54
20
v1v2 (latest)

Optimal Spectral Recovery of a Planted Vector in a Subspace

Abstract

Recovering a planted vector vv in an nn-dimensional random subspace of RN\mathbb{R}^N is a generic task related to many problems in machine learning and statistics, such as dictionary learning, subspace recovery, principal component analysis, and non-Gaussian component analysis. In this work, we study computationally efficient estimation and detection of a planted vector vv whose 4\ell_4 norm differs from that of a Gaussian vector with the same 2\ell_2 norm. For instance, in the special case where vv is an NρN \rho-sparse vector with Bernoulli-Gaussian or Bernoulli-Rademacher entries, our results include the following: (1) We give an improved analysis of a slight variant of the spectral method proposed by Hopkins, Schramm, Shi, and Steurer (2016), showing that it approximately recovers vv with high probability in the regime nρNn \rho \ll \sqrt{N}. This condition subsumes the conditions ρ1/n\rho \ll 1/\sqrt{n} or nρNn \sqrt{\rho} \lesssim \sqrt{N} required by previous work up to polylogarithmic factors. We achieve \ell_\infty error bounds for the spectral estimator via a leave-one-out analysis, from which it follows that a simple thresholding procedure exactly recovers vv with Bernoulli-Rademacher entries, even in the dense case ρ=1\rho = 1. (2) We study the associated detection problem and show that in the regime nρNn \rho \gg \sqrt{N}, any spectral method from a large class (and more generally, any low-degree polynomial of the input) fails to detect the planted vector. This matches the condition for recovery and offers evidence that no polynomial-time algorithm can succeed in recovering a Bernoulli-Gaussian vector vv when nρNn \rho \gg \sqrt{N}.

View on arXiv
Comments on this paper