Papers
Topics
Authors
Recent
Search
2000 character limit reached

Testing Positive Semi-Definiteness via Random Submatrices

Published 13 May 2020 in cs.DS | (2005.06441v3)

Abstract: We study the problem of testing whether a matrix A∈Rn×n\mathbf{A} \in \mathbb{R}^{n \times n} with bounded entries (∥A∥∞≤1\|\mathbf{A}\|_\infty \leq 1) is positive semi-definite (PSD), or ϵ\epsilon-far in Euclidean distance from the PSD cone, meaning that min⁡B⪰0∥A−B∥F2>ϵn2\min_{\mathbf{B} \succeq 0} \|\mathbf{A} - \mathbf{B}\|_F^2 > \epsilon n^2, where B⪰0\mathbf{B} \succeq 0 denotes that B\mathbf{B} is PSD. Our main algorithmic contribution is a non-adaptive tester which distinguishes between these cases using only O~(1/ϵ4)\tilde{O}(1/\epsilon^4) queries to the entries of A\mathbf{A}. If instead of the Euclidean norm we considered the distance in spectral norm, we obtain the "ℓ∞\ell_\infty-gap problem", where A\mathbf{A} is either PSD or satisfies min⁡B⪰0∥A−B∥2>ϵn\min_{\mathbf{B}\succeq 0} \|\mathbf{A}- \mathbf{B}\|_2 > \epsilon n. For this related problem, we give a O~(1/ϵ2)\tilde{O}(1/\epsilon^2) query tester, which we show is optimal up to log⁡(1/ϵ)\log(1/\epsilon) factors. Our testers randomly sample a collection of principal submatrices and check whether these submatrices are PSD. Consequentially, our algorithms achieve one-sided error: whenever they output that A\mathbf{A} is not PSD, they return a certificate that A\mathbf{A} has negative eigenvalues. We complement our upper bound for PSD testing with Euclidean norm distance by giving a Ω~(1/ϵ2)\tilde{\Omega}(1/\epsilon^2) lower bound for any non-adaptive algorithm. Our lower bound construction is general, and can be used to derive lower bounds for a number of spectral testing problems. As an example of the applicability of our construction, we obtain a new Ω~(1/ϵ4)\tilde{\Omega}(1/\epsilon^4) sampling lower bound for testing the Schatten-$1$ norm with a ϵn1.5\epsilon n^{1.5} gap, extending a result of Balcan, Li, Woodruff, and Zhang [SODA'19]. In addition, it yields new sampling lower bounds for estimating the Ky-Fan Norm, and the cost of the best rank-kk approximation.

Citations (14)

Summary

No one has generated a summary of this paper yet.

Paper to Video (Beta)

No one has generated a video about this paper yet.

Whiteboard

No one has generated a whiteboard explanation for this paper yet.

Open Problems

We haven't generated a list of open problems mentioned in this paper yet.

Continue Learning

We haven't generated follow-up questions for this paper yet.

YouTube

Testing Positive Semi-Definiteness via Random Submatrices 156 views