-
Notifications
You must be signed in to change notification settings - Fork 1
/
myMain.py
665 lines (577 loc) · 43.6 KB
/
myMain.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
####################################################################################
####################################################################################
####################################################################################
def main():
import numpy as np
import scipy.io as sio
import os
import sys
sys.path.insert(0,'/content/drive/MyDrive/GoogleColab')
from sklearn.metrics import confusion_matrix, accuracy_score
from tensorflow.keras import utils as np_utils
from tensorflow.keras.callbacks import EarlyStopping
from tensorflow.keras.optimizers import Adam, RMSprop, SGD
from tensorflow.keras import backend as K
K.set_image_data_format('channels_last')
################################################################################
path0 = '/content/drive/MyDrive/GoogleColab/'
import myCNN
import myPlots
import myFunc
################################################################
ppp = '++++++++++++++++++++++++++++++++++++++++++++++++++++++++' #@param {type:"raw"}
b_optimizer_selected = 'AMSgrad' #@param ["Adam", "AMSgrad", "RMSprop", "SGD"]
b_learn_rate = 0.0005 #@param {type:"number"}
b_param1 = 0.9 #@param {type:"number"}
b_param2 = 0.99 #@param {type:"number"}
b_n_epochs = 2000 #@param {type:"slider", min:10, max:3000, step:10}
b_batch_size = 64 #@param {type:"slider", min:15, max:512, step:1}
b_patience = 200 #@param {type:"slider", min:5, max:1000, step:5}
b_class_weights = {0:1, 1:1} #param {type:"raw"}
b_dropoutRate = 0.7 #@param {type:"number"}
b_kernLength = 32 #@param {type:"integer"}
b_F1 = 16 #@param {type:"integer"}
b_D = 2 #@param {type:"integer"}
b_F2 = 32
b_kernLength2 = 16 #@param {type:"integer"}
b_norm_rate = 0.25 #param {type:"number"}
b_dropoutType = 'SpatialDropout2D' #param ["Dropout", "SpatialDropout2D"]
################################################################
qqq = '++++++++++++++++++++++++++++++++++++++++++++++++++++++++' #@param {type:"raw"}
m_optimizer_selected = 'AMSgrad' #@param ["Adam", "AMSgrad", "RMSprop", "SGD"]
m_learn_rate = 0.0005 #@param {type:"number"}
m_param1 = 0.9 #@param {type:"number"}
m_param2 = 0.99 #@param {type:"number"}
m_n_epochs = 10 # 4000 #@param {type:"slider", min:10, max:4000, step:10}
m_batch_size = 64 #@param {type:"slider", min:15, max:512, step:1}
m_patience = 10 # 300 #@param {type:"slider", min:5, max:1000, step:5}
m_class_weights = {0:1, 1:1, 2:1, 3:1, 4:1, 5:1} #param {type:"raw"}
m_dropoutRate = 0.75 #@param {type:"number"}
m_kernLength = 32 #@param {type:"integer"}
m_F1 = 16 #@param {type:"integer"}
m_D = 2 #@param {type:"integer"}
m_F2 = 32
m_kernLength2 = 16 #@param {type:"integer"}
m_norm_rate = 0.25 #param {type:"number"}
m_dropoutType = 'SpatialDropout2D' #param ["Dropout", "SpatialDropout2D"]
################################################################ CHANGE CHANGE
# selects = ["a0", "n0", "f0", "f1", "f2", "f3", "f4"] # with dBdB = "8080"
# selects = ["f03", "f04", "f05", "f07", "f10", "f15", "f20"] # with dBdB = "3457"
# selects = ["f2"]
selects = ["f03", "f04", "f05", "f07", "f10", "f15"]
# subjects = ["AMUSE_faz", "AMUSE_fce", "AMUSE_fcg", "AMUSE_fcj", "AMUSE_kw"]
subjects = ["AMUSE_faz", "AMUSE_kw", "AMUSE_fcg"]
dBdB = "3457" # ["8080", "3457"] filter's bandstop attenuation and ripple attenuation.
bm_C = 45 # [45, 55] select how many channels you want to select.
b_T = 256 # binary dataset's time samples of each sample and each channel.
m_T = 128*3 # multiclass dataset's time samples of each sample and each channel.
T_base = 44 # how many time samples taken from before the stimulus onset you want to average?
T_shift = 0 # how many time samples you want to shift to better capture the ERP?
