Thursday, July 16

MS61
Accuracy and Stability in Numerical Linear Algebra

Sponsored by SIAM Activity Group on Linear Algebra

2:00 PM-4:00 PM
Room: Sidney Smith 2108

Understanding the accuracy and stability of algorithms continues to be a major area of research in numerical linear algebra. The first speaker in this session will discuss the issues involved in defining numerical stability and present a precise definition. The remaining speakers will present recent research in several particular areas: QR factorization and its applications, reduction to bidiagonal form, and generalized eigenvalue and singular value decompositions.

Organizer: Nicholas J. Higham
University of Manchester, United Kingdom
2:00 Should Numerical Stability be Defined by Taking O(machine epsilon) Literally?
Lloyd N. Trefethen, University of Oxford, United Kingdom
2:30 Householder QR Factorization with Complete Pivoting: Why and When to Use It
Anthony J. Cox, University of Manchester, United Kingdom; and Nicholas J. Higham, Organizer
3:00 More Accurate Bidiagonal Reduction for Computing the Singular Value Decomposition
Jesse L. Barlow, Pennsylvania State University
3:30 Backward and Forward Stability in Generalized Eigenvalue and Singular Value Computation
Zlatko Drmac, University of Colorado, Boulder

Program Program Overview Program-at-a-Glance Program Updates Speaker Index Registration Hotel Transportation

LMH Created: 3/19/98; MMD Updated: 4/6/98