On the Convergence of Restarted Krylov Subspace Methods

We are interested in the convergence analysis of restarted Krylov subspace iterative methods for the solution of large nonsymmetric linear systems. Several contributions in the literature have associated the convergence to some spectral properties of the coefficient matrix, while little work has been devoted to investigating how the singular values of A may influence the convergence. In this paper we present new relations that can be used to monitor the behavior of the restarted methods, especially GMRES, when the coefficient matrix has small (but not tiny) singular values and the right-hand side has a dominant component onto the corresponding left singular space. We also present some simple but insightful relations that highlight the dependence of the restarted schemes on new matrices; moreover, closed forms of the restarted solutions are used to relate the approximations of the unrestarted and restarted approaches.

[1]  Anne Greenbaum,et al.  Any Nonincreasing Convergence Curve is Possible for GMRES , 1996, SIAM J. Matrix Anal. Appl..

[2]  Anne Greenbaum,et al.  Iterative methods for solving linear systems , 1997, Frontiers in applied mathematics.

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

[4]  Y. Saad,et al.  GMRES: a generalized minimal residual algorithm for solving nonsymmetric linear systems , 1986 .

[5]  W. Symes,et al.  Deflated Krylov subspace methods for nearly singular linear systems , 1992 .

[6]  Ronald B. Morgan,et al.  Implicitly Restarted GMRES and Arnoldi Methods for Nonsymmetric Systems of Equations , 2000, SIAM J. Matrix Anal. Appl..

[7]  H. V. D. Vorst,et al.  The superlinear convergence behaviour of GMRES , 1993 .

[8]  Meza.  A modification to the GMRES method for ill-conditioned linear systems , 1995 .

[9]  A. Van.  SOME OBSERVATIONS ON THE CONVERGENCE BEHAVIOR OF GMRES(II) , 1990 .

[10]  Wayne Joubert,et al.  On the convergence behavior of the restarted GMRES algorithm for solving nonsymmetric linear systems , 1994, Numer. Linear Algebra Appl..

[11]  E. Sturler,et al.  Truncation Strategies for Optimal Krylov Subspace Methods , 1999 .

[12]  Yousef Saad,et al.  Deflated and Augmented Krylov Subspace Techniques , 1997, Numer. Linear Algebra Appl..

[13]  R. Freund,et al.  QMR: a quasi-minimal residual method for non-Hermitian linear systems , 1991 .

[14]  J. Navarro-Pedreño.  Numerical Methods for Least Squares Problems , 1996 .

[15]  H. Walker,et al.  GMRES On (Nearly) Singular Systems , 1997, SIAM J. Matrix Anal. Appl..

[16]  Ronald B. Morgan,et al.  A Restarted GMRES Method Augmented with Eigenvectors , 1995, SIAM J. Matrix Anal. Appl..

[17]  Valeria Simoncini,et al.  A new variant of restarted GMRES , 1997, Numer. Linear Algebra Appl..

[18]  Anne Greenbaum,et al.  Relations between Galerkin and Norm-Minimizing Iterative Methods for Solving Linear Systems , 1996, SIAM J. Matrix Anal. Appl..

[19]  Peter N. Brown,et al.  A Theoretical Comparison of the Arnoldi and GMRES Algorithms , 1991, SIAM J. Sci. Comput..

[20]  Gene H. Golub,et al.  Matrix computations , 1983 .

[21]  Ronald B. Morgan,et al.  On restarting the Arnoldi method for large nonsymmetric eigenvalue problems , 1996, Math. Comput..

[22]  Valeria Simoncini,et al.  Analysis of a Minimum Perturbation Algorithm for Nonsymmetric Linear Systems , 1997 .

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

[24]  Y. Saad.  Krylov subspace methods for solving large unsymmetric linear systems , 1981 .

[25]  Y. Saad.  Analysis of Augmented Krylov Subspace Methods , 1997, SIAM J. Matrix Anal. Appl..