On the accept-reject mechanism for Metropolis-Hastings algorithms
This work develops a powerful and versatile framework for determining acceptance ratios in Metropolis-Hastings type Markov kernels widely used in statistical sampling problems. Our approach allows us to derive new classes of kernels which unify popular random walk or diffusion-type sampling methods with more complicated "extended phase space" algorithms based on Hamiltonian dynamics. The starting point for our approach is an abstract result developed in the generality of measurable state spaces that addresses proposal kernels that possess a certain involution structure. While this underlying involution proposal structure suggests a scope which includes Hamiltonian-type kernel structures, we demonstrate that our result also recovers a broad range of kernels encompassed by a previous general state space approach in [Tierney, Annals of Applied Probability, 1998]. Altogether, the theoretical unity and reach of our main result provides a basis for deriving novel sampling algorithms while laying bare important relationships between existing methods.
READ FULL TEXT