Error bounds for Metropolis-Hastings algorithms applied to perturbations of Gaussian measures in high dimensions (Q2443195): Difference between revisions

Metropolis-Hastings algorithms for approximate sampling from continuous distributions on high-dimensional state spaces \(\mathbb{R}^d\) are considered. Distributions are assumed to have sufficiently regular densities w.r.t. a Gaussian measure on \(\mathbb{R}^d\). In this case, under mild conditions upper bounds for the contractivity rate in Kantorovich-Rubinstein-Wasserstein distance for sufficiently small step size \(h\) do not depend on dimension size \(d\). As for transition kernels, the approach applies to Metropolis-Hastings chains with Ornstein-Uhlenbeck proposals and to Metropolis-adjusted Langevin algorithm.