-
Notifications
You must be signed in to change notification settings - Fork 7
/
Model.py
133 lines (99 loc) · 3.76 KB
/
Model.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
# -*- coding: utf-8 -*-
"""
Created on Thu Jul 20 09:47:45 2017
@author: lenovo
"""
from keras.layers import merge
from keras.layers.convolutional import Conv1D
from keras.layers.pooling import MaxPooling1D
from keras.layers import *
from keras.layers.core import Dense,Activation,Flatten,Dropout,Flatten
from keras.layers.normalization import BatchNormalization
from keras.models import Model
from keras.layers import Input
def block_type1(x,nb_filter,filter_len=16):
out = Conv1D(nb_filter,filter_len,padding='same')(x)
out = BatchNormalization()(out)
out = Activation('relu')(out)
out=Dropout(0.5)(out)
out = Conv1D(nb_filter,filter_len,padding='same')(out)
#out = merge([out,x],mode='sum')
return out
def block_type2(x,nb_filter,filter_len=16):
out = BatchNormalization()(x)
out = Activation('relu')(out)
out=Dropout(0.5)(out)
out = Conv1D(nb_filter,filter_len,padding='same')(out)
out = BatchNormalization()(out)
out = Activation('relu')(out)
out=Dropout(0.5)(out)
out = Conv1D(nb_filter,filter_len,padding='same')(out)
#out = merge([out,x],mode='sum')
return out
###
#inp = Input(shape=(200,30))
inp = Input(shape=(15191,1))
inp_begin= Conv1D(64,16,padding='same')(inp)
inp_begin = BatchNormalization()(inp_begin)
inp_begin= Activation('relu')(inp_begin)
###
out_1=block_type1(inp_begin,64,filter_len=16)
#maxpooling1
inp_1=MaxPooling1D(pool_size=2,padding='valid')(inp_begin)
out_1=MaxPooling1D(pool_size=2,padding='valid')(out_1)
out_1=merge([out_1,inp_1],mode='sum')
out_2=block_type2(out_1,64,filter_len=16)
out_2=block_type2(out_2,64,filter_len=16)
#maxpooling2
out_2=MaxPooling1D(pool_size=2,padding='valid')(out_2)
inp_2=MaxPooling1D(pool_size=2,padding='valid')(inp_1)
out_2=merge([out_2,inp_2],mode='sum')
out_3=block_type2(out_2,64,filter_len=16)
inp_3= Conv1D(64*2,1,padding='same')(inp_2)
out_3=block_type2(out_3,64*2,filter_len=16)
#maxpooling3
out_3=MaxPooling1D(pool_size=2,padding='valid')(out_3)
inp_3=MaxPooling1D(pool_size=2,padding='valid')(inp_3)
out_3=merge([out_3,inp_3],mode='sum')
out_4=block_type2(out_3,64*2)
out_4=block_type2(out_4,64*2)
#maxpooling4
out_4=MaxPooling1D(pool_size=2,padding='valid')(out_4)
inp_4=MaxPooling1D(pool_size=2,padding='valid')(inp_3)
out_4=merge([out_4,inp_4],mode='sum')
out_5=block_type2(out_4,64*2)
inp_5= Conv1D(64*3,1,padding='same')(inp_4)
out_5=block_type2(out_5,64*3)
#maxpooling5
out_5=MaxPooling1D(pool_size=2,padding='valid')(out_5)
inp_5=MaxPooling1D(pool_size=2,padding='valid')(inp_5)
out_5=merge([out_5,inp_5],mode='sum')
out_6=block_type2(out_5,64*3)
out_6=block_type2(out_6,64*3)
#maxpooling 6
out_6=MaxPooling1D(pool_size=2,padding='valid')(out_6)
inp_6=MaxPooling1D(pool_size=2,padding='valid')(inp_5)
out_6=merge([out_6,inp_6],mode='sum')
out_7=block_type2(out_6,64*3)
inp_7= Conv1D(64*4,1,padding='same')(inp_6)
out_7=block_type2(out_7,64*4)
#maxpooling 7
out_7=MaxPooling1D(pool_size=2,padding='valid')(out_7)
inp_7=MaxPooling1D(pool_size=2,padding='valid')(inp_7)
out_7=merge([out_7,inp_7],mode='sum')
#out_8=block_type2(out_7,64*4)
#out_8=block_type2(out_8,64*4)
#out_8=MaxPooling1D(pool_size=2,padding='valid')(out_8)
#inp_8=MaxPooling1D(pool_size=2,padding='valid')(inp_7)
#out_8=merge([out_8,inp_8],mode='sum')
#out_8=block_type2(out_8,64*4)
out_final = BatchNormalization()(out_7)
out_final= Activation('relu')(out_final)
#out_final=Dense(14)(out_final)
out_final=Flatten()(out_final)
out_final=Dense(15191)(out_final)
out_final= Activation('softmax')(out_final)
model = Model(inp,out_final)
model.compile(optimizer='adam',loss='mse')
bspdn=np.reshape(bspdn,[54,15191,1])
model.fit(bspdn,epdn[0:54,:],epochs=50,batch_size=2)