Thursday, July 17

10:30 AM-12:30 PM
Law School, Room 190

MS61
Theory and Applications of Orthogonal Decompositions

Many numerical algorithms in applications require good approximations to the largest or smallest singular values and corresponding singular subspaces associated with a nearly rank-deficient m X n matrix A. Some of these applications also involve updating and downdating. The singular value decomposition is not practical for some problems due to real-time or storage constraints. However, the needed information can often be obtained by a more general orthogonal decomposition of the matrix A = UTV^T, which is a product of three matrices: an orthogonal matrix U, a middle matrix T, and another orthogonal matrix V. We consider fast and reliable methods based on orthogonal decompositions to solve problems arising in signal processing, spectroscopy, and information retrieval applications.

Organizers: Ricardo D. Fierro, California State University, San Marcos; and Sabine Van Huffel, Katholieke Universiteit Leuven, Belgium

10:30 Recursive Algorithm for Approximate Rank-Revealing Subspace Projection Based Upon a One-Sided Decomposition

Mark J. Smith and I. K. Proudler, Defence Research Agency, United Kingdom

11:00 Downdating Orthogonal Decompositions for Information Retrieval Models

Michael W. Berry and Dian I. Witter, University of Tennessee, Knoxville

11:30 Modifying the ULV Decomposition and Recursive TLS Problems

Jesse L. Barlow, Pennsylvania State University; and Zhenyue Zhang, Zheijiang University, People's Republic of China

12:00 Fast Low Rank-Revealing ULV Algorithms for Toeplitz Matrix Approximations with Applications in Magnetic Resonance Spectroscopy

Leentje Vanhamme, Katholieke Universiteit Leuven, Belgium; Sabine Van Huffel and Ricardo D. Fierro, Organizers

AN97 Homepage | Program Updates|
Registration | Hotel and Dormitory Information | Transportation | Program-at-a-Glance | Program Overview