Support Vector Machines: Parameters, Solutions, and Boundaries

This lecture covers the comparison of KNN, SVM, and GMM in terms of the number of parameters required for prediction. It discusses the minimum number of Support Vectors needed in linear SVM and the uniqueness of solutions. The impact of adding more datapoints on the number of Support Vectors is explained, along with the construction of decision boundaries using SVM. The lecture also delves into non-linear classification and the computation of class labels in a multi-class SVM, emphasizing the effect of kernel width on decision boundaries.