Linear Models for Classification

This lecture covers linear models for classification, starting with a recap on simple parametric models and hyperplanes. It delves into linear regression, binary classification, adding non-linearity, logistic regression, and evaluating classifiers. The instructor explains the concept of margin between classes, the formulation for a maximum margin classifier (SVM), and basic notions of constrained optimization. The lecture also touches on decision boundaries, support vectors, and the SVM formulation. An interlude on constrained optimization and Lagrange duality is included, followed by the derivation of the Lagrange dual problem. The use of slack variables in SVM is discussed, along with the Lagrangian and the optimal values. The lecture concludes with a comparison of linear models for classification.