosome.py
import math
import random
import gene
from gene import *
import matenc
import numpy as np
import tensorflow as tf
import deep_net
import time
inputnumber = 32
outputnumber = 5 # here could be an error, after all that's why I don't use global variables
innov_ctr = inputnumber * outputnumber + 1
# import network
class Chromosome:
"""
def __init__(self,dob,node_arr=[],conn_arr=[],bias_arr=[]):
self.node_arr = node_arr #list of node objects
self.conn_arr = conn_arr #list of conn objects
self.bias_conn_arr = bias_arr #list of BiasNode objects
self.dob = dob #the generation in which it was created.
self.node_ctr=len(node_arr)+1
"""
# here initialization is always with simplest chromosome (AND mainly for innov ctr) , here could be an error
def __init__(self, inputdim, outputdim, old_chromosome = None):
if old_chromosome == None:
global innov_ctr
self.node_ctr = inputdim + outputdim + 1
# NO MORE # Warning!! these two lines change(reset) global variables, here might be some error
lisI = [gene.Node(num_setter, 'I') for num_setter in range(1, self.node_ctr - outputdim)]
lisO = [gene.Node(num_setter, 'O') for num_setter in range(inputdim + 1, self.node_ctr)]
self.node_arr = lisI + lisO
self.conn_arr = []
p = 1
for inputt in lisI:
for outputt in lisO:
self.conn_arr.append(gene.Conn(p, (inputt, outputt), random.random(), status=True))
p += 1
# print(p)
# assert (p == innov_ctr)
self.bias_conn_arr = []
self.bias_conn_arr = [gene.BiasConn(outputt, random.random() / 1000) for outputt in lisO]
self.dob = 0
else:
self.node_ctr = old_chromosome.node_ctr
self.conn_arr = old_chromosome.conn_arr
self.bias_conn_arr = old_chromosome.bias_conn_arr
self.node_arr = old_chromosome.node_arr
self.dob = old_chromosome.dob
def reset_chromo_to_zero(self):
self.node_ctr = 0
self.node_arr = []
self.conn_arr = []
self.bias_conn_arr = []
def set_node_ctr(self, ctr=None):
if not ctr:
ctr = len(self.node_arr) + 1
self.node_ctr = ctr
def pp(self):
print("\nNode List")
[item.pp() for item in self.node_arr]
print("\n\nConnection List")
[item.pp() for item in self.conn_arr]
print("\n\nBias Connection List")
[item.pp() for item in self.bias_conn_arr]
print("dob", self.dob, "node counter", self.node_ctr)
print("--------------------------------------------")
def convert_to_MatEnc(self, inputdim, outputdim):
ConnMatrix = {} # Connection Matrix
WeightMatrix = {} # Weight Matrix
NatureCtrDict = {} # Contains Counter of Nature { 'I', 'H1', 'H2', 'O' }
NatureCtrDict['I'] = 0
NatureCtrDict['H1'] = 0
NatureCtrDict['H2'] = 0
NatureCtrDict['O'] = 0
dictionary = {} # Contains node numbers mapping starting from 0, nature-wise
dictionary['I'] = {}
dictionary['H1'] = {}
dictionary['H2'] = {}
dictionary['O'] = {}
couple_to_conn_map = {}
for i in self.node_arr:
dictionary[i.nature][i] = NatureCtrDict[i.nature]
NatureCtrDict[i.nature] += 1
"""
ConnMatrix['IO'] = np.zeros((inputdim, outputdim))
ConnMatrix['IH1'] = np.zeros((inputdim, NatureCtrDict['H1']))
ConnMatrix['IH2'] = np.zeros((inputdim, NatureCtrDict['H2']))
ConnMatrix['H1H2'] = np.zeros((NatureCtrDict['H1'], NatureCtrDict['H2']))
ConnMatrix['H1O'] = np.zeros((NatureCtrDict['H1'], outputdim))
ConnMatrix['H2O'] = np.zeros((NatureCtrDict['H2'], outputdim))
WeightMatrix['IO'] = np.zeros((inputdim, outputdim))
WeightMatrix['IH1'] = np.zeros((inputdim, NatureCtrDict['H1']))
WeightMatrix['IH2'] = np.zeros((inputdim, NatureCtrDict['H2']))
WeightMatrix['H1H2'] = np.zeros((NatureCtrDict['H1'], NatureCtrDict['H2']))
WeightMatrix['H1O'] = np.zeros((NatureCtrDict['H1'], outputdim))
WeightMatrix['H2O'] = np.zeros((NatureCtrDict['H2'], outputdim))
"""
for con in self.conn_arr:
