Abstract
The paper gives a characterization of all linear systems such that when restarted GMRES is applied, prescribed admissible residual norms and (harmonic) Ritz values for all iterations inside the individual cycles are generated. Additionally, the system matrices can have any nonzero eigenvalues. The total number of GMRES iterations inside all cycles considered is assumed to be smaller than the system size. It is shown that stagnation at the end of a restart cycle must be mirrored at the beginning of the next cycle and that this is the only restriction for prescribed residual norms of restarted GMRES. The relation between prescribed residual norms of restarted GMRES and those of the corresponding full GMRES process is studied and linear systems are given where full and restarted GMRES give the same convergence history.
Similar content being viewed by others
References
Arioli, M., Pták, V., Strakoš, Z.: Krylov sequences of maximal length and convergence of GMRES. BIT 38(4), 636–643 (1998)
Brown, P.N.: A theoretical comparison of the Arnoldi and GMRES algorithms. SIAM J. Sci. Statist. Comput. 12(1), 58–78 (1991)
Chapman, A., Saad, Y.: Deflated and augmented Krylov subspace techniques. Numer. Linear Algebra Appl. 4(1), 43–66 (1997)
De Sturler, E.: Truncation strategies for optimal Krylov subspace methods. SIAM J. Numer. Anal. 36(3), 864–889 (1999)
Du, K., Duintjer Tebbens, J., Meurant, G.: Any admissible harmonic Ritz value set is possible for GMRES. Electron. Trans. Numer. Anal. 47(SI), 37–56 (2017)
Duintjer Tebbens, J., Meurant, G.: Any Ritz value behavior is possible for Arnoldi and for GMRES. SIAM J. Matrix Anal. Appl. 33(3), 958–978 (2012)
Duintjer Tebbens, J., Meurant, G.: Prescribing the behavior of early terminating GMRES and Arnoldi iterations. Num. Algor. 65(1), 69–90 (2014)
Duintjer Tebbens, J., Meurant, G.: On the convergence of Q-OR and Q-MR Krylov methods for solving nonsymmetric linear systems. BIT 56(1), 77–97 (2016)
Duintjer Tebbens, J., Meurant, G., Sadok, H., Strakoš, Z.: On investigating GMRES convergence using unitary matrices. Lin. Alg. Appl. 450, 83–107 (2014)
Eiermann, M., Ernst, O.G.: Geometric aspects of the theory of Krylov subspace methods. Acta Numer. 10, 251–312 (2001)
Eiermann, M., Ernst, O.G., Schneider, O.: Analysis of acceleration strategies for restarted minimal residual methods. J. Comput. Appl. Math. 123(1–2), 261–292 (2000)
Elman, H.: Iterative Methods for Large Sparse Nonsymmetric Systems of Linear Equations. PhD thesis. Department of Computer Science. Yale University, New Haven (1982)
Embree, M.: The tortoise and the hare restart GMRES. SIAM Rev. 45(2), 259–266 (2003)
Faber, V., Liesen, J., Tichý, P.: The Faber-Manteuffel theorem for linear operators. SIAM J. Numer. Anal. 46(3), 1323–1337 (2008)
Faber, V., Manteuffel, T.: Necessary and sufficient conditions for the existence of a conjugate gradient method. SIAM J. Numer. Anal. 21(2), 352–362 (1984)
Freund, R.W., Nachtigal, N.M.: QMR: A quasi-minimal residual method for non-Hermitian linear systems. Numer. Math. 60(3), 315–339 (1991)
Gaul, A., Gutknecht, M.H., Liesen, J., Nabben, R.: A framework for deflated and augmented Krylov subspace methods. SIAM J. Matrix Anal. Appl. 34 (2), 495–518 (2013)
Greenbaum, A., Pták, V., Strakoš, Z.: Any nonincreasing convergence curve is possible for GMRES. SIAM J. Matrix Anal. Appl. 17(3), 465–469 (1996)
