Japan Journal of Industrial and Applied Mathematics

Preconditioned GMRES Methods for Least Squares Problems

Tokushi Ito and Ken Hayami

Full-text: Open access

Abstract

For least squares problems of minimizing $\|\boldsymbol{b}-A\boldsymbol{x}\|_2$ where $A$ is a large sparse $m\times n$ ($m\ge n$) matrix, the common method is to apply the conjugate gradient method to the normal equation $A^{\mathrm{T}} A\boldsymbol{x}=A^{\mathrm{T}} \boldsymbol{b}$. However, the condition number of $A^{\mathrm{T}} A$ is square of that of $A$, and convergence becomes problematic for severely ill-conditioned problems even with preconditioning. In this paper, we propose two methods for applying the GMRES method to the least squares problem by using an $n\times m$ matrix $B$. We give the necessary and sufficient condition that $B$ should satisfy in order that the proposed methods give a least squares solution. Then, for implementations for $B$, we propose an incomplete QR decomposition IMGS($l$). Numerical experiments showed that the simplest case $l=0$ gives the best results, and converges faster than previous methods for severely ill-conditioned problems. The preconditioner IMGS(0) is equivalent to the case $B=(\diag (A^{\mathrm{T}} A))^{-1} A^{\mathrm{T}}$, so $(\diag (A^{\mathrm{T}} A))^{-1} A^{\mathrm{T}}$ was the best preconditioner among IMGS($l$) and Jennings' IMGS($\tau$). On the other hand, CG-IMGS(0) was the fastest for well-conditioned problems.

Article information

Source
Japan J. Indust. Appl. Math., Volume 25, Number 2 (2008), 185-207.

Dates
First available in Project Euclid: 3 July 2008

Permanent link to this document
https://projecteuclid.org/euclid.jjiam/1215118762

Mathematical Reviews number (MathSciNet)
MR2431679

Zentralblatt MATH identifier
1154.65022

Keywords
least squares problems GMRES preconditioning incomplete QR decomposition singular systems

Citation

Ito, Tokushi; Hayami, Ken. Preconditioned GMRES Methods for Least Squares Problems. Japan J. Indust. Appl. Math. 25 (2008), no. 2, 185--207. https://projecteuclid.org/euclid.jjiam/1215118762


Export citation