Analysis of GMRES for Low‐Rank and Small‐Norm Perturbations of the Identity Matrix
Analysis of GMRES for Low‐Rank and Small‐Norm Perturbations of the Identity Matrix
复制标题
单位矩阵低秩和小范数扰动的 GMRES 分析
DOI:
10.1002/pamm.202200267
复制
发表时间:
2023
期刊:
影响因子:
--
通讯作者:
Embree, Mark
中科院分区:
文献类型:
--
作者:
Carr, Arielle K.;de Sturler, Eric;Embree, Mark
In many applications, linear systems arise where the coefficient matrix takes the special formI+K+E, whereIis the identity matrix of dimensionn, rank(K) =p≪n, and ∥E∥ ≤ ϵ < 1. GMRES convergence rates for linear systems with coefficient matrices of the formsI+KandI+Eare guaranteed by well‐known theory, but only relatively weak convergence bounds specific to matrices of the formI+K+Ecurrently exist. In this paper, we explore the convergence properties of linear systems with such coefficient matrices by considering the pseudospectrum ofI+K. We derive a bound for the GMRES residual in terms of ϵ when approximately solving the linear system (I+K+E)x=band identify the eigenvalues ofI+Kthat are sensitive to perturbation. In particular, while a clustered spectrum away from the origin is often a good indicator of fast GMRES convergence, that convergence may be slow when some of those eigenvalues are ill‐conditioned. We show there can be at most 2peigenvalues ofI+Kthat are sensitive to small perturbations. We present numerical results when using GMRES to solve a sequence of linear systems of the form (I+Kj+Ej)xj=bjthat arise from the application of Broyden's method to solve a nonlinear partial differential equation.
登录
查看更多内容
DOI:
10.1137/120884328
发表时间:
2013-07
期刊:
SIAM J. Matrix Anal. Appl.
影响因子:
--
作者:
Josef A. Sifuentes;M. Embree;R. Morgan
通讯作者:
Josef A. Sifuentes;M. Embree;R. Morgan
DOI:
--
发表时间:
2009
期刊:
Large-Scale Scientific Computing
影响因子:
--
作者:
N. Gmati;B. Philippe
通讯作者:
B. Philippe
DOI:
10.1137/100805467
发表时间:
2010
期刊:
SIAM J. Sci. Comput.
影响因子:
--
作者:
K. Ahuja;B. Clark;E. D. Sturler;D. Ceperley;Jeongnim Kim
通讯作者:
Jeongnim Kim
DOI:
--
发表时间:
2004
期刊:
影响因子:
--
作者:
C. Kelley;I. Kevrekidis;L. Qiao
通讯作者:
L. Qiao