What is the Perceptron?

The perceptron is is a binary classifier that learns an explicit decision boundary from an example of labelled points, which can then be used to classify new, previously unseen, points. The perceptron is the first published and arguably the simplest artificial neuron, which is the class of models that form the literal basic unit of neural networks and deep learning. The perceptron forms an ideal introductory concept to these topics because of its simplicity.

This article will introduce the three fundamental components in the perceptron model: weights, summation, and activation function. It should be noted that the perceptron has these three elements in common with all artificial neurons, and that it differs from all other types of artificial neurons only by the activation function.

After describing the perceptron model, this article will move into training the model. Then, we will cover the famous limitations of the model, including the XOR problem, which served as motivation for the development of neural networks.