The linear perceptron is one of the simplest learning algorithms for a two-class classifier. Given a set of data points in d-dimensions, belonging to two classes, C₁ and C₂, the algorithm tries to find a linear separating hyper-plane between the samples of the two classes.

Key Concepts

• If the samples are in one dimension, the separating hyperplane would be a point
• If the samples are in two dimensions, the separating hyperplane would be a line
• If the samples are in three dimensions, the separating hyperplane would be a plane

The specific algorithm that we look into is a special case of a class of algorithms that uses gradient descent on a carefully defined objective function to arrive at a solution.