English

On computing approximate Lewis weights

Data Structures and Algorithms 2024-04-04 v1 Optimization and Control

Abstract

In this note we provide and analyze a simple method that given an n×dn \times d matrix, outputs approximate p\ell_p-Lewis weights, a natural measure of the importance of the rows with respect to the p\ell_p norm, for p2p \geq 2. More precisely, we provide a simple post-processing procedure that turns natural one-sided approximate p\ell_p-Lewis weights into two-sided approximations. When combined with a simple one-sided approximation algorithm presented by Lee (PhD thesis, `16) this yields an algorithm for computing two-sided approximations of the p\ell_p-Lewis weights of an n×dn \times d-matrix using poly(d,p)\mathrm{poly}(d,p) approximate leverage score computations. While efficient high-accuracy algorithms for approximating p\ell_p-Lewis had been established previously by Fazel, Lee, Padmanabhan and Sidford (SODA `22), the simple structure and approximation tolerance of our algorithm may make it of use for different applications.

Cite

@article{arxiv.2404.02881,
  title  = {On computing approximate Lewis weights},
  author = {Simon Apers and Sander Gribling and Aaron Sidford},
  journal= {arXiv preprint arXiv:2404.02881},
  year   = {2024}
}
R2 v1 2026-06-28T15:43:15.110Z