Ali Ghodsi, Lec 6: Spectral Clustering, Laplacian Eigenmap, MVU

Ali Ghodsi's lecture on January 24, 2017 for STAT 442/842: Data Visualization, held at the University of Waterloo. Continuation of Spectral Clustering algorithm. Extension to the Laplacian Eigenmap dimensionality reduction technique. Towards a unified framework: how all dimensionality reduction algorithms so far are essentially variations of kernel-PCA. Using this insight, introduce the semidefinite optimization problem of Maximum Variance Unfolding to find the "best" kernel.