ResearchTrend.AI
  • Papers
  • Communities
  • Events
  • Blog
  • Pricing
Papers
Communities
Social Events
Terms and Conditions
Pricing
Parameter LabParameter LabTwitterGitHubLinkedInBlueskyYoutube

© 2025 ResearchTrend.AI, All rights reserved.

  1. Home
  2. Papers
  3. 2004.07986
19
25

Average Case Column Subset Selection for Entrywise ℓ1\ell_1ℓ1​-Norm Loss

16 April 2020
Zhao Song
David P. Woodruff
Peilin Zhong
ArXivPDFHTML
Abstract

We study the column subset selection problem with respect to the entrywise ℓ1\ell_1ℓ1​-norm loss. It is known that in the worst case, to obtain a good rank-kkk approximation to a matrix, one needs an arbitrarily large nΩ(1)n^{\Omega(1)}nΩ(1) number of columns to obtain a (1+ϵ)(1+\epsilon)(1+ϵ)-approximation to the best entrywise ℓ1\ell_1ℓ1​-norm low rank approximation of an n×nn \times nn×n matrix. Nevertheless, we show that under certain minimal and realistic distributional settings, it is possible to obtain a (1+ϵ)(1+\epsilon)(1+ϵ)-approximation with a nearly linear running time and poly(k/ϵ)+O(klog⁡n)(k/\epsilon)+O(k\log n)(k/ϵ)+O(klogn) columns. Namely, we show that if the input matrix AAA has the form A=B+EA = B + EA=B+E, where BBB is an arbitrary rank-kkk matrix, and EEE is a matrix with i.i.d. entries drawn from any distribution μ\muμ for which the (1+γ)(1+\gamma)(1+γ)-th moment exists, for an arbitrarily small constant γ>0\gamma > 0γ>0, then it is possible to obtain a (1+ϵ)(1+\epsilon)(1+ϵ)-approximate column subset selection to the entrywise ℓ1\ell_1ℓ1​-norm in nearly linear time. Conversely we show that if the first moment does not exist, then it is not possible to obtain a (1+ϵ)(1+\epsilon)(1+ϵ)-approximate subset selection algorithm even if one chooses any no(1)n^{o(1)}no(1) columns. This is the first algorithm of any kind for achieving a (1+ϵ)(1+\epsilon)(1+ϵ)-approximation for entrywise ℓ1\ell_1ℓ1​-norm loss low rank approximation.

View on arXiv
Comments on this paper