Pith. sign in

REVIEW 2 major objections 3 minor 5 cited by

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

T0 review · 2 major / 3 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Sufficiency becomes a pure factorization condition in any Markov category, and the classical statistical theorems follow.

desk verdict Real contribution to categorical probability with an honest, substantive gap: the abstract conditional independence notion isn't yet matched to the standard measure-theoretic one in Stoch. read the letter →

arxiv 1908.07021 v8 pith:2ZM2QAB3 submitted 2019-08-19 math.ST cs.LOmath.CTmath.PRstat.TH

Tobias Fritz This is my paper · ORCID
classification math.STcs.LOmath.CTmath.PRstat.TH MSC 60A0562A0162B0518D1068Q55
keywords MarkovcategoriesconditionalindependencesufficientstatisticsFisher–NeymanfactorizationBasu'stheoremsyntheticprobabilitysemigraphoidpropertiesalmostsureequality
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that the core concepts of probability and statistics—Markov kernels, conditioning, conditional independence, almost sure equality, sufficiency, and completeness—can be developed once, abstractly, in any Markov category, meaning any system of noisy maps equipped with operations for copying and discarding values. The payoff is a single set of proofs that specialize to discrete probability, measure-theoretic probability, Gaussian probability, and stochastic processes. On the statistics side, it proves an abstract characterization of sufficient statistics and derives abstract versions of the Fisher–Neyman, Basu, and Bahadur theorems. A reader should care because the result suggests that measure theory is one concrete model of probability rather than its mandatory foundation.

What carries the argument

The central object is a Markov category: a symmetric monoidal category whose morphisms are interpreted as noisy maps, with a copying operation $X \to X \otimes X$ and a discarding operation $X \to I$ on every object. This structure suffices to define joint distributions, conditionals, conditional independence via string-diagram factorization, almost sure equality, statistics, and completeness. The load-bearing identity is the sufficiency equation (14.1), which under strict positivity becomes equivalent to $\alpha s p = p$ together with $s\alpha =_{sp\text{-a.s.}} \mathrm{id}_V$; strict positivity is the axiom that a $p$-almost surely deterministic composite forces a product structure, and it holds in all the paper's main examples.

What would settle it

In Stoch, take a joint distribution $\psi$ on $X\times W\times Y$ and test whether the factorization condition of Definition 12.1 is equivalent to the standard conditional-expectation equality $\mathbb{E}[1_S 1_T \mid \Sigma_W] = \mathbb{E}[1_S \mid \Sigma_W] \mathbb{E}[1_T \mid \Sigma_W]$; a distribution satisfying exactly one of the two would settle whether the synthetic and measure-theoretic notions coincide.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is Theorem 14.5: in a strictly positive Markov category, a deterministic statistic $s: X \to V$ is sufficient for a statistical model $p: \Theta \to X$ exactly when there is $\alpha: V \to X$ with $\alpha s p = p$ and $s\alpha =_{sp\text{-a.s.}} \mathrm{id}_V$. Sufficiency therefore becomes a pure factorization and almost-sure splitting condition, with no densities and no integration. The paper also proves that completeness is a property of a single morphism rather than of a statistic-model pair, that a complete sufficient statistic is independent of every ancillary statistic (Basu), and that a complete sufficient statistic is minimal sufficient whenever a minimal sufficient statistic exists (Bahadur).

Load-bearing premise

The unified-treatment claim rests on the open assumption that the paper's string-diagram conditional independence coincides with the standard conditional-expectation definition in Stoch; if that correspondence fails, some abstract theorems would still hold but their translation to classical statistics would be incomplete.

