Fundamental limits for rank-one matrix estimation with groupwise heteroskedasticity

TL;DR

This paper establishes the fundamental limits of rank-one matrix estimation under groupwise heteroskedastic noise, providing asymptotic formulas for the minimum mean-squared error and analyzing the optimality of PCA-based methods.

Contribution

It introduces a novel reduction from homogeneous to heteroskedastic noise models and derives exact asymptotic formulas for estimation error in high-dimensional settings.

Findings

Weighted PCA can be optimal in some settings but sub-optimal in others.

Numerical results validate the asymptotic formulas against various algorithms.

The reduction simplifies analysis of heteroskedastic noise in matrix estimation.

Abstract

Low-rank matrix recovery problems involving high-dimensional and heterogeneous data appear in applications throughout statistics and machine learning. The contribution of this paper is to establish the fundamental limits of recovery for a broad class of these problems. In particular, we study the problem of estimating a rank-one matrix from Gaussian observations where different blocks of the matrix are observed under different noise levels. In the setting where the number of blocks is fixed while the number of variables tends to infinity, we prove asymptotically exact formulas for the minimum mean-squared error in estimating both the matrix and underlying factors. These results are based on a novel reduction from the low-rank matrix tensor product model (with homogeneous noise) to a rank-one model with heteroskedastic noise. As an application of our main result, we show that recently proposed methods based on applying principal component analysis (PCA) to weighted combinations of the data are optimal in some settings but sub-optimal in others. We also provide numerical results comparing our asymptotic formulas with the performance of methods based on weighted PCA, gradient descent, and approximate message passing.