# AutoBayes: A Compositional Framework for Generalized Variational Inference

arXiv: https://arxiv.org/abs/2503.18608 · HTML: https://mlsanity.com/paper/2503.18608/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 7 of its 15 references that are in the map.
- First posted on arXiv 24 Mar 2025.

## Builds on (most cited first, 7 of 7)

- [Pyro: Deep Universal Probabilistic Programming](https://mlsanity.com/paper/1810.09538/) (2018)
- [The Bayesian Learning Rule](https://mlsanity.com/paper/2107.04562/) (2021)
- [Conjugate-Computation Variational Inference : Converting Variational Inference in Non-Conjugate Models to Inferences in Conjugate Models](https://mlsanity.com/paper/1703.04265/) (2017)
- [Deep Amortized Inference for Probabilistic Programs](https://mlsanity.com/paper/1610.05735/) (2016)
- [Compositional game theory](https://mlsanity.com/paper/1603.04641/) (2016)
- [Mathematical Foundations for a Compositional Account of the Bayesian Brain](https://mlsanity.com/paper/2212.12538/) (2022)
- [The Compositional Structure of Bayesian Inference](https://mlsanity.com/paper/2305.06112/) (2023)

## Abstract (arXiv)

We introduce a new compositional framework for generalized variational inference, clarifying the different parts of a model, how they interact, and how they compose. We explain that both exact Bayesian inference and the loss functions typical of variational inference (such as variational free energy and its generalizations) satisfy chain rules akin to that of reverse-mode automatic differentiation, and we advocate for exploiting this to build and optimize models accordingly. To this end, we construct a series of compositional tools: for building models; for constructing their inversions; for attaching local loss functions; and for exposing parameters. Finally, we explain how the resulting parameterized statistical games may be optimized locally, too. We illustrate our framework with a number of classic examples, pointing to new areas of extensibility that are revealed.