Editorial extensions

If this is right

  • The same abstract proofs instantiate in finite discrete probability, measure-theoretic probability over general measurable spaces, Gaussian probability, and diagram categories representing stochastic processes, so the statistical theorems transfer across all of these.
  • In the category of measurable spaces and Markov kernels, the abstract notion of completeness specializes to bounded completeness, and the abstract Basu theorem becomes the standard Basu theorem.
  • Any Markov category with conditionals supports conditional products that satisfy the Dawid–Studený axioms, giving a compositional calculus for gluing joint distributions.
  • In a causal Markov category, Bayesian inversion is a symmetric monoidal dagger functor on the category of probability spaces, resolving an open problem about the structure of Bayesian inversion.
  • Sufficiency is compositional: if a statistic $s$ is sufficient for $p$ and a statistic $t$ is sufficient for the pushed-forward model $sp$, then the composite $ts$ is sufficient for $p$.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the paper's string-diagram conditional independence were shown to agree with the standard conditional-expectation definition in the measure-theoretic category Stoch, the uniform treatment would extend to the full classical theory; the author leaves that correspondence open.
  • The factorization characterization suggests a proof-search strategy for new statistical results: conjecture a string-diagram identity, check it in a concrete Markov category such as standard Borel spaces, and then export it to every other instance.
  • Because Markov categories can be built from relations, matrices over semirings, and commutative comonoids, the same theorems may have analogues in non-probabilistic settings such as multivalued functions or fuzzy information transformers.
  • A natural next step, left as Problem 16.4 in the paper, is to determine whether every statistical model in a category with sufficiently nice colimits has a minimal sufficient statistic.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 3 minor

Summary. The paper develops Markov categories as a synthetic framework for probability and statistics, following Golubtsov and Cho–Jacobs. It gives purely categorical definitions of conditioning, several variants of conditional independence, almost sure equality, sufficient, complete and ancillary statistics, and minimal sufficient statistics, and proves abstract versions of the semigraphoid properties, the Fisher–Neyman factorization theorem, Basu's theorem, and Bahadur's theorem. The framework is illustrated with many examples: FinStoch, Stoch, BorelStoch, the Radon-monad Kleisli category, Gaussian probability, functor categories for stochastic processes, and hypergraph/cocommutative-comonoid constructions. The main advertised advantage is a uniform treatment of discrete, measure-theoretic, Gaussian, and process-level probability theories.

Significance. If the advertised dictionary with measure-theoretic probability is established, this would be a valuable unifying contribution to categorical probability and to the foundations of statistics. The paper is careful and honest: it states limitations explicitly (Remark 12.4, Example 10.4, the open measure-theoretic Fisher–Neyman instantiation) and provides many worked examples. The abstract proofs are clean string-diagram arguments, and the axiomatization is parameter-free in the sense that the theorems are derived from the stated categorical axioms. The current significance is tempered by the fact that the central example Stoch has not been shown to match the standard measure-theoretic notion of conditional independence, which is a key advertised application.

major comments (2)
  1. [§12, Remark 12.4] The paper leaves open whether Definition 12.1 agrees with the standard measure-theoretic conditional independence defined via conditional expectations, equation (12.2), in Stoch. In Stoch, Definition 12.1 requires a factorization through Markov kernels f and g, which by Remark 12.2 must be regular conditional distributions; this is strictly stronger than the conditional-expectation notion, which does not require regular conditionals, as the paper itself notes in Examples 11.3 and 11.18. Consequently, the semigraphoid properties proven in Lemmas 12.5, 12.13, 12.17, and 12.20 do not automatically transfer to the notion used in measure-theoretic probability. Since the abstract claims a uniform treatment of measure-theoretic probability, this gap weakens a central advertised contribution. The authors should either prove the correspondence (for BorelStoch, or via a Markov category as in Conjecture 11.10), or explicitly state that the framework currently captures a regular-conditional-based notion of conditional independence, which is stronger than the standard one in full Stoch.
  2. [§14, Theorem 14.5] Theorem 14.5 characterizes sufficiency in strictly positive Markov categories via the splitting conditions αsp = p and sα =_{sp-a.s.} id_V, and the paper describes this as an abstract close relative of the Fisher–Neyman factorization theorem. However, it does not verify that the abstract sufficiency condition (14.1) coincides with the classical measure-theoretic definition of sufficiency in Stoch or BorelStoch, and on page 86 it explicitly leaves open whether the theorem specializes to the standard measure-theoretic Fisher–Neyman factorization theorem. Since Stoch is the primary example for measure-theoretic probability, the advertised instantiation of the theorem in that setting is not yet established. The authors should prove the equivalence at least for BorelStoch, where regular conditional distributions exist, or state more precisely which notion of sufficiency the theorem captures in general Stoch.
minor comments (3)
  1. [§12.4, Eq. (12.2)] The notation S ∩ T is abusive because S ∈ Σ_X and T ∈ Σ_Y are subsets of different spaces; the pullbacks of these sets to the product space should be used, or the conditional expectation should be written with the indicator functions explicitly.
  2. [§13.1, Definition 13.1] The definition of p-a.s. equality via equality after composing with copy is clear to category theorists but may be opaque to statisticians; a short concrete instantiation (e.g., in FinStoch, as in Example 13.2) immediately following the definition would help.
  3. [§14, Definition 14.2 and §10, Example 10.4] Since statistics are defined as deterministic morphisms, and in Stoch deterministic morphisms are {0,1}-valued kernels that need not coincide with measurable maps unless the σ-algebras are countably generated and separating, the reader should be reminded that the clean identification of statistics with measurable functions holds in BorelStoch, which is the intended setting for the statistical instantiations.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's results are derived from explicitly stated axioms, and its admitted gaps are limitations rather than circular reductions.

