Pooling Layer

Overview

Pooling is a process, similar to convolution that reduces the size of an image in order to make it more computationally feasible to process. As an example, consider a matrix that represents an image.



Next, take the following filter. It maps sections of the original image to a new value, given by the max of the image pixel values in that region.





Other aggreations can be used rather than max, for example, average could be used instead.