Java Deep Learning Cookbook

Backpropagation

The backbone of a neural network is the backpropagation algorithm. Refer to the sample neural network structure shown as follows:

For any neural network, data flows from the input layer to the output layer during the forward pass. Each circle in the diagram represents a neuron. Every layer has a number of neurons present. Our data will pass through the neurons across layers. The input needs to be in a numerical format to support computational operations in neurons. Each neuron in the neural network is assigned a weight (matrix) and an activation function. Using the input data, weight matrix, and an activation function, a probabilistic ...