full rationale

The derivation chain is self-contained in the sense relevant to circularity: Markov categories, conditional independence, almost-sure equality, sufficiency, completeness, and ancillary statistics are all given by explicit definitions (Definitions 2.1, 12.1, 12.12, 12.19, 13.1, 14.1–14.3, 15.1, 15.7), and the main theorems are proven from these definitions together with stated axioms such as strict positivity (Definition 13.16). Theorem 14.5 does not presuppose sufficiency: sufficiency is defined by the existence of a witness α satisfying (14.1), and the theorem derives the equivalent factorization-type condition αsp = p and sα =_{sp-a.s.} id using strict positivity. The proof is short but not a mere restatement; the 'if' direction is exactly where the strict-positivity axiom is used. Example 14.6 then instantiates the abstract result in FinStoch and derives the Fisher–Neyman factorization, so the abstract theorem is not a renamed version of the classical theorem. The paper's admitted open point, Remark 12.4, is that the abstract conditional independence of Definition 12.1 has not been proved to coincide with the conditional-expectation definition (12.2) in Stoch; this is an honest limitation of the claimed uniform treatment, not a circular step. Citations to Golubtsov and Cho–Jacobs are to prior external work and are used for definitions and prior results; they are not self-citations of the present author. No fitted parameters are renamed as predictions, and no uniqueness theorem from the authors' own prior work is invoked to force a choice.

Assumptions & free parameters 0 free parameters · 6 assumptions · 0 invented entities

This is a pure mathematics paper, so there are no fitted numerical parameters. The central theorems depend on the Markov category axioms and on additional candidate axioms (positivity, causality, strict positivity, conditionals) that the paper introduces and studies. No new empirical entities are postulated.

assumptions (6)
  • domain assumption Every object X has a commutative comonoid (copy_X, del_X) compatible with the symmetric monoidal structure; del is natural (Definition 2.1).
    This is the defining axiom of a Markov category; all subsequent results assume it.
  • domain assumption The monoidal unit I is terminal (semicartesian).
    Derived in Remark 2.3 from del_I = id and naturality; needed for the interpretation of del as discarding and for the subcategory of deterministic morphisms to be cartesian.
  • ad hoc to paper C is strictly positive (Definition 13.16): whenever gf is p-a.s. deterministic, the displayed equation (13.3) holds.
    This axiom is introduced in this paper and is a hypothesis in the abstract Fisher-Neyman theorem (14.5), Basu (15.8), and Bahadur (16.3). The paper proves it holds for FinStoch, Gauss, and Stoch.
  • ad hoc to paper C is causal (Definition 11.31).
    Introduced as a candidate axiom; needed for well-defined composition in the category of probability spaces ProbStoch(C) (Proposition 13.9).
  • ad hoc to paper C has conditionals (Definition 11.5).
    Assumed in many results: conditional products (Section 12), Bayesian inversion dagger (Remark 13.10), weak union properties (Lemmas 12.5, 12.17, 12.20). Not assumed globally; BorelStoch satisfies it via Bogachev-Malofeev, while Stoch does not.
  • standard math Standard measure-theoretic background: Giry monad, product σ-algebras, π-λ theorem, Radon measures.
    Used to construct Stoch, BorelStoch, and CHaus examples and to verify they are Markov categories (Sections 4-5).

how reviews work

0 comments
Cite this review

Pith. "Pith review of A synthetic approach to Markov kernels, conditional independence and theorems on sufficient statistics." pith.science (2026). https://pith.science/paper/2ZM2QAB3

@misc{pith2026190807021,
  author       = {Pith},
  title        = {Pith review of: A synthetic approach to Markov kernels, conditional independence and theorems on sufficient statistics},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2ZM2QAB3}},
  note         = {Machine review of arXiv:1908.07021}
}
read the original 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.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 5 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Finite Observations, Infinite Behaviour: bicategorical semantics for stateful monoidal processes

    cs.LO 2026-07 accept novelty 7.5 of 10

    Behaviours of stateful monoidal processes are equivalence classes of compatible finite observations in discard bicategories, yielding functorial feedback semantics and a categorified compactness theorem for closed relations.

