Training is the process by which we teach a neural network to learn, and it's controlled programmatically in our code. Recall, from Chapter 2, Machine Learning Basics, that there are two forms of learning in the AI world: supervised learning and unsupervised learning. In general, most ANNs are supervised learners, and they learn by example from a training set. A singular unit in a training cycle in a neural network is called an epoch. By the end of one epoch, your network has been exposed to every data point in the dataset once. At the end of one epoch, epochs are defined as a hyperparameter when first setting up your network. Each of these epochs contain two processes: forward propagation and backpropagation.
Within an epoch we have iterations, which tell us what proportion of the data has gone through both forward propagation and backpropagation. For...
