While I was writing my custom program, I realized that some of the functions could be a little tricky to handle. I tried to make this article not just another article on how CNN works or the mathematics behind it, but more about some of the techniques on how data is structured/formatted while it moves from the convolution layers to maxpool, to the fully connected layer, and back again.

I will take an example to explain how the convolve, max pool, FC, and backpropagation datasets will look with details on some of the key functions

Source link