Why Machines Learn: The Elegant Math Behind Modern AI

by Anil Ananthaswamy

If a network requires more than two weight matrices (one for the output layer and one for each hidden layer), then it’s called a deep neural network: