Multi-Class Classification

With binary classification, you were limited to dealing with target variables that can only take two possible values: 0 and 1 (false or true). Multi-class classification can be seen as an extension of this and allows the target variable to have more than two values (or you can say binary classification is just a subset of multi-class classification). For instance, a model that predicts different levels of disease severity for a patient or another one that classifies users into different groups based on their past shopping behaviors will be multi-class classifiers.

In the next section, you will dive into the softmax function, which is used for multi-class classification.

The Softmax Function

Binary classifiers require a specific activation function for the last fully connected layer of a neural network, which is sigmoid. The activation function specific to multi-class classifiers is different. It is softmax. Its formula is as follows:

...