Classification

Classification refers to the supervised learning task of predicting a discrete class label $y \in \{1, \dots, C\}$ from a set of input features $\boldsymbol{x}$, as opposed to predicting a continuous value (regression). Given training data $\{(\boldsymbol{x}_i, y_i)\}_{i=1}^N$, a classification model learns a decision rule that assigns new, unseen inputs to one of the $C$ classes.

Performance is commonly summarized through the confusion matrix (which counts correct and incorrect predictions per class), together with derived metrics such as precision (of the predicted positives, how many were correct), recall (of the true positives, how many were found), and the f1-score (their harmonic mean) — exactly as introduced in the basic neural network notebook.

Most methods in this course can be used for classification (neural networks, decision trees, random forests, gradient boosting), and some are specifically built around the classification task: support vector machines, covered next, directly construct a decision boundary that separates the classes with maximum margin.