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

DOI10.1016/J.AIM.2020.107239zbMATH Open1505.60004arXiv1908.07021OpenAlexW2969875089MaRDI QIDQ2189508FDOQ2189508

Authors: Tobias Fritz 

Publication date: 15 June 2020

Published in: Advances in Mathematics (Search for Journal in Brave)

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.

Full work available at URL: https://arxiv.org/abs/1908.07021

Recommendations

• scientific article; zbMATH DE number 1324650
• On independence of Markov kernels and a generalization of two theorems of Basu
• A categorical foundation for Bayesian probability
• Infinite products and zero-one laws in categorical probability
• Towards a categorical account of conditional probability

zbMATH Keywords

conditional independencefoundations of statisticsfoundations of probabilitysufficient statisticalmost surelymonoidal categories with structure

Mathematics Subject Classification ID

Foundations and philosophical topics in statistics (62A01) Sufficient statistics and fields (62B05) Monads (= standard construction, triple or triad), algebras for monads, homology and derived functors for monads (18C15) Axioms; other general questions in probability (60A05) Semantics in the theory of computing (68Q55) Monoidal categories, symmetric monoidal categories (18M05)

Cites Work

• Title not available (Why is that?)
• Title not available (Why is that?)
• Selected works of E. L. Lehmann. Edited by Javier Rojo
• Title not available (Why is that?)
• Causality. Models, reasoning, and inference
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Conditional independence for statistical operations
• Title not available (Why is that?)
• Application of the Radon-Nikodym Theorem to the Theory of Sufficient Statistics
• The geometry of tensor calculus. I
• Modelling environments in call-by-value programming languages.
• Sur quelques points d'algèbre homologique
• A survey of graphical languages for monoidal categories
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• General Relativity
• Title not available (Why is that?)
• Title not available (Why is that?)
• Cartesian bicategories. I
• Labelled Markov processes.
• Categorical semantics for arrows
• Title not available (Why is that?)
• Title not available (Why is that?)
• Every Standard Construction is Induced by a Pair of Adjoint Functors
• Probability theory. A comprehensive course
• Title not available (Why is that?)
• Measure theory. Vol. 3. Measure algebras
• Extremal families and systems of sufficient statistics
• What is a statistical model? (With comments and rejoinder).
• Hypergraph categories
• From Kleisli Categories to Commutative C*-algebras: Probabilistic Gelfand Duality
• Causal categories: relativistically interacting processes
• Classical and quantum structuralism
• Title not available (Why is that?)
• Terminality implies non-signalling
• Separoids: a mathematical framework for conditional independence and irrelevance
• Categories tannakiennes
• Borel structures for function spaces
• Framed bicategories and monoidal fibrations
• Picturing classical and quantum Bayesian inference
• Title not available (Why is that?)
• The existence of regular conditional probabilities: Necessary and sufficient conditions
• Statistical Isomorphism
• A class of fuzzy theories
• An Algebraic Theory of Markov Processes
• Semantics of weakening and contraction
• Probability Sheaves and the Giry Monad
• A Convenient Category for Higher-Order Probability Theory
• Markov invariant geometry and manifolds of states
• Title not available (Why is that?)
• Functional distribution monads in functional-analytic contexts
• Tensors, monads and actions
• Matrices, relations, and group representations
• Informativity in the category of linear measurement systems
• Combining statistical models
• Title not available (Why is that?)
• Codensity and the Giry monad
• Title not available (Why is that?)
• A predicate/state transformer semantics for Bayesian learning
• Title not available (Why is that?)
• A Note on Statistical Equivalence
• Theory of fuzzy sets as a theory of uncertainty and decision-making problems in fuzzy experiments
• Informativity in the category of multivalued information transformers
• Axiomatic description of categories of information transformers
• Relative informativity and a priori information in the category of linear information transformers
• Method of additional structures on the objects of a monoidal Kleisli category as a background for information transformers theory
• Information transformers: category-theoretical structure, informativeness, decision-making problems
• Beyond Bell's theorem. II: Scenarios with arbitrary causal structure
• Algebras of the extended probabilistic powerdomain monad
• Algebraic foundation of mathematical statistics²
• Title not available (Why is that?)
• Disintegration and Bayesian inversion via string diagrams
• Kantorovich problems and conditional measures depending on a parameter
• Title not available (Why is that?)
• Title not available (Why is that?)
• Compositories and Gleaves
• Borel kernels and their approximation, categorically
• Selected Works of Debabrata Basu

Cited In (41)

• Stick breaking, in coalgebra and probability
• The existence of regular conditional probabilities for Markov kernels
• The groupoidal picture of quantum mechanics
• Weak Cartesian properties of simplicial sets
• Coend Optics for Quantum Combs
• Monoidal Width: Capturing Rank Width
• Open Dynamical Systems as Coalgebras for Polynomial Functors, with Application to Predictive Processing
• Representable Markov categories and comparison of statistical experiments in categorical probability
• Monoidal Width
• A non-commutative Bayes' theorem
• Monotone metric tensors in quantum information geometry
• Title not available (Why is that?)
• Categorical composable cryptography: extended version
• A categorical approach to statistical mechanics
• Categorical information geometry
• Multisets and distributions, in drawing and learning
• A category-theoretic proof of the ergodic decomposition theorem
• Free gs-monoidal categories and free Markov categories
• Cyber Kittens, or Some First Steps Towards Categorical Cybernetics
• Value Iteration is Optic Composition
• On independence of Markov kernels and a generalization of two theorems of Basu
• A principled approach to expectation maximisation and latent Dirichlet allocation using Jeffrey's update rule
• From gs-monoidal to oplax cartesian categories: constructions and functorial completeness
• Non-commutative disintegrations: existence and uniqueness in finite dimensions
• Drawing from an urn is isometric
• Towards a compositional framework for convex analysis (with applications to probability theory)
• Parametric models and information geometry on \(W^\star\)-algebras
• A higher-order language for Markov kernels and linear operators
• Conditional distributions for quantum systems
• Cartesian double theories: a double-categorical framework for categorical doctrines
• Partitions and Ewens distributions in element-free probability theory
• Monoidal streams for dataflow programming
• Probability monads with submonads of deterministic states
• Title not available (Why is that?)
• Learning from what's right and learning from what's wrong
• Multinomial and hypergeometric distributions in Markov categories
• Categorial independence and Lévy processes
• Causal inference via string diagram surgery
• A reconstruction of Ewens' sampling formula via lists of coins
• Title not available (Why is that?)
• Title not available (Why is that?)

Uses Software

• Separoids

This page was built for publication: A synthetic approach to Markov kernels, conditional independence and theorems on sufficient statistics

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2189508)