################################################################ CHANGE CHANGE CHANGE
param_k = 0
# quant1_label = [8, 16, 32]
# for quant1 in quant1_label:
# b_F1 = quant1
# m_F1 = quant1
# quant2_label = range(3)
# for quant2 in quant2_label:
# b_F2 = quant2
# m_F2 = quant2
# for quant3 in quant3_label:
# b_kernLength2 = quant3
# m_kernLength2 = quant3
quant4_label = range(2) # useful for pseudo-crossvalidation
for quant4 in quant4_label:
param_k += 1
sel_k = 0
for select in selects:
sel_k += 1
################################################################
# just select the optimizer.
b_opt = myFunc.SelectOptimizer(b_optimizer_selected, b_learn_rate, b_param1, b_param2)
m_opt = myFunc.SelectOptimizer(m_optimizer_selected, m_learn_rate, m_param1, m_param2)
################################################################################
################################################################################
################################################################################
# update the run's number at each execution of the program.
from datetime import date
date = date.today()
with open(path0 + 'number_of_execution.txt', 'r') as f:
script = f.readlines()
if str(date) == script[0].strip():
number_run = int(script[1]) + 1
else:
number_run = 1
# start the session and write in a file the hyperparameters.
with open(path0 + 'saveAccuracy.txt', 'a') as f:
f.write("------------------ S T A R T ------------------")
f.write("\nS E S S I O N - " + str(date) + ' | run = ' + str(number_run))
f.write('\nBINARY:')
if b_optimizer_selected == 'Adam':
f.write('\t\tAdam(learning_rate = ' + str(b_learn_rate) + ', beta_1 = '
+ str(b_param1) + ', beta_2 = ' + str(b_param2) + ')')
elif m_optimizer_selected == 'AMSgrad':
f.write('\t\tAMSgrad(learning_rate = ' + str(b_learn_rate) + ', beta_1 = '
+ str(b_param1) + ', beta_2 = ' + str(b_param2) + ')')
elif b_optimizer_selected == 'RMSprop':
f.write('\t\tRMSprop(learning_rate = ' + str(b_learn_rate) + ', rho = '
+ str(b_param1) + ', momentum = ' + str(b_param2) + ')')
elif b_optimizer_selected == 'SGD':
f.write('\t\tSGD(learning_rate = ' + str(b_learn_rate) + ', momentum = '
+ str(b_param2) + ')')
f.write(' | n_epochs: ' + str(b_n_epochs) + ' | patience: ' + str(b_patience) +
' | batch_size: ' + str(b_batch_size) + ' | dropoutRate: ' + str(b_dropoutRate) +
'\n\t\tkernLength: ' + str(b_kernLength) + ' | F1: ' + str(b_F1) + ' | D: ' +
str(b_D) + ' | F2: ' + str(b_F2) + ' | kernLength2: ' + str(b_kernLength2))
f.write('\nMULTICLASS:')
if m_optimizer_selected == 'Adam':
f.write('\tAdam(learning_rate = ' + str(m_learn_rate) + ', beta_1 = '
+ str(m_param1) + ', beta_2 = ' + str(m_param2) + ')')
elif m_optimizer_selected == 'AMSgrad':
f.write('\tAMSgrad(learning_rate = ' + str(m_learn_rate) + ', beta_1 = '
+ str(m_param1) + ', beta_2 = ' + str(m_param2) + ')')
elif m_optimizer_selected == 'RMSprop':
f.write('\tRMSprop(learning_rate = ' + str(m_learn_rate) + ', rho = '
+ str(m_param1) + ', momentum = ' + str(m_param2) + ')')
elif m_optimizer_selected == 'SGD':
f.write('\tSGD(learning_rate = ' + str(m_learn_rate) + ', momentum = '
+ str(m_param2) + ')')
f.write(' | n_epochs: ' + str(m_n_epochs) + ' | patience: ' + str(m_patience) +
' | batch_size: ' + str(m_batch_size) + ' | dropoutRate: ' + str(m_dropoutRate) +
'\n\t\tkernLength: ' + str(m_kernLength) + ' | F1: ' + str(m_F1) + ' | D: ' +
str(m_D) + ' | F2: ' + str(m_F2) + ' | kernLength2: ' + str(m_kernLength2) + '\n')
