Adaptive Generalized Elliptical Slice Sampling

TL;DR

This paper introduces an adaptive generalized elliptical slice sampling algorithm that improves sampling efficiency for complex, high-dimensional, and non-elliptical distributions while maintaining ergodicity.

Contribution

It develops a new adaptive elliptical slice sampler that addresses prior limitations and demonstrates robustness across diverse challenging target distributions.

Findings

Enhanced sampling efficiency in various complex models

Robust performance on high-dimensional and multi-modal distributions

Proven ergodicity under general conditions

Abstract

Elliptical slice sampling is a widely used gradient-free Markov chain Monte Carlo algorithm that is tuning-free and capable of adapting to local characteristics of the target distribution. However, its primary limitation is that sampling efficiency can quickly degrade when there is a mismatch between the prior distribution and the target distribution. To directly address this limitation, we introduce an adaptive generalized elliptical slice sampler that offers compelling gains in sampling efficiency while preserving many of the appealing properties of the standard elliptical slice sampler. We demonstrate the utility of the adaptive algorithm on a broad collection of target distributions arising from realistic modeling scenarios; including generalized regression, deep Gaussian process surrogate modeling, and high-dimensional sparse regression. Collectively, these case studies demonstrate the efficiency and robustness of adaptive generalized elliptical slice sampling across target distributions that are non-elliptical, non-differentiable, multi-modal, and/or high-dimensional. Under fairly general regularity conditions, we establish that the proposed adaptive generalized elliptical slice sampling algorithm is ergodic.