[Deep Learning] any layer of the CNN version of the MATLAB implementation

Prior to the implementation of a layer of convolutional neural network, and the implementation of the two-layer convolutional neural network, the next expansion of CNN to any layer is not difficult.

This is the legendary "Daosh One, life two, two three, Sansheng everything"? = =

The code is still on GitHub.

The interesting point is that I used two layers of neural network and a layer of neural network to achieve the results, such as:

Two-story CNN results:

A layer of CNN results:

It can be seen that the first layer of CNN results is better than the two layers, then it is likely that the two-story CNN has been a fitting phenomenon. For mnist This small data set, a layer of CNN plus some parameter tuning is definitely enough.

[Deep Learning] any layer of the CNN version of the MATLAB implementation