DeepLearning5.4, The gradient of a convolutional layer

views comments

Optimizing the filters of a convolutional layer requires to apply the chain rule across the MaxPooling stage. Here the essential steps of the calculation are shown and an interesting interpretation is given.