Volume 36, pp. 126-148, 2009-2010.

IDR explained

Martin H. Gutknecht

Abstract

The Induced Dimension Reduction (IDR) method is a Krylov space method for solving linear systems that was developed by Peter Sonneveld around 1979. It was noticed by only a few people, and mainly as the forerunner of Bi-CGSTAB, which was introduced a decade later. In 2007, Sonneveld and van Gijzen reconsidered IDR and generalized it to IDR$(s)$, claiming that IDR$(1)$ $\approx$ IDR is equally fast but preferable to the closely related Bi-CGSTAB, and that IDR$(s)$ with $s>1$ may be much faster than Bi-CGSTAB. It also turned out that when $s > 1$, IDR$(s)$ is related to ML$(s)$BiCGSTAB of Yeung and Chan, and that there is quite some flexibility in the IDR approach. This approach differs completely from traditional approaches to Krylov space methods, and therefore it requires an extra effort to get familiar with it and to understand the connections as well as the differences to better-known Krylov space methods. This expository paper aims to provide some help in this and to make the method understandable even to non-experts. After presenting the history of IDR and related methods, we summarize some of the basic facts on Krylov space methods. Then we present the original IDR$(s)$ in detail and put it into perspective with other methods. Specifically, we analyze the differences between the IDR method published in 1980, IDR$(1)$, and Bi-CGSTAB. At the end of the paper, we discuss a recently proposed ingenious variant of IDR$(s)$ whose residuals fulfill extra orthogonality conditions. There we dwell on details that have been left out in the publications of van Gijzen and Sonneveld.

Full Text (PDF) [314 KB]

Key words

Krylov space method, iterative method, induced dimension reduction, IDR, CGS, Bi-CGSTAB, ML($k$)BiCGSTAB, large nonsymmetric linear system

AMS subject classifications

Links to the cited ETNA articles

[22]Vol. 1 (1993), pp. 11-32 Gerard L. G. Sleijpen and Diederik R. Fokkema: BiCGstab(l) for linear equations involving unsymmetric matrices with complex spectrum

< Back