# start the session individually for each subject and write in a file the hyperparameters.
for subject in subjects:
with open(path0 + subject +'/' + subject + '_saveAccuracy.txt', 'a') as f:
f.write("------------------ S T A R T ------------------")
f.write("\nS E S S I O N - " + subject + ' | ' + str(date) + ' | run = ' +
str(number_run))
f.write('\nBINARY:')
if b_optimizer_selected == 'Adam':
f.write('\t\tAdam(learning_rate = ' + str(b_learn_rate) + ', beta_1 = '
+ str(b_param1) + ', beta_2 = ' + str(b_param2) + ')')
elif m_optimizer_selected == 'AMSgrad':
f.write('\t\tAMSgrad(learning_rate = ' + str(b_learn_rate) + ', beta_1 = '
+ str(b_param1) + ', beta_2 = ' + str(b_param2) + ')')
elif b_optimizer_selected == 'RMSprop':
f.write('\t\tRMSprop(learning_rate = ' + str(b_learn_rate) + ', rho = '
+ str(b_param1) + ', momentum = ' + str(b_param2) + ')')
elif b_optimizer_selected == 'SGD':
f.write('\t\tSGD(learning_rate = ' + str(b_learn_rate) + ', momentum = '
+ str(b_param2) + ')')
f.write(' | n_epochs: ' + str(b_n_epochs) + ' | patience: ' + str(b_patience) +
' | batch_size: ' + str(b_batch_size) + ' | dropoutRate: ' + str(b_dropoutRate) +
'\n\t\tkernLength: ' + str(b_kernLength) + ' | F1: ' + str(b_F1) + ' | D: ' +
str(b_D) + ' | F2: ' + str(b_F2) + ' | kernLength2: ' + str(b_kernLength2))
f.write('\nMULTICLASS:')
if m_optimizer_selected == 'Adam':
f.write('\tAdam(learning_rate = ' + str(m_learn_rate) + ', beta_1 = '
+ str(m_param1) + ', beta_2 = ' + str(m_param2) + ')')
elif m_optimizer_selected == 'AMSgrad':
f.write('\tAMSgrad(learning_rate = ' + str(m_learn_rate) + ', beta_1 = '
+ str(m_param1) + ', beta_2 = ' + str(m_param2) + ')')
elif m_optimizer_selected == 'RMSprop':
f.write('\tRMSprop(learning_rate = ' + str(m_learn_rate) + ', rho = '
+ str(m_param1) + ', momentum = ' + str(m_param2) + ')')
elif m_optimizer_selected == 'SGD':
f.write('\tSGD(learning_rate = ' + str(m_learn_rate) + ', momentum = '
+ str(m_param2) + ')')
f.write(' | n_epochs: ' + str(m_n_epochs) + ' | patience: ' + str(m_patience) +
' | batch_size: ' + str(m_batch_size) + ' | dropoutRate: ' + str(m_dropoutRate) +
'\n\t\tkernLength: ' + str(m_kernLength) + ' | F1: ' + str(m_F1) + ' | D: ' +
str(m_D) + ' | F2: ' + str(m_F2) + ' | kernLength2: ' + str(m_kernLength2) + '\n')
# hyperParam.mat ###############################################################
print('PATH: ' + path0 + '\nSave the hyperparameters in a .mat...')
sio.savemat('/hyperParam_' + select + '_' + str(date) + '_run-' + str(number_run) + '.mat',
{"b_optimizer":b_optimizer_selected, "b_learn_rate":b_learn_rate,
"b_param1":b_param1, "b_param2":b_param2, "b_batch_size":b_batch_size,
"b_n_epochs":b_n_epochs, "b_patience":b_patience,
"b_dropoutRate":b_dropoutRate, "b_kernLength":b_kernLength,
"b_F1":b_F1, "b_D":b_D, "b_F2":b_F2, "b_kernLength2":b_kernLength2,
"m_optimizer":m_optimizer_selected, "m_learn_rate":m_learn_rate,
"m_param1":m_param1, "m_param2":m_param2, "m_batch_size":m_batch_size,
"m_n_epochs":m_n_epochs, "m_patience":m_patience,
"m_dropoutRate":m_dropoutRate, "m_kernLength":m_kernLength,
"m_F1":m_F1, "m_D":m_D, "m_F2":m_F2, "m_kernLength2":m_kernLength2})
