6
27

The Overlap Gap Property in Principal Submatrix Recovery

Abstract

We study support recovery for a k×kk \times k principal submatrix with elevated mean λ/N\lambda/N, hidden in an N×NN\times N symmetric mean zero Gaussian matrix. Here λ>0\lambda>0 is a universal constant, and we assume k=Nρk = N \rho for some constant ρ(0,1)\rho \in (0,1). We establish that {there exists a constant C>0C>0 such that} the MLE recovers a constant proportion of the hidden submatrix if λC1ρlog1ρ\lambda {\geq C} \sqrt{\frac{1}{\rho} \log \frac{1}{\rho}}, {while such recovery is information theoretically impossible if λ=o(1ρlog1ρ)\lambda = o( \sqrt{\frac{1}{\rho} \log \frac{1}{\rho}} )}. The MLE is computationally intractable in general, and in fact, for ρ>0\rho>0 sufficiently small, this problem is conjectured to exhibit a \emph{statistical-computational gap}. To provide rigorous evidence for this, we study the likelihood landscape for this problem, and establish that for some ε>0\varepsilon>0 and 1ρlog1ρλ1ρ1/2+ε\sqrt{\frac{1}{\rho} \log \frac{1}{\rho} } \ll \lambda \ll \frac{1}{\rho^{1/2 + \varepsilon}}, the problem exhibits a variant of the \emph{Overlap-Gap-Property (OGP)}. As a direct consequence, we establish that a family of local MCMC based algorithms do not achieve optimal recovery. Finally, we establish that for λ>1/ρ\lambda > 1/\rho, a simple spectral method recovers a constant proportion of the hidden submatrix.

View on arXiv
Comments on this paper