构建网络 练习题三道

1 问题

构建网络1：

1. 
   [_,1,28,28]
   [_,16,14,14]
   [_,32,7,7]
   [_,32,1,1]
   [_,32]
   [_,64]
   [_,10]

2. 构建网络2：

   
   [_,1,28,28]
   [_,16,14,14]
   [_,32,7,7]
   [_,32,1,1]
   [_,10,1,1]
   [_,10]

3. 构建网络3：

[_,1,28,28]

[_16,14,14]

[_,128,14,14]

[_,256,7,7]

[_,256,1,1]

[_,256]

[_,10]

2 方法

构建网络1： import torch.nn as nn class Net(nn.Module):     def __init__(self):         super(Net, self).__init__()         self.conv1 = nn.Conv2d(1, 16, 3, padding=1)         self.conv2 = nn.Conv2d(16, 32, 3, padding=1)         self.conv3 = nn.Conv2d(32, 32, 3, padding=1)         self.conv4 = nn.Conv2d(32, 32, 3, padding=1)         self.conv5 = nn.Conv2d(32, 64, 3, padding=1)         self.fc1 = nn.Linear(64, 10)     def forward(self, x):         x = nn.functional.relu(self.conv1(x))         x= nn.functional.max_pool2d(nn.functional.relu (self.conv2(x)), 2)         x = nn.functional.relu(self.conv3(x))         x = nn.functional.max_pool2d(nn.functional.relu (self.conv4(x)), 2)         x = nn.functional.max_pool2d(nn.functional.relu (self.conv5(x)), 2)         x = x.view(-1, 64)         x = self.fc1(x)         return x 构建网络2： import torch.nn as nn class Net(nn.Module):     def __init__(self):         super(Net, self).__init__()         self.conv1 = nn.Conv2d(1, 16, 3, padding=1)         self.conv2 = nn.Conv2d(16, 32, 3, padding=1)         self.conv3 = nn.Conv2d(32, 32, 3, padding=1)         self.conv4 = nn.Conv2d(32, 10, 3, padding=1)         self.fc1 = nn.Linear(10, 10)     def forward(self, x):         x = nn.functional.relu(self.conv1(x))         x = nn.functional.max_pool2d(nn.functional.relu (self.conv2(x)), 2)         x = nn.functional.relu(self.conv3(x))         x = nn.functional.max_pool2d(nn.functional.relu (self.conv4(x)), 2)         x = nn.functional.adaptive_avg_pool2d(x, (1, 1))         x = x.view(-1, 10)         x = self.fc1(x)         return x 构建网络3： import torch.nn as nn class Net(nn.Module):     def __init__(self):         super(Net, self).__init__()         self.conv1 = nn.Conv2d(1, 16, 3, padding=1)         self.conv2 = nn.Conv2d(16, 128, 3, padding=1)         self.conv3 = nn.Conv2d(128, 256, 3, padding=1)         self.conv4 = nn.Conv2d(256, 256, 3, padding=1)         self.conv5 = nn.Conv2d(256, 256, 3, padding=1)         self.fc1 = nn.Linear(256, 10)     def forward(self, x):         x = nn.functional.relu(self.conv1(x))         x = nn.functional.max_pool2d(nn.functional.relu (self.conv2(x)), 2)         x = nn.functional.relu(self.conv3(x))         x = nn.functional.max_pool2d(nn.functional.relu (self.conv4(x)), 2)         x = nn.functional.relu(self.conv5(x))         x = nn.functional.adaptive_avg_pool2d(x, (1, 1))         x = x.view(-1, 256)         x = self.fc1(x)         return x

3 结语

本次实验我们做了三个练习题，需要的知识在课堂上已经有过演示了。在实际编写代码中我们运用了nn. Conv2d()进行卷积层的构建；利用激活函数nn.ReLU()，使分类效果更好；整体来说是比较简单的，编写过程中也没有遇到什么问题。