We consider the GMRES(s), i.e. the restarted GMRES with restart s for the solution of linear systems Ax = b with complex coefficient matrices. It is well known that the GMRES(s) applied on a real system is convergent if the symmetric part of the matrix A is positive definite. This paper introduces s
A new variant of restarted GMRES
β Scribed by V. Simoncini
- Publisher
- John Wiley and Sons
- Year
- 1999
- Tongue
- English
- Weight
- 264 KB
- Volume
- 6
- Category
- Article
- ISSN
- 1070-5325
No coin nor oath required. For personal study only.
β¦ Synopsis
GMRES is an attractive iterative method for solving large non-symmetric algebraic linear systems. Computational and storage constraints usually force the method to be restarted after a fixed (small) number of iterations with subsequent loss of monotonic convergence properties. Trouble may be caused by the presence of eigenvalues close to the origin which are not well detected by early restarts. Unlike recent techniques that propose to include this information in later restarts, we use a new formulation of GMRES to derive a simple variant of the algorithm. The new approach attempts to mitigate stagnation by exploiting the smoothness of a certain polynomial near zero, resorting to the original method once convergence becomes truly monotonic.
π SIMILAR VOLUMES
In the paper, GMRES is applied to various large matrices appearing in 3D field analyses. By changing the condition of the matrices, effective preconditioning of the matrix is sought. Also studied is a relationship between the residual norm and the precision evaluated during GMRES. The paper shows th
GMRES(k) is widely used for solving nonsymmetric linear systems. However, it is inadequate either when it converges only for k close to the problem size or when numerical error in the modified Gram-Schmidt process used in the GMRES orthogonalization phase dramatically affects the algorithm performan