1. LeNet5 network overview

LeNet5 Network is a Convolutional Neural Network (CNN) with five hidden layers, and its basic Network structure is as follows

graph TD
Input --> Convolution1
Convolution1 --> subsampling1
subsampling1 --> Convolution2
Convolution2 --> subsampling2
subsampling2 --> full_connection
full_connection --> Convolution3
Convolution3 --> output

2. Basic idea of LeNet5 network

LeNet5 network has various structures of convolutional neural network, and its main structure is as follows:

  • The input layer
  • Convolution at the Convolution layer: feature extraction is achieved
  • Pooling layer, subsampling, and data dimension reduction
  • Full connection layer: carry on the operation to the obtained feature graph, train the parameters and get the output
  • Output layer: In a common classification task, there are as many output nodes as there are possible outputs. For example, in the classification task of handwritten number recognition, the number of output nodes is 10