Approximate Inference via Fibrations of Statistical Games
arXiv:2306.17009 · doi:10.4204/EPTCS.397.17
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.
arXiv commentIn Proceedings ACT 2023, arXiv:2312.08138
References in corpus (11)
- Auto-Encoding Variational Bayes
- Disintegration and Bayesian Inversion via String Diagrams
- Monoidal Grothendieck construction
- Towards Foundations of Categorical Cybernetics
- Bayesian Updates Compose Optically
- On S-Finite Measures and Kernels
- Diegetic Representation of Feedback in Open Games
- Seeing double through dependent optics
- Dependent Optics
- Cyber Kittens, or Some First Steps Towards Categorical Cybernetics
- Fibre optics