if con.source.nature + con.destination.nature not in ConnMatrix.keys():
ConnMatrix[con.source.nature + con.destination.nature] = np.zeros(
(NatureCtrDict[con.source.nature], NatureCtrDict[con.destination.nature]))
WeightMatrix[con.source.nature + con.destination.nature] = np.zeros(
(NatureCtrDict[con.source.nature], NatureCtrDict[con.destination.nature]))
if con.status == 1:
ConnMatrix[con.source.nature + con.destination.nature][
dictionary[con.source.nature][con.source]][
dictionary[con.destination.nature][con.destination]] = 1
couple_to_conn_map[con.get_couple()] = con
# print(con.source.nature + con.destination.nature)
WeightMatrix[con.source.nature + con.destination.nature][dictionary[con.source.nature][con.source]][
dictionary[con.destination.nature][con.destination]] = con.weight
inv_dic = {key: {v: k for k, v in dictionary[key].items()} for key in dictionary.keys()}
new_encoding = matenc.MatEnc(WeightMatrix, ConnMatrix, self.bias_conn_arr, inv_dic, couple_to_conn_map,
self.node_arr, self.conn_arr)
return new_encoding
def modify_thru_backprop(self, inputdim, outputdim, trainx, trainy, epochs=10, learning_rate=0.1, n_par=10):
x = tf.placeholder(shape=[None, inputdim], dtype=tf.float32)
y = tf.placeholder(shape=[None, ], dtype=tf.int32)
n_par = n_par
par_size = tf.shape(trainx)[0] // n_par
prmsdind = tf.placeholder(name='prmsdind', dtype=tf.int32)
valid_x_to_be = trainx[prmsdind * par_size:(prmsdind + 1) * par_size, :]
valid_y_to_be = trainy[prmsdind * par_size:(prmsdind + 1) * par_size]
train_x_to_be = tf.concat(
(trainx[:(prmsdind) * par_size, :], trainx[(prmsdind + 1) * par_size:, :]),
axis=0)
train_y_to_be = tf.concat(
(trainy[:(prmsdind) * par_size], trainy[(prmsdind + 1) * par_size:]), axis=0)
mat_enc = self.convert_to_MatEnc(inputdim, outputdim)
newneu_net = deep_net.DeepNet(x, inputdim, outputdim, mat_enc)
cost = newneu_net.negative_log_likelihood(y)
# print(newneu_net.mat_enc.CMatrix['IO'])
optmzr = tf.train.GradientDescentOptimizer(learning_rate).minimize(cost, var_list=newneu_net.params)
# savo1 = tf.train.Saver(var_list=[self.srest_setx, self.srest_sety, self.stest_setx, self.stest_sety])
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
# err = sess.run(newneu_net.errors(y), feed_dict={x: trainx, y: trainy})
# print("train error ", err)
# just any no. which does not satisfy below condition
prev = 7
current = 5
start1 = time.time()
for epoch in range(epochs):
listisi = []
for ind in range(n_par):
_, bost = sess.run([optmzr, cost],
feed_dict={x: train_x_to_be.eval(feed_dict={prmsdind: ind}),
y: train_y_to_be.eval(feed_dict={prmsdind: ind})})
if epoch % (epochs // 4) == 0:
q = newneu_net.errors(y).eval(
feed_dict={x: valid_x_to_be.eval(feed_dict={prmsdind: ind}),
y: valid_y_to_be.eval(feed_dict={prmsdind: ind})})
listisi.append(q)
if epoch % (epochs // 4) == 0:
prev = current
current = np.mean(listisi)
print('validation', current)
print(tf.reduce_sum(newneu_net.wei_mat_var_map['IO']).eval())
if current - prev > 0.0002:
break;
end1 = time.time()
print("time ", end1 - start1)
print("now ending")
for key in newneu_net.wei_mat_var_map.keys():
newneu_net.mat_enc.WMatrix[key] = newneu_net.wei_mat_var_map[key].eval()
for i in range(len(newneu_net.bias_wei_arr)):
ar = newneu_net.bias_var.eval()
newneu_net.mat_enc.Bias_conn_arr[i].set_weight(ar[i])
#print(newneu_net.mat_enc.CMatrix['IO'], 'final')
newchromo = newneu_net.mat_enc.convert_to_chromosome(inputdim, outputdim, self.dob)
self.conn_arr = newchromo.conn_arr
self.node_arr = newchromo.node_arr
self.bias_conn_arr = newchromo.bias_conn_arr # list of BiasNode objects
self.dob = newchromo.dob # the generation in which it was created.