################################################################################
################################################################################
################################################################################
################################################################################
################################################################################
################################################################################
################################################################################
################################################################################
# loop for each subject.
sub_k = 0
for subject in subjects:
sub_k += 1
Title = subject + '_' + str(date) + '_run-' + str(number_run) + '__' + select
path1 = path0 + subject + '/models_' + Title + '/'
if os.path.exists(path1):
print("\nFOLDER ALREADY EXISTING AT:\n\t" + path1)
else:
print("\nNEW FOLDER CREATED AT:\n\t" + path1)
os.makedirs(path1)
print('\nSubject ' + str(sub_k) + ' of ' + str(len(subjects)) + '\nTitle: ' + Title)
############################################################################
print("------------------ S T A R T ------------------")
print("S E S S I O N - " + subject + ' | ' + str(date) + ' | run = ' + str(number_run))
print('BINARY:')
if b_optimizer_selected == 'Adam':
print('Adam(learning_rate = ' + str(b_learn_rate) + ', beta_1 = '
+ str(b_param1) + ', beta_2 = ' + str(b_param2) + ')')
if b_optimizer_selected == 'AMSgrad':
print('AMSgrad(learning_rate = ' + str(b_learn_rate) + ', beta_1 = '
+ str(b_param1) + ', beta_2 = ' + str(b_param2) + ')')
elif b_optimizer_selected == 'RMSprop':
print('RMSprop(learning_rate = ' + str(b_learn_rate) + ', rho = '
+ str(b_param1) + ', momentum = ' + str(b_param2) + ')')
elif b_optimizer_selected == 'SGD':
print('SGD(learning_rate = ' + str(b_learn_rate) + ', momentum = '
+ str(b_param2) + ')')
print('#\tn_epochs: ' + str(b_n_epochs) + ' | patience: ' + str(b_patience) +
' | batch_size: ' + str(b_batch_size) + ' | dropoutRate: ' + str(b_dropoutRate) +
'\n#\tkernLength: ' + str(b_kernLength) + ' | F1: ' + str(b_F1) + ' | D: ' +
str(b_D) + ' | F2: ' + str(b_F2) + ' | kernLength2: ' + str(b_kernLength2))
print('MULTICLASS:')
if m_optimizer_selected == 'Adam':
print('Adam(learning_rate = ' + str(m_learn_rate) + ', beta_1 = '
+ str(m_param1) + ', beta_2 = ' + str(m_param2) + ')')
elif m_optimizer_selected == 'AMSgrad':
print('AMSgrad(learning_rate = ' + str(m_learn_rate) + ', beta_1 = '
+ str(m_param1) + ', beta_2 = ' + str(m_param2) + ')')
elif m_optimizer_selected == 'RMSprop':
print('RMSprop(learning_rate = ' + str(m_learn_rate) + ', rho = '
+ str(m_param1) + ', momentum = ' + str(m_param2) + ')')
elif m_optimizer_selected == 'SGD':
print('SGD(learning_rate = ' + str(m_learn_rate) + ', momentum = '
+ str(m_param2) + ')')
print('#\tn_epochs: ' + str(m_n_epochs) + ' | patience: ' + str(m_patience) +
' | batch_size: ' + str(m_batch_size) + ' | dropoutRate: ' + str(m_dropoutRate) +
'\n#\tkernLength: ' + str(m_kernLength) + ' | F1: ' + str(m_F1) + ' | D: ' +
str(m_D) + ' | F2: ' + str(m_F2) + ' | kernLength2: ' + str(m_kernLength2) + '\n')
############################################################################
print('\nLoading datasets...')
b_X_train, b_X_test, m_X_train, m_X_test, b_Y_train, b_Y_test, m_Y_train, m_Y_test = myFunc.LoadDataset(subject,
select, dBdB, bm_C, b_T, m_T, T_base, T_shift, path0)
############################################################################
# print('\nLoading dataset b...') ############################################
# b_path = path0 + subject +'/b'
# mat_contents = sio.loadmat(b_path + select + '_X3D.mat') # ba0_X3D.mat || 4320 samples
# b_X_train = mat_contents['b_X3D_train'] #################################
# mat_contents = sio.loadmat(b_path + select + '_X3D_val.mat') # ba0_X3D_val.mat || 5940 samples
# b_X_test = mat_contents['b_X3D_val'] ###################################
# mat_contents = sio.loadmat(b_path + '_Y.mat') # b_Y.mat || 4320 samples
# b_Y_train = mat_contents['b_Y_train'][:, -1]
# mat_contents = sio.loadmat(b_path + '_Y_val.mat') # b_Y_val.mat || 5940 samples
# b_Y_test = mat_contents['b_Y_val'][:, -1]
############################################################################
# print('Loading dataset m...') ##############################################
# m_path = path0 + subject +'/m'
# mat_contents = sio.loadmat(m_path + select + '_X3D.mat') # ma0_X3D.mat || 720 samples
# m_X_train = mat_contents['m_X3D_train'] #################################
# mat_contents = sio.loadmat(m_path + select + '_X3D_val.mat') # ma0_X3D_val.mat || 990 samples
# m_X_test = mat_contents['m_X3D_val'] ###################################
# mat_contents = sio.loadmat(m_path + '_Y.mat') # m_Y.mat || 720 samples
# m_Y_train = mat_contents['m_Y_train'][:, -1]
# mat_contents = sio.loadmat(m_path + '_Y_val.mat') # m_Y_val.mat || 990 samples
# m_Y_test = mat_contents['m_Y_val'][:, -1]
# mat_contents = sio.loadmat(m_path + '_Y_sound.mat') # m_Y_sound.mat || 720 samples
# m_Y_sound_train = mat_contents['m_Y_sound_train'][:, -1]
# mat_contents = sio.loadmat(m_path + '_Y_sound_val.mat') # m_Y_sound_val.mat || 990 samples
# m_Y_sound_test = mat_contents['m_Y_sound_val'][:, -1]
b_sample_train = b_X_train.shape[0]
m_sample_train = m_X_train.shape[0]
b_sample_test = b_X_test.shape[0]
m_sample_test = m_X_test.shape[0]
b_C = b_X_train.shape[1]
m_C = m_X_train.shape[1]
b_T = b_X_train.shape[2]
m_T = m_X_train.shape[2]
print('Hot encoding...')
