A Robust Preconditioner with Low Memory Requirements for Large Sparse Least Squares Problems

0404727 - UIVT-O 20030115 RIV US eng J - Journal Article
Benzi, M. - Tůma, Miroslav
A Robust Preconditioner with Low Memory Requirements for Large Sparse Least Squares Problems.
SIAM Journal on Scientific Computing. Roč. 25, č. 2 (2003), s. 499-512. ISSN 1064-8275. E-ISSN 1095-7197
R&D Projects: GA AV ČR IAA1030103; GA AV ČR IAA2030801
Institutional research plan: AV0Z1030915
Keywords : large sparse least squares problems * preconditioned CGLS * robust incomplete factorization * incomplete C-orthogonalization * incomplete QR
Subject RIV: BA - General Mathematics
Impact factor: 1.379, year: 2003

This paper describes a technique for constructing robust preconditioners for the CGLS method applied to the solution of large and sparse least squares problems. The algorithm computes an incomplete LDLT factorization of the normal equations matrix without the need to form the normal matrix itself. The preconditioner is reliable 9 pivot breakdowns cannot occur0 and has low intermediate storage requirements. Numerical experiments illustrating the performance of the preconditioner are presented. A comparison with incomplete QR preconditioners is also included.
Permanent Link: http://hdl.handle.net/11104/0124965