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. 1704.03371
88
58
v1v2v3 (latest)

Sublinear Time Low-Rank Approximation of Positive Semidefinite Matrices

11 April 2017
Cameron Musco
David P. Woodruff
ArXiv (abs)PDFHTML
Abstract

We show how to compute a relative-error low-rank approximation to any positive semidefinite (PSD) matrix in sublinear time, i.e., for any n×nn \times nn×n PSD matrix AAA, in O~(n⋅poly(k/ϵ))\tilde O(n \cdot poly(k/\epsilon))O~(n⋅poly(k/ϵ)) time we output a rank-kkk matrix BBB, in factored form, for which ∥A−B∥F2≤(1+ϵ)∥A−Ak∥F2\|A-B\|_F^2 \leq (1+\epsilon)\|A-A_k\|_F^2∥A−B∥F2​≤(1+ϵ)∥A−Ak​∥F2​, where AkA_kAk​ is the best rank-kkk approximation to AAA. When kkk and 1/ϵ1/\epsilon1/ϵ are not too large compared to the sparsity of AAA, our algorithm does not need to read all entries of the matrix. Hence, we significantly improve upon previous nnz(A)nnz(A)nnz(A) time algorithms based on oblivious subspace embeddings, and bypass an nnz(A)nnz(A)nnz(A) time lower bound for general matrices (where nnz(A)nnz(A)nnz(A) denotes the number of non-zero entries in the matrix). We prove time lower bounds for low-rank approximation of PSD matrices, showing that our algorithm is close to optimal. Finally, we extend our techniques to give sublinear time algorithms for low-rank approximation of AAA in the (often stronger) spectral norm metric ∥A−B∥22\|A-B\|_2^2∥A−B∥22​ and for ridge regression on PSD matrices.

View on arXiv
Comments on this paper