b_Y_train_hot = np_utils.to_categorical(b_Y_train)
b_Y_test_hot = np_utils.to_categorical(b_Y_test)
m_Y_train_hot = np_utils.to_categorical(m_Y_train-1)
m_Y_test_hot = np_utils.to_categorical(m_Y_test-1)
# m_Y_sound_train_hot = np_utils.to_categorical(m_Y_sound_train-1)
# m_Y_sound_test_hot = np_utils.to_categorical(m_Y_sound_test-1)
b_X_train_hot = b_X_train.reshape(b_sample_train, b_C, b_T, 1)
b_X_test_hot = b_X_test.reshape(b_sample_test, b_C, b_T, 1)
m_X_train_hot = m_X_train.reshape(m_sample_train, m_C, m_T, 1)
m_X_test_hot = m_X_test.reshape(m_sample_test, m_C, m_T, 1)
#########################################################################################################
#########################################################################################################
#########################################################################################################
#########################################################################################################
print('\n################################################################################')
print('################################################################################')
print('BINARY MODEL:')
if quant4 == 0:
b_tra = np.array(range(int(b_sample_train*3/4)))
b_val = np.array(range(int(b_sample_train/4))) + int(b_sample_train*3/4)
else:
b_val = np.array(range(int(b_sample_train/4)))
b_tra = np.array(range(int(b_sample_train*3/4))) + int(b_sample_train/4)
K.clear_session()
path2 = '_' + Title
b_path = '_b' + path2
print('\n### PARAM = ' + str(param_k) + ', SELECT = ' + str(sel_k) +
'\n###\n### Subject = ' + str(sub_k) + ' of ' + str(len(subjects)) +
'\n###\n###\nSET: ' + path1 + 'SWAG' + b_path + '\n\tBINARY: ' + Title)
# call the binary model #############################################################################
b_model = myCNN.EEGNet(nb_classes = 2, Chans = b_C, Samples = b_T,
dropoutRate = b_dropoutRate, kernLength = b_kernLength,
F1 = b_F1, D = b_D, F2 = b_F2, kernLength2 = b_kernLength2,
norm_rate = b_norm_rate, dropoutType = b_dropoutType,
activFunct = 'sigmoid')
# compile the model.
b_model.compile(loss = 'binary_crossentropy', optimizer = b_opt, metrics = ['accuracy'])
b_numParams = b_model.count_params()
if sub_k * sel_k == 1:
# save the model's weights only at the the first execution.
b_model.save_weights(path0 + 'b_model_weights.h5')
else:
b_model.load_weights(path0 + 'b_model_weights.h5')
# set the checkpointer.
b_callback = myCNN.ReturnBestEarlyStopping(monitor = "val_loss", mode='min', patience = b_patience,
verbose = 1, restore_best_weights = True)
# fit the model.
print('\nFit the binary model:')
b_fittedModel = b_model.fit(b_X_train_hot[b_tra], b_Y_train_hot[b_tra],
batch_size = b_batch_size, epochs = b_n_epochs,
verbose = 2, validation_data = (b_X_train_hot[b_val],
b_Y_train_hot[b_val]), callbacks = [b_callback],
class_weight = b_class_weights)
b_n_epochs_actual = len(b_fittedModel.history['loss'])
print('\nn_epochs = ' + str(b_n_epochs_actual))
# evaluate the model on the validation set (1/4 of the train dataset).
b_scores = b_model.evaluate(b_X_train_hot[b_val], b_Y_train_hot[b_val], verbose=1)
print('\nModel metrics:')
print("%s: %.2f%%" % (b_model.metrics_names[1], b_scores[1]*100))
#####################################################################################################
b_probs = b_model.predict(b_X_test_hot)
# fittedModel.jpg ###################################################################################
text = path1 + 'fittedModel' + b_path + '.jpg'
myPlots.fitModel('binary', b_fittedModel.history, 0, text)
# fittedModelZoom.jpg ###############################################################################
text = path1 + 'fittedModelZoom' + b_path + '.jpg'
myPlots.fitModel('binary', b_fittedModel.history, 1, text)
# fittedModelScale.jpg ###############################################################################
text = path1 + 'fittedModelScale' + b_path + '.jpg'
myPlots.fitModel('binary', b_fittedModel.history, 2, text)
# fittedModel.mat ###################################################################################
sio.savemat(path1 + 'fittedModelMAT' + b_path + '.mat',
{"b_fittedModel_loss":b_fittedModel.history['loss'],
"b_fittedModel_loss_val":b_fittedModel.history['val_loss'],
"b_fittedModel_acc":b_fittedModel.history['accuracy'],
"b_fittedModel_acc_val":b_fittedModel.history['val_accuracy']})
#########################################################################################################
#########################################################################################################
#########################################################################################################
#########################################################################################################
print('\n################################################################################')
print('################################################################################')
print('MULTICLASS MODEL:')
if quant4 == 0:
m_tra = np.array(range(int(m_sample_train*3/4)))
m_val = np.array(range(int(m_sample_train/4))) + int(m_sample_train*3/4)
else:
m_val = np.array(range(int(m_sample_train/4)))
m_tra = np.array(range(int(m_sample_train*3/4))) + int(m_sample_train/4)
K.clear_session()
path2 = '_' + Title
m_path = '_m' + path2
print('\n### PARAM = ' + str(param_k) + ', SELECT = ' + str(sel_k) +
'\n###\n### Subject = ' + str(sub_k) + ' of ' + str(len(subjects)) +
'\n###\n###\nSET: ' + path1 + 'SWAG' + m_path + '\n\tMULTICLASS: ' + Title)
# call the multiclass model #########################################################################
m_model = myCNN.EEGNet(nb_classes = 6, Chans = m_C, Samples = m_T,
dropoutRate = m_dropoutRate, kernLength = m_kernLength,
F1 = m_F1, D = m_D, F2 = m_F2, kernLength2 = m_kernLength2,
norm_rate = m_norm_rate, dropoutType = m_dropoutType,
activFunct = 'softmax')
