-
Notifications
You must be signed in to change notification settings - Fork 0
/
main_eegnet.py
124 lines (97 loc) · 3.66 KB
/
main_eegnet.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
import tensorflow as tf
import numpy
import random
from tensorflow.keras.callbacks import ModelCheckpoint
from model.EEGModels import EEGNet
from get_data_data import eegmat,stew,nback,eegmat_test,stew_test,nback_test,eegmat_val,stew_val,nback_val
from sklearn.metrics import accuracy_score,f1_score
numpy.random.seed(42)
tf.random.set_seed(42)
random.seed(42)
class Main:
def __init__(self,train=None,valid=None,test=None):
self.train=train
self.valid=valid
self.test=test
def select_data(self,data):
d=[]
l=[]
c=[]
for i,j,k in data:
i=i.reshape(i.shape+(1,))
j=tf.one_hot(j,2)
d.append(i)
l.append(j)
c.append(k)
d=numpy.stack(d)
l=numpy.stack(l)
c=numpy.stack(c)
return d,l,c
def cut_sub(self,x,y,c):
sub=set(c.tolist())
num_sub=len(sub)
num_sample=len(c)
sps=num_sample//num_sub
xx=x.reshape(num_sub,sps,x.shape[-3],x.shape[-2],x.shape[-1])
yy=y.reshape(num_sub,sps,y.shape[-1])
cc=c.reshape(num_sub,sps)
return xx,yy,cc
def score(self,labels,probs):
preds=probs.argmax(axis=-1)
labels=labels.argmax(axis=-1)
acc=accuracy_score(y_true=labels,y_pred=preds)
f1=f1_score(y_true=labels,y_pred=preds,average="macro")
#print(acc,f1)
return acc,f1
def clf(self,d1,t1,c1,d2,t2,c2,d3,t3,c3):
eegnet=EEGNet(nb_classes=2,Chans=10,Samples=2560,dropoutRate=0.25,norm_rate=0.25)
opt=tf.keras.optimizers.Adam(learning_rate=0.0001)
crt=tf.keras.losses.CategoricalCrossentropy()
inp=tf.data.Dataset.from_tensor_slices((d1,t1))
inp=inp.shuffle(buffer_size=1024).batch(10)
for i in range(5):
for d,t in inp:
with tf.GradientTape() as tape:
p=eegnet(d,training=True)
loss=crt(t,p)
grads=tape.gradient(loss,eegnet.trainable_weights)
opt.apply_gradients(zip(grads,eegnet.trainable_weights))
print(i,loss)
#eegnet.compile(loss='categorical_crossentropy',optimizer='adam',metrics=['accuracy'])
#eegnet.fit(d1,t1,batch_size=10,epochs=10,verbose=2)
#d3,t3,c3=self.cut_sub(d3,t3,c3)
#d2,t2,c2=self.cut_sub(d2,t2,c2)
p1=eegnet.predict(d1,verbose=3)
#acc=eegnet.evaluate(d1,t1,verbose=2)
acc,f1=self.score(t1,p1)
print(acc,f1)
print()
p2=eegnet.predict(d2,verbose=3)
p3=eegnet.predict(d3,verbose=3)
acc2,f12=self.score(t2,p2)
acc3,f13=self.score(t3,p3)
print(acc2,'\t',f12)
print(acc3,'\t',f13)
'''
for idx,(d,t,c) in enumerate(zip(d2,t2,c2)):
p=eegnet.predict(d,verbose=3)
acc,f1=self.score(t,p)
print(acc,"\t",f1)
print()
for idx,(d,t,c) in enumerate(zip(d3,t3,c3)):
p=eegnet.predict(d,verbose=3)
acc,f1=self.score(t,p)
print(acc,"\t",f1)
'''
return
def forward(self):
d1,t1,c1=self.select_data(self.train)
d2,t2,c2=self.select_data(self.valid)
d3,t3,c3=self.select_data(self.test)
self.clf(d1,t1,c1,d2,t2,c2,d3,t3,c3)
return
if __name__=='__main__':
#print(len(nback),len(eegmat),len(stew))
main=Main(train=stew,valid=nback_val,test=eegmat)
main.forward()
print("a")