Treatment of near-breakdown in the CGS algorithm
The authors consider the system of linear algebraic equations (1) \(Ax = b\) with nonsingular matrix \(A\). They define the linear functional \(c\) on the space of polynomials by the relation \[ c(\xi^ i) = (y, A^ i r_ 0), \] where \(y\) is a given nonzero vector, \(r_ 0 = b - Ax_ 0\), and \(x_ 0\) is an initial approximation, by using Lanczos' method for solving (1). The residue \(r_ k = b - Ax_ k = P_ k(A)r_ 0\) and Lanczos' method imply orthogonality relations for the polynomials \(P_ k\), \(k = 0,1,\dots\) with respect to \(c\). The behaviour and form of the \(P_ k\) are studied in detail. The coefficients appearing in the recurrence relations for \(P_ k\) are given as ratios of scalar products. When a denominator is close to zero then a ``near-breakdown situation occurs. The stopping of the algorithm can be avoided by jumping over non-existing polynomials. In this sense the conjugate gradient squared (CGS) algorithm, i.e. \(r_ k = P^ 2_ k(A)r_ 0\), is investigated and a pseudocode of the corresponding BSMRZS algorithm (modification of the method of recursive zoom algorithm) and informations about this code in FORTRAN 77 are presented. Five illustrative examples complete this paper.
- A Completed Theory of the Unsymmetric Lanczos Process and Related Algorithms, Part I
- A composite step conjugate gradients squared algorithm for solving nonsymmetric linear systems
- A Look-Ahead Lanczos Algorithm for Unsymmetric Matrices
- A new presentation of orthogonal polynomials with applications to their computation
- A Theoretical Comparison of the Arnoldi and GMRES Algorithms
- Addendum to ``Avoiding breakdown and near-breakdown in Lanczos type algorithms
- An analysis of the composite step biconjugate gradient method
- An Implementation of the Look-Ahead Lanczos Algorithm for Non-Hermitian Matrices
- Avoiding breakdown and near-breakdown in Lanczos type algorithms
- Avoiding breakdown in the CGS algorithm
- Bi-CGSTAB: A Fast and Smoothly Converging Variant of Bi-CG for the Solution of Nonsymmetric Linear Systems
- Breakdowns in the implementation of the Lánczos method for solving linear systems
- CGS, A Fast Lanczos-Type Solver for Nonsymmetric Linear systems
- Extrapolation methods theory and practice
- Generalized conjugate-gradient acceleration of nonsymmetrizable iterative methods
- How Fast are Nonsymmetric Matrix Iterations?
- scientific article; zbMATH DE number 440776 (Why is no real title available?)
- scientific article; zbMATH DE number 176389 (Why is no real title available?)
- scientific article; zbMATH DE number 481965 (Why is no real title available?)
- scientific article; zbMATH DE number 671989 (Why is no real title available?)
- scientific article; zbMATH DE number 221694 (Why is no real title available?)
- Hybrid procedures for solving linear systems
- Lanczos-type algorithms for solving systems of linear equations
- Look-ahead in Bi-CGSTAB and other product methods for linear systems
- On the simplification of generalized conjugate-gradient methods for nonsymmetrizable linear systems
- Padé-type approximation and general orthogonal polynomials
- Polynômes orthogonaux formels - applications
- Residual Smoothing Techniques for Iterative Methods
- Review on stochastic approach to round-off error analysis and its applications
- Treatment of near-breakdown in the CGS algorithm
- Variants of BICGSTAB for Matrices with Complex Spectrum
- Avoiding breakdown in the CGS algorithm
- A composite step conjugate gradients squared algorithm for solving nonsymmetric linear systems
- Treatment of near-breakdown in the CGS algorithm
- On vector Hankel determinants
- A review of formal orthogonality in Lanczos-based methods
- Lanczos-type algorithms for solving systems of linear equations
- Look-ahead in Bi-CGSTAB and other product methods for linear systems
- The methods of Vorobyev and Lanczos
- A look-ahead strategy for the implementation of some old and new extrapolation methods
- Breakdown and near-breakdown control in the CGS algorithm using stochastic arithmetic
- An overview of approaches for the stable computation of hybrid BiCG methods
- Extrapolation algorithms and Padé approximations: A historical survey
- Derivation of BiCG from the conditions defining Lanczos' method for solving a system of linear equations
- na5
- A stabilization of the CGS method by avoiding near-breakdown
- Lanczos Methods for the Solution of Nonsymmetric Systems of Linear Equations
- Breakdowns in the implementation of the Lánczos method for solving linear systems
- A breakdown-free Lanczos type algorithm for solving linear systems
This page was built for publication: Treatment of near-breakdown in the CGS algorithm
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1334209)
