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. 2312.08893
47
8

Solving Dense Linear Systems Faster Than via Preconditioning

14 December 2023
Michal Dereziñski
Jiaming Yang
ArXivPDFHTML
Abstract

We give a stochastic optimization algorithm that solves a dense n×nn\times nn×n real-valued linear system Ax=bAx=bAx=b, returning x~\tilde xx~ such that ∥Ax~−b∥≤ϵ∥b∥\|A\tilde x-b\|\leq \epsilon\|b\|∥Ax~−b∥≤ϵ∥b∥ in time: \tilde O((n^2+nk^{\omega-1})\log1/\epsilon), where kkk is the number of singular values of AAA larger than O(1)O(1)O(1) times its smallest positive singular value, ω<2.372\omega < 2.372ω<2.372 is the matrix multiplication exponent, and O~\tilde OO~ hides a poly-logarithmic in nnn factor. When k=O(n1−θ)k=O(n^{1-\theta})k=O(n1−θ) (namely, AAA has a flat-tailed spectrum, e.g., due to noisy data or regularization), this improves on both the cost of solving the system directly, as well as on the cost of preconditioning an iterative method such as conjugate gradient. In particular, our algorithm has an O~(n2)\tilde O(n^2)O~(n2) runtime when k=O(n0.729)k=O(n^{0.729})k=O(n0.729). We further adapt this result to sparse positive semidefinite matrices and least squares regression. Our main algorithm can be viewed as a randomized block coordinate descent method, where the key challenge is simultaneously ensuring good convergence and fast per-iteration time. In our analysis, we use theory of majorization for elementary symmetric polynomials to establish a sharp convergence guarantee when coordinate blocks are sampled using a determinantal point process. We then use a Markov chain coupling argument to show that similar convergence can be attained with a cheaper sampling scheme, and accelerate the block coordinate descent update via matrix sketching.

View on arXiv
Comments on this paper