self.node_ctr = len(self.node_arr) + 1
return newchromo
def weight_mutation(self, rng, factor=0.01, individual_change_probablity=0.1):
import copy
lis = self.conn_arr + self.bias_conn_arr
# chosen_ind = rng.choice(range(len(lis)))
for item in lis:
if rng.random() < individual_change_probablity:
item.weight += (rng.random() - 0.5) * 2 * factor
# return chosen_ind
def edge_mutation(self, inputdim, outputdim, rng): # not tested, might just have some error!
newmatenc = self.convert_to_MatEnc(inputdim, outputdim)
key_list = list(newmatenc.WMatrix.keys())
# key_list.remove('IO')
# print(key_list)
chosen_key = rng.choice(key_list)
while chosen_key not in newmatenc.CMatrix.keys():
chosen_key = rng.choice(key_list)
mat = newmatenc.CMatrix[chosen_key]
# print(chosen_key, mat.shape, list(newmatenc.node_map.items()))
i = rng.randint(0, mat.shape[0]-1)
if mat.shape[1] > 1:
j = rng.randint(0, mat.shape[1]-1)
else:
j = 0
split_key1, split_key2 = matenc.split_key(chosen_key)
# print(split_key1, split_key2)
# couple = (newmatenc.node_map[split_key1][i], newmatenc.node_map[split_key2][j])
ctr = 0
# print(mat[i][j])
while mat[i][j] != 0:
i = rng.randint(0, mat.shape[0]-1)
if mat.shape[1] > 1:
j = rng.randint(0, mat.shape[1]-1)
else:
j = 0
if ctr > 10:
return
ctr += 1
couple = (newmatenc.node_map[split_key1][i], newmatenc.node_map[split_key2][j])
mat[i][j] = 1
if not newmatenc.WMatrix[chosen_key][i][j]:
innov_num = normalize_conn_arr_for_this_gen(self, couple)
con_obj = gene.Conn(innov_num, couple, (rng.random() - 0.5) * 2, True)
self.conn_arr.append(con_obj)
# con_obj.pp()
else:
con_obj = newmatenc.couple_to_conn_map[couple]
con_obj.status = True
def node_mutation(self, inputdim, outputdim, rng):
# global innov_ctr
type = 0
newmatenc = self.convert_to_MatEnc(inputdim, outputdim)
key_list = ['IH2', 'H1O', 'IO']
stlis = ['H1', 'H2', 'H2']
# prob_list = [0.1, 0.3, 0.6]
prndm = rng.random()
if prndm > 0.4:
ind = 2
elif prndm > 0.1:
ind = 1
elif prndm > 0:
ind = 0
chosen_key = key_list[ind]
while chosen_key not in newmatenc.CMatrix.keys():
prndm = rng.random()
if prndm > 0.4:
ind = 2
elif prndm > 0.1:
ind = 1
elif prndm > 0:
ind = 0
chosen_key = key_list[ind]
# key_list.remove('IO')
# print(key_list)
# chosen_key = (key_list)
mat = newmatenc.CMatrix[chosen_key]
# print(chosen_key, mat.shape, list(newmatenc.node_map.items()))
i = rng.randint(0, mat.shape[0]-1)
if mat.shape[1] > 1:
j = rng.randint(0, mat.shape[1]-1)
else:
j = 0
split_key1, split_key2 = matenc.split_key(chosen_key)
# (split_key1, split_key2)
ctr = 0
if not newmatenc.WMatrix[chosen_key][i][j] and not type:
while mat[i][j] == 0:
i = rng.randint(0, mat.shape[0]-1)
if mat.shape[1] > 1:
j = rng.randint(0, mat.shape[1]-1)
else:
j = 0
if ctr > 10:
return
ctr += 1
couple = (newmatenc.node_map[split_key1][i], newmatenc.node_map[split_key2][j])
con_obj = newmatenc.couple_to_conn_map[couple]
con_obj.status = False
