Linear Models for Classification

This lecture covers linear models for classification, starting with the basic linear model in dimension D and its application to binary classification. It then explores adding non-linearity using the logistic sigmoid function, leading to logistic regression. Decision boundaries and the Support Vector Machine (SVM) formulation are discussed, along with handling overlapping classes and introducing slack variables. The lecture also delves into multi-class classification, comparing least-square classification, logistic regression, and SVM. Practical examples include predicting fetal state from cardiotocography data and thyroid disease classification. The lecture concludes with a comparison of linear classifiers on datasets like Iris and MNIST.