Methods for overcoming breakdown problems in the Unsymmetric Lanczos Reduction method
Linear systems of the form \[ My'+ Ky= f(t), \] where \(M\) is symmetric and positive definite can be solved by a Krylov subspace approach in which a much smaller problem of the form \[ \widehat Mz'+\widehat Kz=\widehat f(t) \] is solved. In this paper, the unsymmetric Lanczos reduction method is used using a two-sided biorthogonalization process so that \(\widehat M\) is tridiagonal and \(\widehat K\) the identity matrix. In order to avoid difficulties with breakdown when a near zero pivot is produced, a maximum pivot new-start vector technique is developed. Some numerical results show that there can be up to a 95\% computational saving compared with a standard approach. However, the reduced system can become unstable and lose accuracy during time stepping even though the original problem is stable.
- A linear system solver based on a modified Krylov subspace method for breakdown recovery
- Eigenvalue translation method for stabilizing an unsymmetric Lanczos reduction process
- A unified approach to some strategies for the treatment of breakdown in Lanczos-type algorithms
- Breakdowns in the implementation of the Lánczos method for solving linear systems
This page was built for publication: Methods for overcoming breakdown problems in the Unsymmetric Lanczos Reduction method
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4216630)