Greenbaum, A., Strakoš, Z.: Matrices that generate the same Krylov residual spaces. In: Recent Advances in Iterative Methods, IMA Vol. Math. Appl., vol. 60, pp 95–118. Springer, New York (1994)
Liesen, J., Strakoš, Z.: GMRES convergence analysis for a convection-diffusion model problem. SIAM J. Sci. Comput. 26(6), 1989–2009 (2005). (electronic)
Liesen, J., Strakoš, Z.: Krylov Subspace Methods, Principles and Analysis. Numerical Mathematics and Scientific Computation. Oxford University Press, Oxford (2013)
Meurant, G.: Necessary and sufficient conditions for GMRES complete and partial stagnation. Submitted
Meurant, G., Duintjer Tebbens, J.: The role eigenvalues play in forming GMRES residual norms with non-normal matrices. Numer. Algor. 68, 143–165 (2015)
Morgan, R.B.: A restarted GMRES method augmented with eigenvectors. SIAM J. Matrix Anal. Appl. 16(4), 1154–1171 (1995)
Morgan, R.B.: Implicitly restarted GMRES and Arnoldi methods for nonsymmetric systems of equations. SIAM J. Matrix Anal. Appl. 21(4), 1112–1135 (2000)
Morgan, R.B.: GMRES with deflated restarting. SIAM J. Sci. Comput. 24(1), 20–37 (2002)
Nachtigal, N.M., Reichel, L., Trefethen, L.N.: A hybrid GMRES algorithm for nonsymmetric linear systems. SIAM J. Matrix Anal. Appl. 13(3), 796–825 (1992)
Parlett, B., Strang, G.: Matrices with prescribed Ritz values. Linear Algebra Appl. 428(7), 1725–1739 (2008)
Saad, Y.: Analysis of augmented Krylov subspace methods. SIAM J. Matrix Anal. Appl. 18(2), 435–449 (1997)
Saad, Y., Schultz, M.H.: GMRES: A generalized minimal residual algorithm for solving nonsymmetric linear systems. SIAM J. Sci. Statist. Comput. 7(3), 856–869 (1986)
Schweitzer, M.: Any finite convergence curve is possible in the initial iterations of restarted FOM. Electron. Trans. Numer. Anal. 45, 133–145 (2016)
Simoncini, V.: On the convergence of restarted Krylov subspace methods. SIAM J. Matrix Anal. Appl. 22(2), 430–452 (2000)
Simoncini, V., Szyld, D.B.: New conditions for non-stagnation of minimal residual methods. Numer. Math. 109(3), 477–487 (2008)
de Surler, E.: Personal communication (2013)
Vecharynski, E., Langou, J.: The cycle-convergence of restarted GMRES for normal matrices is sublinear. SIAM J. Sci. Comput. 32(1), 186–196 (2010)
Vecharynski, E., Langou, J.: Any admissible cycle-convergence behavior is possible for restarted GMRES at its initial cycles. Num. Lin. Algebr. Appl. 18, 499–511 (2011)
Zavorin, I., O’Leary, D.P., Elman, H.: Complete stagnation of GMRES. Linear Algebra Appl. 367, 165–183 (2003)
Zhong, B., Morgan, R.B.: Complementary cycles of restarted GMRES. Numer. Linear Algebra Appl. 15(6), 559–571 (2008)
Zítko, J.: Generalization of convergence conditions for a restarted GMRES. Numer. Linear Algebra Appl. 7(3), 117–131 (2000)
Zítko, J.: Some remarks on the restarted and augmented GMRES method. Electron. Trans. Numer. Anal. 31, 221–227 (2008)
Acknowledgments
We thank Eric de Sturler for a discussion that has stimulated the search for some of the results in Section 4.
Funding
The work of J. Duintjer Tebbens was supported by the long-term strategic development financing of the Institute of Computer Science (RVO: 67985807) of the Czech Academy of Sciences.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
Duintjer Tebbens, J., Meurant, G. On the residual norms, the Ritz values and the harmonic Ritz values that can be generated by restarted GMRES. Numer Algor 84, 1329–1352 (2020). https://doi.org/10.1007/s11075-019-00846-z
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11075-019-00846-z