# compile the model.
m_model.compile(loss = 'categorical_crossentropy', optimizer = m_opt, metrics = ['accuracy'])
m_numParams = m_model.count_params()
if sub_k * sel_k == 1:
# save the model's weights only at the the first execution.
m_model.save_weights(path0 + 'm_model_weights.h5')
else:
m_model.load_weights(path0 + 'm_model_weights.h5')
# set the checkpointer.
m_callback = myCNN.ReturnBestEarlyStopping(monitor = "val_loss", mode='min', patience = m_patience,
verbose = 1, restore_best_weights = True)
# fit the model.
print('\nFit the multiclass model:')
m_fittedModel = m_model.fit(m_X_train_hot[m_tra], m_Y_train_hot[m_tra],
batch_size = m_batch_size, epochs = m_n_epochs,
verbose = 2, validation_data = (m_X_train_hot[m_val],
m_Y_train_hot[m_val]), callbacks = [m_callback],
class_weight = m_class_weights)
m_n_epochs_actual = len(m_fittedModel.history['loss'])
print('\nn_epochs = ' + str(m_n_epochs_actual))
# evaluate the model on the validation set (1/4 of the training dataset).
m_scores = m_model.evaluate(m_X_train_hot[m_val], m_Y_train_hot[m_val], verbose=1)
print('\nModel metrics:')
print("%s: %.2f%%" % (m_model.metrics_names[1], m_scores[1]*100))
#########################################################################################################
m_probs = m_model.predict(m_X_test_hot)
# fittedModel.jpg ###################################################################################
text = path1 + 'fittedModel' + m_path + '.jpg'
myPlots.fitModel('multiclass', m_fittedModel.history, 0, text)
# fittedModelZoom.jpg ###############################################################################
text = path1 + 'fittedModelZoom' + m_path + '.jpg'
myPlots.fitModel('multiclass', m_fittedModel.history, 1, text)
# fittedModelScale.jpg ###############################################################################
text = path1 + 'fittedModelScale' + m_path + '.jpg'
myPlots.fitModel('multiclass', m_fittedModel.history, 2, text)
# fittedModel.mat ###################################################################################
sio.savemat(path1 + 'fittedModelMAT' + m_path + '.mat',
{"m_fittedModel_loss":m_fittedModel.history['loss'],
"m_fittedModel_loss_val":m_fittedModel.history['val_loss'],
"m_fittedModel_acc":m_fittedModel.history['accuracy'],
"m_fittedModel_acc_val":m_fittedModel.history['val_accuracy']})
#########################################################################################################
#########################################################################################################
#########################################################################################################
#########################################################################################################
#########################################################################################################
#########################################################################################################
print('\n\n\nL O A D - M O D E L S #########################################################')
path1 = path0 + subject +'/models_' + Title + '/'
path2 = '_' + Title
b_path = '_b' + path2
m_path = '_m' + path2
print('\n###\n###\nSET: ' + path1 + 'SWAG' + path2)
#########################################################################################################
b_preds_bin_old = b_probs.argmax(axis = -1)
m_preds_multi = m_probs.argmax(axis = -1)
b_preds_multi = np.zeros(m_sample_test, dtype = int)
for k in range(m_sample_test):
b_preds_multi[k] = b_probs[np.array(range(6)) + 6*k, 1].argmax()
b_preds_bin = np.zeros(b_sample_test, dtype=int)
for k in range(m_sample_test):
b_preds_bin[b_preds_multi[k] + k*6] = 1
m_preds_bin = np.zeros(b_sample_test, dtype=int)
for k in range(m_sample_test):
m_preds_bin[m_preds_multi[k] + k*6] = 1
b_preds_multi = b_preds_multi + 1
m_preds_multi = m_preds_multi + 1
#########################################################################################################
print('\nPlot the confusion matrix...')
