Neural Network Projects with Python

Max pooling

In CNNs, it is common to place a max pooling layer immediately after a convolution layer. The objective of the max pooling layer is to reduce the number of weights after each convolution layer, thereby reducing model complexity and avoiding overfitting.

The max pooling layer does this simply by looking at each subset of the input passed to it, and throwing out all but the maximum value in the subset. Let's take a look at an example to see what this means. Assume that our input to the max pooling layer is a 4 x 4 tensor (a tensor is just an n-dimensional array, such as those output by a convolutional layer), and we are using a 2 x 2 max pooling layer. The following diagram illustrates the Max Pooling operation:

As we can see from ...