Approximate spectral gaps for Markov chains mixing times in high dimensions

TL;DR

This paper introduces the concept of approximate spectral gaps to analyze the mixing times of high-dimensional Markov Chain Monte Carlo algorithms, especially when traditional spectral gaps are degenerate, with applications to Bayesian variable selection.

Contribution

It proposes a new framework of approximate spectral gaps for analyzing MCMC mixing times in high dimensions, applicable to complex models like linear regression.

Findings

Mixing time is polynomial in dimension under regularity conditions.

Applicable to Gibbs samplers for variable selection.

Provides a new tool for high-dimensional MCMC analysis.

Abstract

This paper introduces a concept of approximate spectral gap to analyze the mixing time of Markov Chain Monte Carlo (MCMC) algorithms for which the usual spectral gap is degenerate or almost degenerate. We use the idea to analyze a class of MCMC algorithms to sample from mixtures of densities. As an application we study the mixing time of a Gibbs sampler for variable selection in linear regression models. Under some regularity conditions on the signal and the design matrix of the regression problem, we show that for well-chosen initial distributions the mixing time of the Gibbs sampler is polynomial in the dimension of the space.