Numerical Instabilities in the Kaczmarz Method and Stabilization by Iterative Refinement

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Dereziński, Michał, Epperly, Ethan N., Needell, Deanna, Xue, Alexander
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866916020264894464
author Dereziński, Michał
Epperly, Ethan N.
Needell, Deanna
Xue, Alexander
author_facet Dereziński, Michał
Epperly, Ethan N.
Needell, Deanna
Xue, Alexander
contents The randomized Kaczmarz method and its accelerated variants are a powerful class of iterative methods for solving large-scale linear systems, offering guaranteed convergence with low per-iteration cost. However, their numerical stability remains poorly understood. In this work, we investigate the stability properties of both classical and accelerated randomized Kaczmarz methods, with an emphasis on how error propagates across iterations and interacts with acceleration. We show that both classical and accelerated randomized Kaczmarz fail to be forward stable. To address this issue, we propose the integration of iterative refinement into randomized Kaczmarz frameworks. We demonstrate that refinement can effectively control error accumulation and recover high-accuracy solutions, even when the system is ill-conditioned. Numerical experiments corroborate our theoretical findings and illustrate the practical benefits of combining refinement with both classical and accelerated Kaczmarz methods.
format Preprint
id arxiv_https___arxiv_org_abs_2605_17185
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Numerical Instabilities in the Kaczmarz Method and Stabilization by Iterative Refinement
Dereziński, Michał
Epperly, Ethan N.
Needell, Deanna
Xue, Alexander
Numerical Analysis
65F10, 65F35, 65F20, 65G50, 68W20
The randomized Kaczmarz method and its accelerated variants are a powerful class of iterative methods for solving large-scale linear systems, offering guaranteed convergence with low per-iteration cost. However, their numerical stability remains poorly understood. In this work, we investigate the stability properties of both classical and accelerated randomized Kaczmarz methods, with an emphasis on how error propagates across iterations and interacts with acceleration. We show that both classical and accelerated randomized Kaczmarz fail to be forward stable. To address this issue, we propose the integration of iterative refinement into randomized Kaczmarz frameworks. We demonstrate that refinement can effectively control error accumulation and recover high-accuracy solutions, even when the system is ill-conditioned. Numerical experiments corroborate our theoretical findings and illustrate the practical benefits of combining refinement with both classical and accelerated Kaczmarz methods.
title Numerical Instabilities in the Kaczmarz Method and Stabilization by Iterative Refinement
topic Numerical Analysis
65F10, 65F35, 65F20, 65G50, 68W20
url https://arxiv.org/abs/2605.17185