一个简单的卷积神经网络

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
input=[3,4,6,5,7,

       2,4,6,8,2,

       1,6,7,8,4,

       9,7,4,6,2,

       3,7,5,4,1]

#转化成(B,C,W,H)，批量维度，通道，宽，高

input=torch.Tensor(input).view(1,1,5,5)

conv_layer=torch.nn.Conv2d(1,1,kernel_size=3,padding=1,bias=False)

kennel=torch.Tensor([1,2,3,4,5,6,7,8,9]).view(1,1,3,3)

conv_layer.weight.data=kennel.data



output=conv_layer(input)

print(output)

1
样本数量、通道数、宽度、高度确立了一个卷积层。

1
（batch，20,4,4）展开为1为一行，进行变换得到10*1的结果。

1
 利用卷积神经网络做手写数字的识别：

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
import torch

from torchvision import transforms

from torchvision import datasets

from torch.utils.data import  DataLoader

import torch.nn.functional as F

import torch.optim as optim



batch_size=64

#利用transforms 转换为tensor 类型的图片，

transform =transforms.Compose([transforms.ToTensor(),transforms.Normalize((0.1307,),(0.3081,))

                               ])

train_dataset=datasets.MNIST(root='../data/mnist',train=True,download=True,transform=transform)

train_loader= DataLoader(train_dataset,shuffle=True,batch_size=batch_size)





test_dataset=datasets.MNIST(root='../data/mnist',train=False,download=True,transform=transform)

test_loader=DataLoader(train_dataset,shuffle=False,batch_size=batch_size)





class Net(torch.nn.Module):

    def __init__(self):

        super(Net,self).__init__()

        self.conv1=torch.nn.Conv2d(1,10,kernel_size=5)

        self.conv2=torch.nn.Conv2d(10,20,kernel_size=5)

        self.pooling=torch.nn.MaxPool2d(20)

        self.fc=torch.nn.Linear(320,10)

    def forward(self,x):

        batch_size=x.size(0)

        x=F.relu(self.pooling(self.conv1(x)))

        x=F.relu(self.pooling(self.conv2(x)))

        x=x.view(batch_size,-1)  #变成全连接网络需要的输入

        x=self.fc(x)

        return  x



model=Net()

crierion=torch.nn.CrossEntropyLoss()

optimizer=optim.SGD(model.parameters(),lr=0.01,momentum=0.5)



def train(epoch):

    running_loss=0.0

    for batch_idx,data in enumerate(train_loader,0):

        inputs,target=data

        optimizer.zero_grad()

        #forward +backward+update

        outputs=model(inputs)

        loss=crierion(outputs,target)

        loss.backward()

        optimizer.step()

        running_loss+=loss.item()

        if batch_idx%300==299:

            print('[%d,%5d] loss:%.3f' %(epoch+1,batch_idx+1,running_loss/300))

            running_loss==0.0



def test():

    correct=0

    total=0

    with torch.no_grad():

        for data in test_loader:

            images,labels=data

            outputs=model(images)

            _,predicted=torch.max(outputs.data,dim=1)

            total+=labels.size(0)

            correct+=(predicted==labels).sum().item()

    print('Accuacy on test set：%d %%'%(100*correct/total))





if __name__=='__main__':

    for epoch in range(10):

        train(epoch)

        test()

1