Approximate Inference via Fibrations of Statistical Games

Toby St Clere Smithe

We characterize a number of well known systems of approximate inference as loss models: lax sections of 2-fibrations of statistical games, constructed by attaching internally-defined loss functions to Bayesian lenses. Our examples include the relative entropy, which constitutes a strict section, and whose chain rule is formalized by the horizontal composition of the 2-fibration. In order to capture this compositional structure, we first introduce the notion of 'copy-composition', alongside corresponding bicategories through which the composition of copy-discard categories factorizes. These bicategories are a variant of the Copara construction, and so we additionally introduce coparameterized Bayesian lenses, proving that coparameterized Bayesian updates compose optically, as in the non-coparameterized case.

In Sam Staton and Christina Vasilakopoulou: Proceedings of the Sixth International Conference on Applied Category Theory 2023 (ACT 2023), University of Maryland, 31 July - 4 August 2023, Electronic Proceedings in Theoretical Computer Science 397, pp. 279–298.
Published: 14th December 2023.

ArXived at: https://dx.doi.org/10.4204/EPTCS.397.17 bibtex PDF

Comments and questions to: eptcs@eptcs.org
For website issues: webmaster@eptcs.org