Practical Convolutional Neural Networks

Pooling layer

As we have seen, a convolutional layer is a stack of feature maps, with one feature map for each filter. More filters increase the dimensionality of convolution. Higher dimensionality indicates more parameters. So, the pooling layer controls overfitting by progressively reducing the spatial size of the representation to reduce the number of parameters and computation. The pooling layer often takes the convolutional layer as input. The most commonly used pooling approach is max pooling. In addition to max pooling, pooling units can also perform other functions such as average pooling. In a CNN, we can control the behavior of the convolutional layer by specifying the size of each filter and the number of filters. To increase the ...