-
Notifications
You must be signed in to change notification settings - Fork 0
/
MLP.py
100 lines (89 loc) · 4.12 KB
/
MLP.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
import torch.nn as nn
class my_MLP(nn.Module):
def __init__(self):
super(my_MLP,self).__init__()
self.input_h=28
self.input_w=28
self.passageway=1
self.hiddendim1=20#隐藏层的维度
self.hiddendim2=40#隐藏层的维度
self.classes=1000#drop数
self.outputdim=10#输出的类别数
# self.linear1=nn.Linear(self.passageway*self.input_h*self.input_w,self.hiddendim1)
# self.conv1=nn.Conv2d(1,self.hiddendim1,kernel_size=5,padding=1)#1*28*28->20*24*24
# self.relu=nn.ReLU()
# self.conv2=nn.Conv2d(self.hiddendim1,self.hiddendim2,kernel_size=5,padding=1)#20*24*24->40*20*20
# self.linear1=nn.Linear(self.hiddendim2*4*4,self.classes)
# self.linear2=nn.Linear(self.classes,self.outputdim)
# self.linear=nn.Linear(self.hiddendim2*4*4,self.outputdim)
# self.softmax=nn.Softmax(dim=1)
# self.pool=nn.MaxPool2d(kernel_size=2,stride=2)
# self.dropout=nn.Dropout(0.5)
self.conv1=nn.Sequential(
nn.Conv2d( #--> (1,28,28)
in_channels=1, #传入的图片是几层的，灰色为1层，RGB为三层
out_channels=20, #输出的图片是几层
kernel_size=5, #代表扫描的区域点为5*5
stride=1, #就是每隔多少步跳一下
padding=2, #边框补全，其计算公式=（kernel_size-1）/2=(5-1)/2=2
), # 2d代表二维卷积 --> (16,28,28)
nn.ReLU(), #非线性激活层
nn.MaxPool2d(kernel_size=2), #设定这里的扫描区域为2*2，且取出该2*2中的最大值 --> (16,14,14)
)
self.conv2=nn.Sequential(
nn.Conv2d( #--> (20,14,14)
in_channels=20, #传入的图片是几层的，灰色为1层，RGB为三层
out_channels=40, #输出的图片是几层
kernel_size=5, #代表扫描的区域点为5*5
stride=1, #就是每隔多少步跳一下
padding=2, #边框补全，其计算公式=（kernel_size-1）/2=(5-1)/2=2
), # 2d代表二维卷积 --> (40,14,14)
nn.ReLU(), #非线性激活层
nn.MaxPool2d(kernel_size=2), #设定这里的扫描区域为2*2，且取出该2*2中的最大值 --> (40,7,7)
)
self.linear1=nn.Linear(40*7*7,100)
self.linear2=nn.Linear(100,100)
self.linear3=nn.Linear(100,10)
self.softmax=nn.Softmax(dim=1)
self.dropout=nn.Dropout(0.5)
self.relu=nn.ReLU()
'''self.net=nn.Sequential(
nn.Linear(self.input_h*self.input_w,256),
nn.ReLu(),
nn.Linear(256,10)
)'''
def forward(self,input):
#input B*(P*H*W) -> B*F
# input=input.view(input.shape[0],-1)
# input_linear1=self.linear1(input)
# input_relu1=self.relu1(input_linear1)
# input_linear2=self.linear2(input_relu1)
# input_relu2=self.relu1(input_linear2)
# output=self.linear3(input_relu2)
'''
input_conv1=self.conv1(input)
input_pool1=self.pool(input_conv1)
input_relu1=self.relu(input_pool1)
input_conv2=self.conv2(input_relu1)
input_pool2=self.pool(input_conv2)
input_relu2=self.relu(input_pool2)
input_dropout=self.dropout(input_relu2)
input_linear1=self.linear1(input_dropout)
# input_linear1=self.linear1(input_relu2)
input_relu=self.relu(input_linear1)
input_dropout=self.dropout(input_relu)
input_linear2=self.linear2(input_dropout)
input_relu=self.relu(input_linear2)
output=self.softmax(input_relu)
#output=self.softmax(input_relu)
'''
x=self.conv1(input)
x=self.conv2(x)
x=x.view(x.shape[0],-1)
x=self.linear1(x)
x=self.linear2(x)
output=self.linear3(x)
# output=self.softmax(x)
# x=self.relu(x)
# output=self.softmax(x)
return output