English

Backward errors for multiple eigenpairs in structured and unstructured nonlinear eigenvalue problems

Numerical Analysis 2025-02-27 v2 Numerical Analysis

Abstract

Given a nonlinear matrix-valued function F(λ)F(\lambda) and approximate eigenpairs (λi,vi)(\lambda_i, v_i), we discuss how to determine the smallest perturbation δF\delta F such that [F+δF](λi)vi=0[F + \delta F](\lambda_i) v_i = 0; we call the distance between the FF and F+δFF + \delta F the backward error for this set of approximate eigenpairs. We focus on the case where F(λ)F(\lambda) is given as a linear combination of scalar functions multiplying matrix coefficients FiF_i, and the perturbation is done on the matrix coefficients. We provide inexpensive upper bounds, and a way to accurately compute the backward error by means of direct computations or through Riemannian optimization. We also discuss how the backward error can be determined when the FiF_i have particular structures (such as symmetry, sparsity, or low-rank), and the perturbations are required to preserve them. For special cases (such as for symmetric coefficients), explicit and inexpensive formulas to compute the δFi\delta F_i are also given.

Keywords

Cite

@article{arxiv.2405.06327,
  title  = {Backward errors for multiple eigenpairs in structured and unstructured nonlinear eigenvalue problems},
  author = {Miryam Gnazzo and Leonardo Robol},
  journal= {arXiv preprint arXiv:2405.06327},
  year   = {2025}
}

Comments

26 pages, 6 figures, 1 table