The perceptron is a special type of neural network that is especially useful for recognizing and classifiying various types of patterns. The simplest perceptron is the single layer perceptron, developed by Rosenblatt in 1958, which is capable of classifying linearly separable patterns as explained below. More complicated perceptrons involve hidden layers and additional nonlinearities introduced into neuronal output functions. In the 1970's, after the breakthrough discovery of the back propagation algorithm used to train these multilayer perceptrons, this particular class of neural networks gained popularity and were successfully used in many pattern classification problems.

Like any neural network, a perceptron consists of a set of nodes, or neurons. Each neuron N has a set of inputs x_i and one output which can subsequently serve as the input to another neuron. Each input is multiplied by a weight w_i. The neuron N computes the weighted sum of its...