Slice Sampling for Probabilistic Programming

TL;DR

This paper introduces a versatile slice sampling inference engine integrated into StocPy, a Turing-complete probabilistic programming language, demonstrating superior performance and flexibility over existing methods across various models.

Contribution

The paper presents the first general-purpose slice sampling inference engine for probabilistic programs, implemented in a new Python-based language, StocPy, with a novel transdimensional extension.

Findings

Slice sampling outperforms previous inference methods in experiments.

StocPy offers enhanced flexibility and usability compared to other PPLs.

The inference engine effectively handles models like logistic regression, HMM, and Bayesian Neural Net.

Abstract

We introduce the first, general purpose, slice sampling inference engine for probabilistic programs. This engine is released as part of StocPy, a new Turing-Complete probabilistic programming language, available as a Python library. We present a transdimensional generalisation of slice sampling which is necessary for the inference engine to work on traces with different numbers of random variables. We show that StocPy compares favourably to other PPLs in terms of flexibility and usability, and that slice sampling can outperform previously introduced inference methods. Our experiments include a logistic regression, HMM, and Bayesian Neural Net.