newnode = gene.Node(self.node_ctr, stlis[ind])
self.node_ctr += 1
innov_num = normalize_conn_arr_for_this_gen(self, (con_obj.source, newnode))
new_conn1 = gene.Conn(innov_num, (con_obj.source, newnode), 1.0, True)
innov_num = normalize_conn_arr_for_this_gen(self, (newnode, con_obj.destination))
new_conn2 = gene.Conn(innov_num, (newnode, con_obj.destination), con_obj.weight, True)
self.node_arr.append(newnode)
self.conn_arr.append(new_conn1)
self.conn_arr.append(new_conn2)
def do_mutation(self, rate_conn_weight, rate_conn_itself, rate_node, weight_factor , inputdim, outputdim, max_hidden_unit, rng):
# rate_conn_weight > rate_conn_itself >> rate_node
# 0.2, 0.1, 0.05
p = len(self.conn_arr)
flag = 0
rate_conn_itself += rate_node
rate_conn_weight += rate_conn_itself
prnd = rng.random()
if prnd < rate_node:
if self.node_ctr <= (max_hidden_unit + inputdim + outputdim):
self.node_mutation(inputdim, outputdim, rng)
flag = 1
elif prnd < rate_conn_itself:
self.edge_mutation(inputdim, outputdim, rng)
flag = 1
elif prnd < rate_conn_weight:
self.weight_mutation(rng, weight_factor)
flag = 1
"""if flag:
print("before mutation length", p)
print("after mutation length", len(self.conn_arr))
"""
def convert_to_empirical_string(self):
st = ''
for con in self.conn_arr:
tup = con.get_couple()
st += tup[0].nature + tup[1].nature + str(con.innov_num)
return st
def normalize_conn_arr_for_this_gen(chromo, tup):
st = chromo.convert_to_empirical_string()
global innov_ctr
if (st, (tup[0].node_num, tup[1].node_num)) in gene.dict_of_sm_so_far.keys():
innov_num = gene.dict_of_sm_so_far[(st, (tup[0].node_num, tup[1].node_num))]
# print("matches")
else:
innov_num = innov_ctr
gene.dict_of_sm_so_far[(st, (tup[0].node_num, tup[1].node_num))] = innov_ctr
innov_ctr += 1
# print([item for item in gene.dict_of_sm_so_far.items()])
return innov_num
# 0.5, 0.1, 0.2, 0.2
def aux_weighted(parent1, parent2):
fitness_tup1 = parent1.fitness
fitness_tup2 = parent2.fitness
if fitness_tup1 <= fitness_tup2:
return parent1
elif fitness_tup1 > fitness_tup2:
return parent2
else:
theta = 0.5 * fitness_tup1[0] + 0.0001 * fitness_tup1[1] + 0.2 * fitness_tup1[2] + 0.2 * fitness_tup1[3]
fi = 0.5 * fitness_tup2[0] + 0.0001 * fitness_tup2[1] + 0.2 * fitness_tup2[2] + 0.2 * fitness_tup2[3]
if theta < fi:
return parent1
else:
return parent2
'''
def aux_weightedTest(parentx, parenty): # parentx, parenty represents a tuple (parent, fitness_arr)
parent1 = parentx[0]
parent2 = parenty[0]
fitness_tup1 = parentx[1]
fitness_tup2 = parenty[1]
if fitness_tup1 <= fitness_tup2:
return parent1
elif fitness_tup1 > fitness_tup2:
return parent2
else:
theta = 0.5 * fitness_tup1[0] + 0.0001 * fitness_tup1[1] + 0.2 * fitness_tup1[2] + 0.2 * fitness_tup1[3]
fi = 0.5 * fitness_tup2[0] + 0.0001 * fitness_tup2[1] + 0.2 * fitness_tup2[2] + 0.2 * fitness_tup2[3]
if theta < fi:
return parent1
else:
return parent2
'''
def aux_non_weighted(parent1, parent2):
