Autoconj: Recognizing and Exploiting Conjugacy Without a Domain-Specific Language
This work addresses a bottleneck for researchers and developers in probabilistic modeling by enabling faster and more reliable inference algorithm development, though it is incremental as it builds on existing conjugacy concepts.
The paper tackles the problem of automating derivations of conditional and marginal distributions using conjugacy relationships, which are often time-consuming and error-prone, by introducing Autoconj, a system that operates directly on Python functions for log-joint distributions and supports conjugacy-exploiting algorithms in any Python-embedded probabilistic programming language.
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.