On greedy randomized coordinate descent methods for computing the extreme eigenpairs of symmetric matrices
IMA Journal of Numerical Analysis · 27.12.2025
DOI: 10.1093/imanum/draf103 ↗Kurz gefasst
Fachartikel in der Zeitschrift IMA Journal of Numerical Analysis, online erschienen am 27.12.2025. Verfasst von Zhong-Zhi Bai. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 28 Literaturangaben.
Abstract (Originalsprache)
We propose a class of greedy randomized coordinate descent methods for computing the smallest eigenvalue and the corresponding eigenvector of a large-scale, real and symmetric matrix. We prove the convergence in terms of expectation for the sequence of iterates and the sequence of Rayleigh quotients, generated by these iteration methods, and show that these iteration methods can distinctly outperform the parameterized power method and the coordinate descent method in terms of both iteration counts and computing times. Moreover, with both theory and computation, we confirm that the convergence property of this class of greedy randomized coordinate descent methods can be improved with a suitable choice of the iteration parameter involved in the probability criterion.
Zitieren
APA
Bai, Z. (2025). On greedy randomized coordinate descent methods for computing the extreme eigenpairs of symmetric matrices. IMA Journal of Numerical Analysis. https://doi.org/10.1093/imanum/draf103
Vancouver
Bai Z. On greedy randomized coordinate descent methods for computing the extreme eigenpairs of symmetric matrices. IMA Journal of Numerical Analysis. 2025. doi:10.1093/imanum/draf103
BibTeX
@article{Bai2025,
author = {Bai, Zhong-Zhi},
title = {On greedy randomized coordinate descent methods for computing the extreme eigenpairs of symmetric matrices},
journal = {IMA Journal of Numerical Analysis},
year = {2025},
doi = {10.1093/imanum/draf103}
}
EUVIMED