Unifilar Machines and the Adjoint Structure of Bayesian Filtering

DOI10.4204/EPTCS.397.18arXiv2305.02826MaRDI QIDQ6435332FDOQ6435332

Publication date: 4 May 2023

Abstract: We apply recent work on category theoretical probability to the idea of Bayesian filtering, making use of the concept of a strongly representable Markov category. We show that there is an adjunction between 'dynamical' and 'epistemic' models of a hidden Markov process. Concepts such as Bayesian filtering and conjugate priors arise as natural consequences of this adjunction. Along the way we define a notion of unifilar machine, which is a kind of stochastic Moore machine in which the output is chosen stochastically, but the update function is deterministic given the output. Unifilar machines are useful as models of the behaviour of stochastic systems; we show that in the Kleisli category of the distribution monad there is a terminal unifilar machine, and its elements are controlled stochastic processes, mapping sequences of the input alphabet probabilistically to sequences of the output alphabet.

This page was built for publication: Unifilar Machines and the Adjoint Structure of Bayesian Filtering

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6435332)