-
Notifications
You must be signed in to change notification settings - Fork 3
/
model.py
128 lines (105 loc) · 5.33 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
import torch
from torch import nn
from functions import ReverseLayerF
class UDAModel(nn.Module):
def __init__(self, n_flattens, n_hiddens, n_class, bn=False):
super(UDAModel, self).__init__()
# feature之后的维度
self.n_flattens = n_flattens
# 全连接层的hidden size
self.n_hiddens = n_hiddens
# 分类器输出的类别个数
self.n_class = n_class
self.feature = nn.Sequential()
self.class_classifier = nn.Sequential()
self.domain_classifier = nn.Sequential()
self.domain_critic = nn.Sequential()
# features 使用1-D CNN提取特征
self.feature.add_module('f_conv1', nn.Conv1d(1, 8, kernel_size=32, stride=2))
self.feature.add_module('f_relu1', nn.ReLU(True))
self.feature.add_module('f_pool1', nn.MaxPool1d(kernel_size=2, stride=2))
self.feature.add_module('f_drop1', nn.Dropout(0.2))
self.feature.add_module('f_conv2', nn.Conv1d(8, 16, kernel_size=16, stride=2))
self.feature.add_module('f_relu2', nn.ReLU(True))
self.feature.add_module('f_pool2', nn.MaxPool1d(kernel_size=2, stride=2))
self.feature.add_module('f_drop2', nn.Dropout(0.2))
self.feature.add_module('f_conv3', nn.Conv1d(16, 32, kernel_size=8, stride=2))
self.feature.add_module('f_relu3', nn.ReLU(True))
self.feature.add_module('f_pool3', nn.MaxPool1d(kernel_size=2, stride=2))
self.feature.add_module('f_drop3', nn.Dropout(0.2))
self.feature.add_module('f_conv4', nn.Conv1d(32, 32, kernel_size=3, stride=2))
self.feature.add_module('f_relu4', nn.ReLU(True))
self.feature.add_module('f_pool4', nn.MaxPool1d(kernel_size=2, stride=2))
self.feature.add_module('f_drop4', nn.Dropout(0.2))
# class_classifier 使用全连接层进行分类
self.class_classifier = nn.Sequential()
self.class_classifier.add_module('c_fc1', nn.Linear(self.n_flattens, self.n_hiddens))
self.class_classifier.add_module('c_relu1', nn.ReLU(True))
self.class_classifier.add_module('c_drop1', nn.Dropout(0.2))
self.class_classifier.add_module('c_fc2', nn.Linear(self.n_hiddens, self.n_hiddens))
self.class_classifier.add_module('c_relu2', nn.ReLU(True))
self.class_classifier.add_module('c_drop2', nn.Dropout(0.2))
self.class_classifier.add_module('c_fc3', nn.Linear(self.n_hiddens, self.n_class))
# domain_classifier 使用全连接层进行领域分类
# 注意区分最后输出是1还是2，loss对应选择BCELoss或CE
self.domain_classifier = nn.Sequential()
self.domain_classifier.add_module('d_fc1', nn.Linear(self.n_flattens, self.n_hiddens))
self.domain_classifier.add_module('d_relu1', nn.ReLU(True))
self.domain_classifier.add_module('d_drop1', nn.Dropout(0.2))
self.domain_classifier.add_module('d_fc2', nn.Linear(self.n_hiddens, self.n_hiddens))
self.domain_classifier.add_module('d_relu2', nn.ReLU(True))
self.domain_classifier.add_module('d_drop2', nn.Dropout(0.2))
self.domain_classifier.add_module('d_fc3', nn.Linear(self.n_hiddens, 2))
# 用于w距离
self.domain_critic = nn.Sequential()
self.domain_critic.add_module('dc_fc1', nn.Linear(self.n_flattens, self.n_hiddens))
self.domain_critic.add_module('dc_relu1', nn.ReLU(True))
self.domain_critic.add_module('dc_drop1', nn.Dropout(0.2))
self.domain_critic.add_module('dc_fc2', nn.Linear(self.n_hiddens, self.n_hiddens))
self.domain_critic.add_module('dc_relu2', nn.ReLU(True))
self.domain_critic.add_module('dc_drop2', nn.Dropout(0.2))
self.domain_critic.add_module('dc_fc3', nn.Linear(self.n_hiddens, 1))
def extract_feature(self, x):
x = self.feature(x)
x = x.view(x.size(0), -1)
return x
def class_classify(self, x):
x = self.feature(x)
x = x.view(x.size(0), -1)
x = self.class_classifier(x)
return x
def dann(self, input_data, alpha):
feature = self.feature(input_data)
feature = feature.view(feature.size(0), -1)
reverse_feature = ReverseLayerF.apply(feature, alpha)
class_output = self.class_classifier(feature)
domain_output = self.domain_classifier(reverse_feature)
return class_output, domain_output, feature
def deepcoral(self, source, target):
loss = 0
source = self.feature(source)
source = source.view(source.size(0), -1)
if self.training == True:
target = self.feature(target)
target = target.view(target.size(0), -1)
loss += CORAL(source, target)
source = self.class_classifier(source)
return source, loss
def ddc(self, source, target):
loss = 0
source = self.feature(source)
source = source.view(source.size(0), -1)
if self.training == True:
target = self.feature(target)
target = target.view(target.size(0), -1)
loss = mmd.mmd_linear(source, target)
source = self.class_classifier(source)
return source, loss
def forward(self, x):
print('forward called')
x = self.feature(x)
x = x.view(x.size(0), -1)
output = self.class_classifier(x)
return output
def output_num(self):
return self.__in_features