This question already has an answer here:

In convolutional neural net architectures, I found many Convolutional layers stacked with ReLU layers. What is its purpose? Why would we want to threshold the output of the convolutional layer since it's basically a representation of the input image.


marked as duplicate by Franck Dernoncourt, Michael Chernick, Firebug, mdewey, John Jan 14 '17 at 17:21

This question has been asked before and already has an answer. If those answers do not fully address your question, please ask a new question.