b_confusion_bin = confusion_matrix(b_Y_test, b_preds_bin)
b_confusion_multi = confusion_matrix(m_Y_test, b_preds_multi)
m_confusion_bin = confusion_matrix(b_Y_test, m_preds_bin)
m_confusion_multi = confusion_matrix(m_Y_test, m_preds_multi)
text = path1 + 'confusionMatrix' + path2 + '.jpg'
myPlots.confusionMat(b_confusion_bin, b_confusion_multi, m_confusion_bin, m_confusion_multi, text)
# accuracy&confusionMat.mat #########################################################################
print('\nSave the accuracy variables and the confusion matrix on test data in a .mat...')
b_acc_bin = accuracy_score(b_Y_test, b_preds_bin)
b_acc_multi = accuracy_score(m_Y_test, b_preds_multi)
m_acc_bin = accuracy_score(b_Y_test, m_preds_bin)
m_acc_multi = accuracy_score(m_Y_test, m_preds_multi)
sio.savemat(path1 + 'accuracy' + path2 + '.mat',
{"b_acc_val":b_scores[1], "m_acc_val":m_scores[1],
"b_probs":b_probs, "m_probs":m_probs,
"b_acc_bin":b_acc_bin, "m_acc_bin":m_acc_bin,
"b_acc_multi":b_acc_multi, "m_acc_multi":m_acc_multi,
"b_confusion_bin":b_confusion_bin, "m_confusion_bin":m_confusion_bin,
"b_confusion_multi":b_confusion_multi, "m_confusion_multi":m_confusion_multi,
"b_n_epochs_actual":b_n_epochs_actual, "m_n_epochs_actual":m_n_epochs_actual})
#####################################################################################################
print('\nPrint accuracy in a txt file to compare the same options accross subjects...')
with open(path0 + 'saveAccuracy.txt', 'a') as f:
f.write(subject + " | " + select)
f.write("\nBinary-corrected acc: " + str(round(b_acc_bin*100, 1)) + "%")
f.write(" | Binary-multiclass acc: " + str(round(b_acc_multi*100, 1)) + "%")
f.write(" | Multiclass-binary acc: " + str(round(m_acc_bin*100, 1)) + "%")
f.write(" | Multiclass acc: " + str(round(m_acc_multi*100, 1)) + "%")
f.write("\nBinary Validation acc: " + str(np.around(b_scores[1]*100, 1)) + "%")
f.write("\t\t\t\t\t\t| Multiclass Validation acc: " + str(np.around(m_scores[1]*100, 1)) + "%")
f.write("\nActual Epochs (bin): " + str(b_n_epochs_actual) + "/" + str(b_n_epochs))
f.write(" | Actual Epochs (multi): " + str(m_n_epochs_actual) + "/" + str(m_n_epochs))
f.write("\n\n")
#####################################################################################################
print('\nPrint accuracy in a txt file to compare different options within each subject...')
with open(path0 + subject + '/' + subject + '_saveAccuracy.txt', 'a') as f:
f.write(subject + " | " + select)
f.write("\nBinary-corrected acc: " + str(round(b_acc_bin*100, 1)) + "%")
f.write(" | Binary-multiclass acc: " + str(round(b_acc_multi*100, 1)) + "%")
f.write(" | Multiclass-binary acc: " + str(round(m_acc_bin*100, 1)) + "%")
f.write(" | Multiclass acc: " + str(round(m_acc_multi*100, 1)) + "%")
f.write("\nBinary Validation acc: " + str(np.around(b_scores[1]*100, 1)) + "%")
f.write("\t\t\t\t\t\t| Multiclass Validation acc: " + str(np.around(m_scores[1]*100, 1)) + "%")
f.write("\nActual Epochs (bin): " + str(b_n_epochs_actual) + "/" + str(b_n_epochs))
f.write(" | Actual Epochs (multi): " + str(m_n_epochs_actual) + "/" + str(m_n_epochs))
f.write("\n\n")
#####################################################################################################
print('\nSave status of parameters in a txt file...')
