An illustrated and intuitive information on how Neural Networks be taught

Welcome to Half 2 of the Deep Studying Illustrated sequence. Within the earlier article (undoubtedly learn that first!), we lined how a neural community works and the way a educated neural community makes predictions.

On this article, we’ll delve into the coaching course of and discover how a neural community learns.

📣 In case you haven’t learn my earlier articles, I extremely advocate you begin with my sequence of articles protecting the fundamentals of machine studying, particularly the one on Gradient Descent since you’ll discover that numerous the fabric lined there may be related right here.

Machine Studying Starter Pack

Let’s say we need to create a neural community that predicts the day by day income of ice cream gross sales utilizing the options temperature and day of the week.

That is the (artificial) coaching dataset we’re utilizing:

To construct a neural community, as we discovered within the earlier article, we have to first determine on its structure. This consists of figuring out the variety of hidden layers, the variety of neurons in every layer, and the activation operate of every neuron.

Let’s say we determined our structure is: 1 hidden layer with 2 neurons, and 1 output neuron, all utilizing the rectifier activation operate.

Terminology segue: Within the earlier article, we discovered about utilizing subscripts to distinguish between completely different weights. We’re sticking with the identical conference right here, and as well as, we‘ll use superscripts to…