AutoBayes: A Compositional Framework for Generalized Variational Inference

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Smithe, Toby St Clere, Perin, Marco
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912292362256384
author Smithe, Toby St Clere
Perin, Marco
author_facet Smithe, Toby St Clere
Perin, Marco
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2503_18608
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle AutoBayes: A Compositional Framework for Generalized Variational Inference
Smithe, Toby St Clere
Perin, Marco
Machine Learning
Statistics Theory
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.
title AutoBayes: A Compositional Framework for Generalized Variational Inference
topic Machine Learning
Statistics Theory
url https://arxiv.org/abs/2503.18608