Group-averaged Markov chains: mixing improvement
Abstract
For Markov kernels on a general state space , we introduce a new class of averaged Markov kernels of induced by a group that acts on and a probability measure on . Notable special cases are the group-orbit average , left-average , right-average and the independent-double-average . For -stationary in which is invariant with respect to , we show that in general enjoys favorable convergence properties than based on metrics such as spectral gap or asymptotic variance, and within the family of the most preferable kernel is in general . We demonstrate that are comparable in terms of mixing times, which supports the use of in practice as computationally cheaper alternatives over . These averaged kernels also admit natural geometric interpretations: they emerge as unique projections of onto specific -invariant structures under the Kullback-Leibler divergence or the Hilbert-Schmidt norm and satisfy Pythagorean identities. On the other hand, in the general case if is not invariant with respect to , we propose and study a technique that we call state-dependent averaging of Markov kernels which generalizes the earlier results to this setting. As examples and applications, this averaging perspective not only allows us to recast state-of-the-art Markov chain samplers such as Hamiltonian Monte Carlo or piecewise-deterministic Markov processes as specific cases of , but also enables improvements to existing samplers such as Metropolis-Hastings, achieving rapid mixing in some toy models or when is the discrete uniform distribution.
Keywords
Cite
@article{arxiv.2509.02996,
title = {Group-averaged Markov chains: mixing improvement},
author = {Michael C. H. Choi and Youjia Wang},
journal= {arXiv preprint arXiv:2509.02996},
year = {2025}
}
Comments
68 pages