1:30 PM-2:30 PM
Chair: Ake Bjorck, Linkoping University, Sweden
Room: Ballroom 3
Fast algorithms based on generalizations of the classical Schur algorithm,for solving least squares problems with Toeplitz structure, are discussed. Such methods are potentially very fast, but they may suffer from stability problems, when the matrix is ill-conditioned. The speaker will give an overview of generalized Schur methods, and discuss their stability properties. Postprocessing methods that can be used in order to obtain a more accurate solution will be presented. Some numerical tests will also be reported, where different variants of the generalized Schur algorithm and other fast methods for Toeplitz least squares and Toeplitz linear systems will be compared for their accuracy and speed. (Joint work with Lars Elden, Linkoping University, Sweden).
Computer Science Department
University of Minnesota, Minneapolis
LA97 Homepage | Program Updates| Registration | Hotel Information | Transportation | Program-at-a-Glance |