Spectral Clustering: Finding Clusters

This lecture covers the concept of Spectral Clustering, focusing on building similarity graphs, measuring distances, and identifying connected components. The instructor explains how to perform eigenvalue decomposition and Laplacian matrix construction to determine the number of clusters in a dataset. Various exercises are provided to practice building similarity matrices and Laplacian matrices using different kernels. The lecture also discusses the role of eigenvalues in spectral clustering and the process of finding clusters through eigenvector projections. Additionally, the lecture explores the equivalency of Laplacian Eigenmaps with other non-linear embeddings and the effect of distance functions on clustering results.