classifier

A trained model that assigns one of a fixed set of labels to an input.

A classifier is a function $g$ that maps an input -- here, an image or an image's feature vector -- to one label from a fixed, known set of classes. k-NN, an SVM, and a random forest (Section 2) are three different ways to build $g$ from the same labeled feature vectors; a neural network (Chapter 11) is a fourth. All four are judged by the same yardstick: how often $g$ 's label matches the true label on data it was not trained on.

Examples

A classifier trained on labeled "cat" and "dog" feature vectors outputs one of those two labels for every new feature vector it is given, never a third option it was not trained on.