论文信息 - Stopping Criteria for Eigensolvers

Stopping Criteria for Eigensolvers

In most iterative methods for solving linear systems, the stopping criterion is based on the backward error. In this paper, after having recalled the situation for linear systems and showed that backward analysis can be used on eigenproblems, we present similar stopping criteria for eigensolvers. We also present link between the backward error and the forward error using the condition number. Numerical experiments with symmetric eigensolvers (Jacobi and Lanczos method) and with nonsymmetric ones (QR algorithm, subspace iteration and Arnoldi-Tchebychee method) illustrate the choice of the backward error.

Thierry BraconnieryCERFACS

[1] A. Deif,et al. A relative backward perturbation theorem for the eigenvalue problem , 1989 .

[2] A. J. Geurts,et al. A contribution to the theory of condition , 1982 .

[3] Iain S. Duff,et al. Stopping Criteria for Iterative Solvers , 1992, SIAM J. Matrix Anal. Appl..

[4] F. Chatelin. Valeurs propres de matrices , 1988 .

[5] Françoise Chatelin. The Influence of Nonnormality on Matrix Computations , 1993 .

[6] L. Trefethen. Approximation theory and numerical linear algebra , 1990 .

[7] Jack Dongarra,et al. LINPACK Users' Guide , 1987 .

[8] W. Arnoldi. The principle of minimized iterations in the solution of the matrix eigenvalue problem , 1951 .

[9] Nicholas J. Higham,et al. Experience with a Matrix Norm Estimator , 1990, SIAM J. Sci. Comput..

[10] Jennifer A. Scott,et al. An Arnoldi code for computing selected eigenvalues of sparse, real, unsymmetric matrices , 1995, TOMS.

[11] B. Parlett. The Symmetric Eigenvalue Problem , 1981 .

[12] Richard Barrett,et al. Templates for the Solution of Linear Systems: Building Blocks for Iterative Methods , 1994, Other Titles in Applied Mathematics.