On this page

4. Discriminative Classification

So far we have studied continuous prediction via regression, where we try to predict the value of some unknown given input features. However, we often encounter discrete problems where we are presented with a number of options and want to make the most appropriate choice. Simply put, we want to maximize the probability of choosing correctly. Traditionally, this is denoted as a “classification” problem where we choose the correct class for a given object.

This section will introduce the most popular discriminative classification algorithms, meaning that they directly model the probabilities of a data point belonging to a specific class.