Sunday, May 11

MS36
Optimization Techniques for Nonnegative Matrix Factorization

3:00 PM - 5:00 PM
Room: Beacon Hill - CC

Nonnegative matrix factorization (NMF) has emerged as a powerful tool for information retrieval with applications to text, images, biochemical experimental results, and even music. NMF is most naturally posed as a nonconvex continuous optimization problem. The language and techniques of optimization such as convergence behavior, local and global optima, complexity, etc., can be usefully applied to NMF. Speakers in this minisymposium will present recent work in optimization approaches to NMF and applications.

Organizer: Stephen A. Vavasis
University of Waterloo, Canada

3:00-3:25 Using Nonnegative Matrix and Tensor Factorizations for Topic Detection and Tracking

Michael W. Berry, University of Tennessee; Brett W. Bader, Sandia National Laboratories

3:30-3:55 Effective Initializations for NMF algorithms

Efstratios Gallopoulos, University of Patras, Greece; Christos Boutsidis, Rensselaer Polytechnic Institute

4:00-4:25 Greedy Algorithms and Complexity for Nonnegative Matrix Factorization

Stephen A. Vavasis, Ali Ghodsi, and Michael Biggs, University of Waterloo, Canada

4:30-4:55 Nonnegative Matrix Factorizations and Clustering

Haesun Park and Jingu Kim, Georgia Institute of Technology