Deflated Lanczos Procedures for Solving Nearly Singular Systems.
Abstract
This document considers the solution of the linear sytem Ax = b, where A is nearly singular. The solution x can be uniquely decomposed into two parts: a generally large component in the direction of the approximate null space of A, and a part that is orthogonal to it. In many applications, it is desirable to compute this deflated decomposition in a stable and efficient manner. This apper proposes an iterative algorithm based on the Lanczos process, for the case where A is symmetric. The method requires access to A only in the form of a matrix-vector product Av and is efficient for large problems. No a priori knowledge of the approximate null space is needed. Keywords: Orthogonality; Convergence; Matrices(Mathematics); and Numerical analysis. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1985
- Accession Number
- ADA161971
Entities
People
- Tony F. Chan
- Youcef Saad
Organizations
- Yale University