with open(path1 + 'info' + path2 + '.txt', 'w') as f:
f.write('####### ' + Title + ' #######')
f.write('\n\nBINARY: ')
if b_optimizer_selected == 'Adam':
f.write('\t\tAdam(learning_rate = ' + str(b_learn_rate) + ', beta_1 = '
+ str(b_param1) + ', beta_2 = ' + str(b_param2) + ')')
elif b_optimizer_selected == 'AMSgrad':
f.write('\t\tAMSgrad(learning_rate = ' + str(b_learn_rate) + ', beta_1 = '
+ str(b_param1) + ', beta_2 = ' + str(b_param2) + ')')
elif b_optimizer_selected == 'RMSprop':
f.write('\t\tRMSprop(learning_rate = ' + str(b_learn_rate) + ', rho = '
+ str(b_param1) + ', momentum = ' + str(b_param2) + ')')
elif b_optimizer_selected == 'SGD':
f.write('\t\tSGD(learning_rate = ' + str(b_learn_rate) + ', momentum = '
+ str(b_param2) + ')')
f.write(' | n_epochs: ' + str(b_n_epochs_actual) + "/" + str(b_n_epochs) +
' | patience: ' + str(b_patience) + ' | batch_size: ' + str(b_batch_size) +
' | dropoutRate: ' + str(b_dropoutRate) +
'\n\t\tkernLength: ' + str(b_kernLength) + ' | F1: ' + str(b_F1) + ' | D: ' +
str(b_D) + ' | F2: ' + str(b_F2) + ' | kernLength2: ' + str(b_kernLength2))
f.write('\nBINARY DATASETS ########')
f.write('\n X_train shape: ' + str(b_X_train.shape))
f.write('\n X_test shape: ' + str(b_X_test.shape))
f.write('\n Y_train: ' + str(b_Y_train.shape))
f.write('\n Y_test : ' + str(b_Y_test.shape))
f.write('\n\nBINARY training & binary evaluation -------------------------\n')
f.write('Val score = ' + str(np.around(b_scores[1]*100, 1)) + '%')
f.write('\nTest accuracy = ' + str(round(b_acc_bin*100, 1)) + '%')
f.write('\nConfusion matrix:\n' + str(b_confusion_bin))
f.write('\n\nBINARY training & multiclass evaluation ---------------------\n')
f.write('Test accuracy = ' + str(round(b_acc_multi*100, 1)) + '%')
f.write('\nConfusion matrix:\n' + str(b_confusion_multi))
f.write('\n\nMULTICLASS: ')
if m_optimizer_selected == 'Adam':
f.write('\tAdam(learning_rate = ' + str(m_learn_rate) + ', beta_1 = '
+ str(m_param1) + ', beta_2 = ' + str(m_param2) + ')')
elif m_optimizer_selected == 'AMSgrad':
f.write('\tAMSgrad(learning_rate = ' + str(m_learn_rate) + ', beta_1 = '
+ str(m_param1) + ', beta_2 = ' + str(m_param2) + ')')
elif m_optimizer_selected == 'RMSprop':
f.write('\tRMSprop(learning_rate = ' + str(m_learn_rate) + ', rho = '
+ str(m_param1) + ', momentum = ' + str(m_param2) + ')')
elif m_optimizer_selected == 'SGD':
f.write('\tSGD(learning_rate = ' + str(m_learn_rate) + ', momentum = '
+ str(m_param2) + ')')
f.write(' | n_epochs: ' + str(m_n_epochs_actual) + "/" + str(m_n_epochs) +
' | patience: ' + str(m_patience) + ' | batch_size: ' + str(m_batch_size) +
' | dropoutRate: ' + str(m_dropoutRate) +
'\n\t\tkernLength: ' + str(m_kernLength) + ' | F1: ' + str(m_F1) + ' | D: ' +
str(m_D) + ' | F2: ' + str(m_F2) + ' | kernLength2: ' + str(m_kernLength2))
f.write('\nMULTICLASS DATASETS #####')
f.write('\n X_train shape: ' + str(m_X_train.shape))
f.write('\n X_test shape: ' + str(m_X_test.shape))
f.write('\n Y_train: ' + str(m_Y_train.shape))
f.write('\n Y_test : ' + str(m_Y_test.shape))
f.write('\n\nMULTICLASS training & binary evaluation ----------------------\n')
f.write('Test accuracy = ' + str(round(m_acc_bin*100, 1)) + '%')
f.write('\nConfusion matrix:\n' + str(m_confusion_bin))
f.write('\n\nMULTICLASS training & multiclass evaluation ------------------\n')
f.write('Val score = ' + str(np.around(m_scores[1]*100, 1)) + '%')
f.write('\nTest accuracy = ' + str(round(m_acc_multi*100, 1)) + '%')
f.write('\nConfusion matrix:\n' + str(m_confusion_multi))
print("\n###\n###\nTest accuracy: b_acc = %.2f%% | m_acc = %.2f%%" % (b_acc_bin, m_acc_bin))
print("\n\n--------- E N D O F S E S S I O N ---------\n\n\n")
with open(path0 + 'saveAccuracy.txt', 'a') as f:
f.write("--------- E N D O F S E S S I O N ---------\n\n\n")
with open(path0 + 'number_of_execution.txt', 'w') as f:
f.write(str(date) + '\n' + str(number_run))