Hands-On One-shot Learning with Python

Neural networks – learning procedure

As we know, neural networks learn through minimizing the loss function (or the objective function) using the stochastic gradient descent optimization method. So, loss functions are one of the major factors that determine the objective of neural network architecture. For example, if we want to classify data points, we will choose loss functions such as categorical cross-entropy, 0-1 loss, and hinge loss; whereas, if our objective is regression, we will choose loss functions such as mean squared error, root mean squared error, and Huber loss. Some of the common equations are as follows:

The first thing that ...