1.對LeNet-5的改進

將kernel_size=5的卷積層變為兩層kernel_size=3的卷積層，並加入batch_normalization，具體實現如下：

self.conv1=nn.Sequential(nn.Conv2d(in_channels=1,out_channels=6,kernel_size=3),
                                   nn.Conv2d(in_channels=6,out_channels=6,kernel_size=3),
                                   nn.BatchNorm2d(
 
6))
self.conv2=nn.Sequential(nn.Conv2d(in_channels=6,out_channels=16,kernel_size=3),
                                   nn.Conv2d(in_channels=16,out_channels=16,kernel_size=3),
                                   nn.BatchNorm2d(16))

共100epoch，在第63個epoch達到了99.25%的準確率

優化器選用SGD，訓練中學習率的調整如下所示：

optimizer=
 
optim.SGD(network.parameters(),lr=0.05)
scheduler= MultiStepLR(optimizer, milestones=[30,70], gamma=0.1)