AF: Small: RUI: Eigenvalue Computation via the QR algorithm: Advanced deflation techniques

South Dakota School Of Mines And Technology, Rapid City SD

Investigators

Abstract

Eigenvalue problems arise in many diverse areas of science and engineering, including, but by no means limited to nuclear magnetic resonance, signal processing, speech processing, and control theory. This project will build on previous algorithmic innovations to further accelerate the numerical solution of one class of these important problems. This research is typical of the field of Numerical Linear Algebra in that it is a blending of mathematics and computer science. Specifically, new deflation strategies, which should allow large eigenvalue problems to be broken into smaller sub-problems, will be implemented and tested. If successful, a production quality code will be submitted to netlib.org for inclusion in the Linear Algebra Package (LAPACK), an open-source repository of reliable scientific computational software.