English

Approximate Inference via Fibrations of Statistical Games

Category Theory 2023-12-15 v2 Statistics Theory Statistics Theory

Abstract

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.

Keywords

Cite

@article{arxiv.2306.17009,
  title  = {Approximate Inference via Fibrations of Statistical Games},
  author = {Toby St Clere Smithe},
  journal= {arXiv preprint arXiv:2306.17009},
  year   = {2023}
}

Comments

In Proceedings ACT 2023, arXiv:2312.08138