English

Randomized Kaczmarz converges along small singular vectors

Numerical Analysis 2021-09-15 v2 Numerical Analysis Functional Analysis Optimization and Control

Abstract

Randomized Kaczmarz is a simple iterative method for finding solutions of linear systems Ax=bAx = b. We point out that the arising sequence (xk)k=1(x_k)_{k=1}^{\infty} tends to converge to the solution xx in an interesting way: generically, as kk \rightarrow \infty, xkxx_k - x tends to the singular vector of AA corresponding to the smallest singular value. This has interesting consequences: in particular, the error analysis of Strohmer \& Vershynin is optimal. It also quantifies the `pre-convergence' phenomenon where the method initially seems to converge faster. This fact also allows for a fast computation of vectors xx for which the Rayleigh quotient Ax/x\|Ax\|/\|x\| is small: solve Ax=0Ax = 0 via Randomized Kaczmarz.

Keywords

Cite

@article{arxiv.2006.16978,
  title  = {Randomized Kaczmarz converges along small singular vectors},
  author = {Stefan Steinerberger},
  journal= {arXiv preprint arXiv:2006.16978},
  year   = {2021}
}
R2 v1 2026-06-23T16:44:43.151Z