Feature maps from the convolutional layer are down sampled to a size determined by the size of the pooling kernel and the size of the pooling kernel’s stride. An activation function is then applied to the resulting image, and a bias is finally added to the output of the activation function. 3.9 illustrates an example …