Autoconj: Recognizing and Exploiting Conjugacy Without a Domain-Specific Language

TL;DR

Autoconj automates the derivation of conjugate distributions directly from Python functions, simplifying and accelerating the development of probabilistic inference algorithms without relying on domain-specific languages.

Contribution

It introduces a system that recognizes and exploits conjugacy relationships directly from log-joint functions in Python, unlike prior systems limited to variable pairs.

Findings

Automates conjugacy derivations from Python functions

Supports any Python embedded probabilistic programming language

Enables faster development of inference algorithms

Abstract

Deriving conditional and marginal distributions using conjugacy relationships can be time consuming and error prone. In this paper, we propose a strategy for automating such derivations. Unlike previous systems which focus on relationships between pairs of random variables, our system (which we call Autoconj) operates directly on Python functions that compute log-joint distribution functions. Autoconj provides support for conjugacy-exploiting algorithms in any Python embedded PPL. This paves the way for accelerating development of novel inference algorithms and structure-exploiting modeling strategies.