Mathematics Colloquia and Seminars

Return to Colloquia & Seminar listing

In this talk we consider data clustering using graph models --- the pairwise similarities between all data objects form a weighted graph adjacency matrix that contains all necessary information for clustering. We propose new algorithms for graph partition with an objective function that follows the min-max clustering principle. We show relationships between these new algorithms and certain algebraic structures in the eigenvector and singular vector matrices computed from the clustering data. We demonstrate the effectiveness of these methods via data extracted from newsgroup articles. Brief Bio: Ming Gu received his PhD (1993) degree in Computer Science from Yale University. He has been with the UCLA math faculty since 1996 and joined the Berkeley faculty in July 2000. His research interests include numerical linear algebra, fast algorithm and optimization.

Coffee/cookies @ 693 Kerr