Matrix concentration inequalities via the method of exchangeable pairs

Abstract

This paper derives exponential concentration inequalities and polynomial moment inequalities for the spectral norm of a random matrix. The analysis requires a matrix extension of the scalar concentration theory developed by Sourav Chatterjee using Stein's method of exchangeable pairs. When applied to a sum of independent random matrices, this approach yields matrix generalizations of the classical inequalities due to Hoeffding, Bernstein, Khintchine and Rosenthal. The same technique delivers bounds for sums of dependent random matrices and more general matrix-valued functions of dependent random variables.Comment: Published in at http://dx.doi.org/10.1214/13-AOP892 the Annals of Probability (http://www.imstat.org/aop/) by the Institute of Mathematical Statistics (http://www.imstat.org

An Introduction to Matrix Concentration Inequalities

“…The annotated bibliography includes references to the large literature on moment inequalities for random matrices. Recently, Lester Mackey and the author, in collaboration with Daniel Paulin and several other researchers [118,143], have developed another framework for establishing matrix concentration. This approach extends a scalar argument, introduced by Chatterjee [38,39], that depends on exchangeable pairs and Markov chain couplings.…”

Lasso Based Performance Evaluation for Sparse One-Dimensional Radar Problem Under Random Sub-Sampling and Gaussian Noise

“…Secondly, the spectrum bound for N S n=1 ς n h * Λ,n h Λ,n , the sum of independent random matrix, can be estimated according to the Bornstein's inequality [23], which states…”

“…For instance, let the bandwidth and time duration of waveform given by (2) be chosen as 150 MHz and 6 µs, respectively, then N H µ H ∼ = 14.81. Assume ρ K ≤ 1/10N H µ H = 0.6753%, then it can be calculated from (41) that C(ρ K ) takes value in (14,23.125].…”

A Stein characterisation of the generalized hyperbolic distribution

“…For a detailed account of the method see the monograph [10] or the review article [40]. Over the years, the method has been adapted to many other probability distributions, such as the Poisson [9], gamma [26,32], exponential [7,36] and Laplace distribution [14,39], and has been applied to a wide range of applications, including random matrix theory [33], random graph theory [3], urn models [13,28,38], goodness-of-fit statistics [27,26] and statistical physics [8,19]. For an overview of the current literature see [31].…”