On hitting time, mixing time and geometric interpretations of Metropolis-Hastings reversiblizations
Michael C.H. Choi, Lu-Jing Huang

TL;DR
This paper introduces a new Metropolis-Hastings generator $M_2$ that, along with the classical $M_1$, offers improved mixing properties and a geometric interpretation as $\, ext{l}^1$-minimizers between the proposal generator and reversible generators.
Contribution
The paper defines a novel MH generator $M_2$, compares it with $M_1$ across multiple metrics, and provides a geometric framework extending previous results.
Findings
$M_2$ has superior mixing properties than $M_1$.
Explicit spectral analysis for Metropolised independent sampling.
Laplace transform order of the fastest strong stationary time between $M_1$ and $M_2$.
Abstract
Given a target distribution and a proposal chain with generator on a finite state space, in this paper we study two types of Metropolis-Hastings (MH) generator and in a continuous-time setting. While is the classical MH generator, we define a new generator that captures the opposite movement of and provide a comprehensive suite of comparison results ranging from hitting time and mixing time to asymptotic variance, large deviations and capacity, which demonstrate that enjoys superior mixing properties than . To see that and are natural transformations, we offer an interesting geometric interpretation of , and their convex combinations as minimizers between and the set of -reversible generators, extending the results by Billera and Diaconis (2001). We provide two examples as…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
