Tail bounds for all eigenvalues of a sum of random matrices

TL;DR

This paper develops a new minimax Laplace transform method to derive tight eigenvalue bounds for sums of random matrices, enabling improved spectral analysis in matrix sparsification and covariance estimation.

Contribution

It introduces the minimax Laplace transform technique for eigenvalue bounds, extending classical tail bounds to individual eigenvalues of random matrices.

Findings

Derived eigenvalue Chernoff, Bennett, Bernstein bounds.

Analyzed spectral effects of column sparsification.

Established sample complexity for eigenvalue accuracy in covariance matrices.

Abstract

This work introduces the minimax Laplace transform method, a modification of the cumulant-based matrix Laplace transform method developed in "User-friendly tail bounds for sums of random matrices" (arXiv:1004.4389v6) that yields both upper and lower bounds on each eigenvalue of a sum of random self-adjoint matrices. This machinery is used to derive eigenvalue analogues of the classical Chernoff, Bennett, and Bernstein bounds. Two examples demonstrate the efficacy of the minimax Laplace transform. The first concerns the effects of column sparsification on the spectrum of a matrix with orthonormal rows. Here, the behavior of the singular values can be described in terms of coherence-like quantities. The second example addresses the question of relative accuracy in the estimation of eigenvalues of the covariance matrix of a random process. Standard results on the convergence of sample covariance matrices provide bounds on the number of samples needed to obtain relative accuracy in the spectral norm, but these results only guarantee relative accuracy in the estimate of the maximum eigenvalue. The minimax Laplace transform argument establishes that if the lowest eigenvalues decay sufficiently fast, on the order of (K^2*r*log(p))/eps^2 samples, where K is the condition number of an optimal rank-r approximation to C, are sufficient to ensure that the dominant r eigenvalues of the covariance matrix of a N(0, C) random vector are estimated to within a factor of 1+-eps with high probability.