The parameters of the convolutional neural network are actually the parameters of the convolution kernel

  • The convolution kernel is equivalent to a feature extractor, which has four hyperparameters: stride(step),padding,K(quantity) and F(length).

  • The activation function adds a nonlinear element.

  1. The pooling layer is equivalent to scaling the size of the image to work with the feature extractor.
  2. It essentially says that the convolution kernel, as defined, does not count as a parameter.

  • The full connection layer is generally used for classification to break the spatial limitation of convolution features.

  • The calculation formula of convolutional neural network is: N=(W-F+2P)/S+1

N: output size W: input size F: convolution kernel size P: fill value size S: step size