fitness_tup1 = parent1.fitness
fitness_tup2 = parent2.fitness
arr1 = np.array(fitness_tup1)
arr2 = np.array(fitness_tup2)
if np.all(arr1 <= arr2):
return parent1
elif np.all(arr1 > arr2):
return parent2
else:
if fitness_tup1[0] <= fitness_tup2[0] and np.all(arr1[2:4] <= arr2[2:4]):
return parent1
elif fitness_tup1[0] > fitness_tup2[0] and np.all(arr1[2:4] > arr2[2:4]):
return parent2
elif ((fitness_tup1[0] <= fitness_tup2[0]) and (
(fitness_tup1[2] <= fitness_tup2[2]) or (fitness_tup1[3] <= fitness_tup2[3]))):
return parent1
elif ((fitness_tup1[0] > fitness_tup2[0]) and (
(fitness_tup1[2] > fitness_tup2[2]) or (fitness_tup1[3] > fitness_tup2[3]))):
return parent2
else:
return random.choice((parent1, parent2))
def aux_non_weighted_1(parent1, parent2):
fitness_tup1 = parent1.fitness
fitness_tup2 = parent2.fitness
arr1 = np.array(fitness_tup1)
arr2 = np.array(fitness_tup2)
if np.all(arr1 <= arr2):
return parent1
elif np.all(arr1 > arr2):
return parent2
else:
return random.choice((parent1, parent2))
def aux_non_weightedTest(parentx, parenty):
fitness_tup1 = parentx[1]
fitness_tup2 = parenty[1]
parent1 = parentx[0]
parent2 = parenty[0]
arr1 = np.array(fitness_tup1)
arr2 = np.array(fitness_tup2)
if np.all(arr1 <= arr2):
return parent1
elif np.all(arr1 > arr2):
return parent2
else:
if fitness_tup1[0] <= fitness_tup2[0] and np.all(arr1[2:] <= arr2[2:]):
return parent1
elif fitness_tup1[0] > fitness_tup2[0] and np.all(arr1[2:] > arr2[2:]):
return parent2
elif ((fitness_tup1[0] <= fitness_tup2[0]) and (
(fitness_tup1[2] <= fitness_tup2[2]) or (fitness_tup1[3] <= fitness_tup2[3]))):
return parent1
elif ((fitness_tup1[0] > fitness_tup2[0]) and (
(fitness_tup1[2] > fitness_tup2[2]) or (fitness_tup1[3] > fitness_tup2[3]))):
return parent2
else:
return random.choice((parent1, parent2))
def crossover(parent1, parent2, gen_no, inputdim, outputdim):
# print("cross between lengths", len(parent1.conn_arr), len(parent2.conn_arr))
if gen_no > gene.curr_gen_no:
gene.dict_of_sm_so_far = {}
gene.curr_gen_no = gen_no
# print("yes changed",gene.curr_gen_no)
child = Chromosome(inputdim, outputdim)
child.reset_chromo_to_zero()
child.dob = gen_no + 1
len1 = len(parent1.conn_arr)
len2 = len(parent2.conn_arr)
nodeDict = {}
c1 = 0
c2 = 0
dominating_parent = None
input_nodes = []
output_nodes = []
hidden_nodes = []
while c1 < len1 or c2 < len2:
f1 = f2 = 0
if c1 < len1:
i = parent1.conn_arr[c1]
f1 = 1
if c2 < len2:
j = parent2.conn_arr[c2]
f2 = 1
if (f1 == 1 and f1 == f2 and i.innov_num == j.innov_num):
alpha = random.uniform(0, 1)
wt = alpha * i.weight + (1 - alpha) * j.weight
stat = False
if i.status == j.status:
stat = i.status
else:
stat = random.choice((True, False))
nodeObj1 = nodeObj2 = None
if i.source.node_num not in nodeDict.keys():
nodeObj1 = Node(i.source.node_num, i.source.nature)
nodeDict[i.source.node_num] = nodeObj1
if i.source.nature == 'I':
input_nodes.append(nodeObj1)
else:
hidden_nodes.append(nodeObj1)
