Bayesian Inference for Discrete Markov Random Fields Through Coordinate Rescaling

TL;DR

This paper introduces coordinate-rescaling sampling methods for Bayesian inference in discrete Markov random fields, improving uncertainty quantification while maintaining computational efficiency and scalability.

Contribution

It proposes a novel class of coordinate-rescaling samplers that enhance posterior variability estimates in discrete MRFs, addressing limitations of existing methods.

Findings

Coordinate-rescaling improves posterior variability estimates.

Methods scale well to large systems.

Simulation studies show better accuracy than existing approaches.

Abstract

Discrete Markov random fields are undirected graphical models that capture complex conditional dependencies between discrete variables. Conducting exact posterior inference in these models is often computationally challenging because evaluating their normalizing constant requires summation over all possible state configurations, and the size of this state space grows exponentially with the number of variables and their possible states. As a result, exact likelihood-based inference is infeasible in many practical settings, and existing methods, such as Double Metropolis-Hastings or pseudo-likelihood approximations, either scale poorly to large systems or underestimate posterior variability. To address these limitations, we propose a new class of coordinate-rescaling sampling methods that transform pseudo-likelihood-based posteriors toward the target posterior while preserving computational efficiency. The resulting samplers retain scalability while improving uncertainty quantification. In simulation studies, we compare the proposed methods to existing approaches and demonstrate that coordinate-rescaling sampling yields more accurate estimates of posterior variability, providing a scalable and reliable approach to Bayesian inference in discrete MRFs.