scientific article; zbMATH DE number 1174460
The main scope of the paper is to present a variant of the hybrid GMRES algorithm; this variant is obtained by substitution of the Householder procedure for the Gram-Schmidt algorithm in the orthogonalizations involved in hybrid GMRES. The analytic justification is provided by a result by Björck who showed that the residue produced by Householder's orthogonalization has spectral norm of the order of the unit rounding error, while in the Gram-Schmidt case that value appears multiplied by the condition number of the starting linearly independent set. When it comes to the computational comparison of both approaches, the author discusses two examples. The first one ``is an artificial example designed elaborately to show the greater reliability of the new implementation (sic). As for the second, it is a real life one; its analysis suggests that for certain dimensions, Householder based hybrid GMRES procedures have a lower computational cost than the one employing the Gram-Schmidt orthogonalization, while in other cases the opposite seems to hold. The author concludes that substitution of Housholder orthogonalization procedures for those employing the classical Gram-Schmidt should lead to improvement in other hybrid methods.
- Implementation of the GMRES Method Using Householder Transformations
- Implementations of the GMRES method
- A flexible hybrid GMRES algorithm
- A Hybrid GMRES Algorithm for Nonsymmetric Linear Systems
- A fast implementation for GMRES method
- scientific article; zbMATH DE number 592765
- A generalized GMRES iterative method
- Hybrid methods based on LCG and GMRES
- Generalizations and modifications of the GMRES iterative method
- A new computational GMRES method
- A flexible hybrid GMRES algorithm
- Hybrid methods based on LCG and GMRES
- Implementation of the GMRES Method Using Householder Transformations
- scientific article; zbMATH DE number 1500400 (Why is no real title available?)
- The use of the CADNA library for validating the numerical results of the hybrid GMRES algorithm
This page was built for publication:
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4397886)