Convolutional neural network (including code and corresponding diagram)

1. Related calculation problems in Convolutional Neural Networks

（1） Simple two-dimensional convolution

（2） Add fill （padding）
notes ： Image below ph Is set in the code padding Twice the value

（3） Add fill （padding） And stride （stride）

Convolution kernel and filter （fliter） There is a difference , Convolution kernel is a two-dimensional concept , The filter is composed of one or more convolution kernels .

2. Learn the construction of convolutional neural network through an example

Build the neural network in the figure above through the following code

class ConvNet(nn.Module):
    def __init__(self):
        super(ConvNet, self).__init__()
        self.layer1 = nn.Sequential(
            nn.Conv2d(1, 32, kernel_size=5, stride=1, padding=2),
            nn.ReLU(),
            nn.MaxPool2d(kernel_size=2, stride=2))
        self.layer2 = nn.Sequential(
            nn.Conv2d(32, 64, kernel_size=5, stride=1, padding=2),
            nn.ReLU(),
            nn.MaxPool2d(kernel_size=2, stride=2))
        self.drop_out = nn.Dropout()
        self.fc1 = nn.Linear(7 * 7 * 64, 1000)
        self.fc2 = nn.Linear(1000, 10)
	def forward(self, x):
    	out = self.layer1(x)
    	out = self.layer2(out)
    	out = out.reshape(out.size(0), -1)
    	out = self.drop_out(out)
    	out = self.fc1(out)
    	out = self.fc2(out)
    	return out

3. Convolution neural network related content

Filling and stride in convolution layer

When the convolution kernel is relatively large or the image is too small after multiple convolutions, consider filling operation to alleviate

When the input picture size is relatively large , Under the small convolution kernel, it needs many layers of calculation to realize






8-5+1+4=8
8-3+1+2=8

8 / 2 = 4

（8-3+0+3）/3=2
（8-5+2+4）/4=2

Multiple input multiple output channels in convolution layer