A synthetic approach to Markov kernels, conditional independence and theorems on sufficient statistics

Abstract

We develop Markov categories as a framework for synthetic probability and statistics, following work of Golubtsov as well as Cho and Jacobs. This means that we treat the following concepts in purely abstract categorical terms: conditioning and disintegration; various versions of conditional independence and its standard properties; conditional products; almost surely; sufficient statistics; versions of theorems on sufficient statistics due to Fisher--Neyman, Basu, and Bahadur. Besides the conceptual clarity offered by our categorical setup, its main advantage is that it provides a uniform treatment of various types of probability theory, including discrete probability theory, measure-theoretic probability with general measurable spaces, Gaussian probability, stochastic processes of either of these kinds, and many others.

98 pages. v6: fixed error in Section 7. v7: incorporates referee's comments. v8: minor correction

References in corpus (7)

• A generalized no-broadcasting theorem
• Denotational validation of higher-order Bayesian inference
• Terminality implies non-signalling
• Equivalence of relativistic causal structure and process terminality
• Supplying bells and whistles in symmetric monoidal categories
• Method of Additional Structures on the Objects of a Monoidal Kleisli Category as a Background for Information Transformers Theory
• Algebras of the extended probabilistic powerdomain monad

Cited by in corpus (66)

• Axioms for retrodiction: achieving time-reversal symmetry with a prior
• Representable Markov Categories and Comparison of Statistical Experiments in Categorical Probability
• On quantum states over time
• A non-commutative Bayes' theorem
• Infinite products and zero-one laws in categorical probability
• Free gs-monoidal categories and free Markov categories
• Bayesian Updates Compose Optically
• A category-theoretic proof of the ergodic decomposition theorem
• Dilations and information flow axioms in categorical probability
• Markov Categories and Entropy
• The information loss of a stochastic map
• Category Theory in Machine Learning
• Parametric models and information geometry on W*-algebras
• Inverses, disintegrations, and Bayesian inversion in quantum Markov categories
• Bayesian inversion and the Tomita-Takesaki modular group
• A functorial characterization of von Neumann entropy
• Bayesian open games
• Multinomial and Hypergeometric Distributions in Markov Categories
• Coend Optics for Quantum Combs
• Learning from What's Right and Learning from What's Wrong
• Probability monads with submonads of deterministic states - Extended version
• Homotopy Theoretic and Categorical Models of Neural Information Networks
• Categorical Stochastic Processes and Likelihood
• An Operational Environment for Quantum Self-Testing
• The algebra and machine representation of statistical models
• Categorical composable cryptography: extended version
• Conditional Distributions for Quantum Systems
• A Framework for Universality in Physics, Computer Science, and Beyond
• Cyber Kittens, or Some First Steps Towards Categorical Cybernetics
• Functorial Manifold Learning
• Probabilistic programming interfaces for random graphs: Markov categories, graphons, and nominal sets
• Open Dynamical Systems as Coalgebras for Polynomial Functors, with Application to Predictive Processing
• A Categorical Semantics of Fuzzy Concepts in Conceptual Spaces
• From Gs-monoidal to Oplax Cartesian Categories: Constructions and Functorial Completeness
• Structured and Decorated Cospans from the Viewpoint of Double Category Theory
• Monoidal Width
• Sufficient Statistics and Split Idempotents in Discrete Probability Theory
• Value Iteration is Optic Composition
• A category theory framework for Bayesian learning
• A Categorical Construction of the Real Unit Interval
• Weak cartesian properties of simplicial sets
• Hidden Markov Models and the Bayes Filter in Categorical Probability
• A Mathematical Framework for Causally Structured Dilations and its Relation to Quantum Self-Testing
• Bipartite quantum states admitting a causal explanation
• On the categorical foundations of quantum information theory: Categories and the Cramer-Rao inequality
• Lifting couplings in Wasserstein spaces
• Reinforcement Learning in Categorical Cybernetics
• Compositional Semantics for Probabilistic Programs with Exact Conditioning
• Overdrawing Urns using Categories of Signed Probabilities
• Compositional Abstraction Error and a Category of Causal Models
• On a 2-relative entropy
• Unifilar Machines and the Adjoint Structure of Bayesian Filtering
• String Diagrams with Factorized Densities
• A model of stochastic memoization and name generation in probabilistic programming: categorical semantics via monads on presheaf categories
• Cartesian double theories: A double-categorical framework for categorical doctrines
• The Aldous$\unicode{x2013}$Hoover Theorem in Categorical Probability
• Functors induced by comma categories
• De Finetti's construction as a categorical limit
• Biased elementary doctrines and quotient completions
• Drawing with Distance
• Quasi-Measurable Spaces: A Convenient Foundation of Probability Theory
• Categorical algebra of conditional probability
• Identification of Causal Influences in Quantum Processes
• Some Notions of (Open) Dynamical System on Polynomial Interfaces
• Monoidal Width: Capturing Rank Width
• Categories of abstract and noncommutative measurable spaces