else:
nodeObj1 = nodeDict[i.source.node_num]
if i.destination.node_num not in nodeDict.keys():
nodeObj2 = Node(i.destination.node_num, i.destination.nature)
nodeDict[i.destination.node_num] = nodeObj2
if i.destination.nature == 'H1' or i.destination.nature == 'H2':
hidden_nodes.append(nodeObj2)
else:
output_nodes.append(nodeObj2)
else:
nodeObj2 = nodeDict[i.destination.node_num]
conObj = Conn(i.innov_num, (nodeObj1, nodeObj2), wt, stat) # conn object
child.conn_arr.append(conObj)
c1 += 1
c2 += 1
else:
dominating_parent = aux_non_weighted_1(parent1, parent2)
length = len(dominating_parent.conn_arr)
while c1 < length:
i = dominating_parent.conn_arr[c1]
nodeObj1 = nodeObj2 = None
if i.source.node_num not in nodeDict.keys():
nodeObj1 = Node(i.source.node_num, i.source.nature)
nodeDict[i.source.node_num] = nodeObj1
if i.source.nature == 'I':
input_nodes.append(nodeObj1)
else:
hidden_nodes.append(nodeObj1)
else:
nodeObj1 = nodeDict[i.source.node_num]
if i.destination.node_num not in nodeDict.keys():
nodeObj2 = Node(i.destination.node_num, i.destination.nature)
nodeDict[i.destination.node_num] = nodeObj2
if i.destination.nature == 'H1' or i.destination.nature == 'H2':
hidden_nodes.append(nodeObj2)
else:
output_nodes.append(nodeObj2)
else:
nodeObj2 = nodeDict[i.destination.node_num]
c1 += 1
connObj = Conn(i.innov_num, (nodeObj1, nodeObj2), i.weight, i.status)
child.conn_arr.append(connObj)
break
input_nodes.sort(key=lambda x: x.node_num)
output_nodes.sort(key=lambda x: x.node_num)
child.node_arr = input_nodes + output_nodes + hidden_nodes
if (outputdim != 1):
point_of_crossover = random.randint(0, outputdim)
for i in range(len(output_nodes)):
if i < point_of_crossover:
wt = parent1.bias_conn_arr[i].weight
else:
wt = parent2.bias_conn_arr[i].weight
new_bias_conn = gene.BiasConn(output_nodes[i], wt)
child.bias_conn_arr.append(new_bias_conn)
elif outputdim == 1:
p = random.random()
if p > 0.5:
wt = parent1.bias_conn_arr[0].weight
else:
wt = parent2.bias_conn_arr[0].weight
new_bias_conn = gene.BiasConn(output_nodes[0], wt)
child.bias_conn_arr.append(new_bias_conn)
child.set_node_ctr()
return child
"""
assert ( parent1.node_ctr == child.node_ctr or parent2.node_ctr == child.node_ctr)
assert ( len(parent1.conn_arr) == len(child.conn_arr) or len(parent2.conn_arr) == len(child.conn_arr))
if dominating_parent:
#print("FOUND ONE")
for i in range(len(dominating_parent.conn_arr)):
assert( dominating_parent.conn_arr[i].innov_num == child.conn_arr[i].innov_num)
assert( dominating_parent.conn_arr[i].source.nature + dominating_parent.conn_arr[i].destination.nature == child.conn_arr[i].source.nature + child.conn_arr[i].destination.nature)
assert ( set([item.node_num for item in dominating_parent.node_arr]) == set([item.node_num for item in child.node_arr]) )
return child
"""
def crossoverTest(parentx, parenty, gen_no, inputdim, outputdim):
parent1 = parentx[0]
parent2 = parenty[0]
# if gen_no > gene.curr_gen_num:
# gene.dict_of_sm_so_far = {}
# gene.curr_gen_num = gen_no
child = Chromosome(inputdim, outputdim)
child.reset_chromo_to_zero()
child.dob = gen_no
len1 = len(parent1.conn_arr)
len2 = len(parent2.conn_arr)
nodeDict = {}
c1 = 0
c2 = 0
input_nodes = []
output_nodes = []
hidden_nodes = []
while c1 < len1 or c2 < len2:
f1 = f2 = 0
if c1 < len1:
i = parent1.conn_arr[c1]
f1 = 1
if c2 < len2:
j = parent2.conn_arr[c2]
f2 = 1
if (f1 == 1 and f1 == f2 and i.innov_num == j.innov_num):
alpha = random.uniform(0, 1)
print("alpha", alpha)
wt = alpha * i.weight + (1 - alpha) * j.weight
stat = False
if i.status == j.status:
stat = i.status
else:
stat = random.choice((True, False))
print(stat)
nodeObj1 = nodeObj2 = None
if i.source.node_num not in nodeDict.keys():
nodeObj1 = Node(i.source.node_num, i.source.nature)
nodeDict[i.source.node_num] = nodeObj1
if i.source.nature == 'I':
input_nodes.append(nodeObj1)
else:
hidden_nodes.append(nodeObj1)
else:
nodeObj1 = nodeDict[i.source.node_num]
if i.destination.node_num not in nodeDict.keys():
nodeObj2 = Node(i.destination.node_num, i.destination.nature)
nodeDict[i.destination.node_num] = nodeObj2
if i.destination.nature == 'H1' or i.destination.nature == 'H2':
hidden_nodes.append(nodeObj2)
else:
output_nodes.append(nodeObj2)
else:
nodeObj2 = nodeDict[i.destination.node_num]
conObj = Conn(i.innov_num, (nodeObj1, nodeObj2), wt, stat) # conn object
child.conn_arr.append(conObj)
c1 += 1
c2 += 1
else:
dominating_parent = aux_non_weightedTest(parentx, parenty)
length = len(dominating_parent.conn_arr)
while c1 < length:
i = dominating_parent.conn_arr[c1]
nodeObj1 = nodeObj2 = None
if i.source.node_num not in nodeDict.keys():
nodeObj1 = Node(i.source.node_num, i.source.nature)
nodeDict[i.source.node_num] = nodeObj1
if i.source.nature == 'I':
input_nodes.append(nodeObj1)
else:
hidden_nodes.append(nodeObj1)
else:
nodeObj1 = nodeDict[i.source.node_num]
if i.destination.node_num not in nodeDict.keys():
nodeObj2 = Node(i.destination.node_num, i.destination.nature)
nodeDict[i.destination.node_num] = nodeObj2
if i.destination.nature == 'H1' or i.destination.nature == 'H2':
hidden_nodes.append(nodeObj2)
else:
output_nodes.append(nodeObj2)
else:
nodeObj2 = nodeDict[i.destination.node_num]
c1 += 1
connObj = Conn(i.innov_num, (nodeObj1, nodeObj2), i.weight, i.status)
child.conn_arr.append(connObj)
break
input_nodes.sort(key=lambda x: x.node_num)
output_nodes.sort(key=lambda x: x.node_num)
child.node_arr = input_nodes + output_nodes + hidden_nodes
point_of_crossover = random.randint(1, outputdim - 1)
for i in range(len(output_nodes)):
if i < point_of_crossover:
wt = parent1.bias_conn_arr[i].weight
else:
wt = parent2.bias_conn_arr[i].weight
new_bias_conn = gene.BiasConn(output_nodes[i], wt)
# Conn(-1, (Node(-1, -1), output_nodes[i]), wt, True)
child.bias_conn_arr.append(new_bias_conn)
child.set_node_ctr()
return child