  2. Convex Biproducts, Stochastic Matrices and Tape Diagrams

    cs.LO 2026-07 accept novelty 7.0 of 10

    Convex biproducts free-generate substochastic matrix categories isomorphic to probabilistic tape diagrams, giving a complete axiomatisation of probabilistic Boolean circuits.

  3. Supplying bells and whistles in symmetric monoidal categories

    math.CT 2019-08 accept novelty 7.0 of 10

    The paper defines 'supply' of a prop in a symmetric monoidal category and proves that associators, unitors, and braidings are automatically homomorphisms for any supply.

  4. Local Uniqueness of the Born Rule on Categories with Complex-Weighted Morphisms

    quant-ph 2026-08 accept novelty 4.0 of 10

    A polynomial probability functional on complex amplitudes that is phase invariant, phase-averaged additive, and normalized must equal the squared modulus, the Born rule.

  5. Categorical and geometric methods in statistical, manifold, and machine learning

    stat.ML 2025-05 unverdicted novelty 2.0 of 10

    The paper is a survey of the authors' categorical and geometric framework for statistical learning, with no substantially new results.

Reference graph

Works this paper leans on

118 extracted references · 68 canonical work pages · cited by 5 Pith papers

  1. [1]

    Robert J. Aumann. Borel structures for function spaces. Illinois J. Math. , 5:614–630, 1961. ↑ 50

    1961

  2. [2]

    Codensity and the Giry monad

    Tom A very. Codensity and the Giry monad. J. Pure Appl. Algebra , 220(3):1229–1251, 2016. ↑ 7, 19

    2016

  3. [3]

    An algebraic theory of Markov processes

    Giorgio Bacci, Radu Mardare, Prakash Panangaden, and Go rdon Plotkin. An algebraic theory of Markov processes. In LICS ’18/emdash.cyr33rd Annual ACM/IEEE Symposium on Logic in Computer Science , page 10. ACM, New York, 2018. cs.mcgill.ca/∼ prakash/Pubs/Markov-LICS.pdf. ↑ 7

    2018

  4. [4]

    A generalized no-broadcasting theorem

    Howard Barnum, Jonathan Barrett, Matthew Leifer, and Al exander Wilce. A generalized no-broadcasting theorem. Phys. Rev. Lett. , 99:240501, 2007. arXiv:0707.0620. ↑ 10

    work page Pith review arXiv 2007

  5. [5]

    On statistics independent of a complete sufficient statistic

    Debabrata Basu. On statistics independent of a complete sufficient statistic. Sankhy¯ a, 15:377–380, 1955. Reprinted at [ 6, p. 62]. ↑ 87

    1955

  6. [6]

    Selected works of Debabrata Basu

    Debabrata Basu. Selected works of Debabrata Basu. Selected Works in Probability and Statistics. Springer, New York, 2011. Edited by Anirban DasGupta. ↑ 93

    2011

  7. [7]

    Probability and measure

    Patrick Billingsley. Probability and measure . Wiley Series in Probability and Mathematical Statistics. John Wiley & Sons, Inc., New York, third edition, 1995. A Wile y-Interscience Publication. ↑ 41

    1995

  8. [8]

    Kantorovich problems and conditional measures depending on a parameter

    Vladimir I. Bogachev and Ilya I. Malofeev. Kantorovich p roblems and conditional measures depending on a parameter. J. Math. Anal. Appl. , 486(1):123883, 2020. arXiv:1904.03642. ↑ 43

    work page Pith review arXiv 2020

Show all 118 references
  1. [9]

    Carboni and R

    A. Carboni and R. F. C. Walters. Cartesian bicategories. I. J. Pure Appl. Algebra , 49(1-2):11–32, 1987. ↑ 30

    1987

  2. [10]

    Matrices, relations, and group repre sentations

    Aurelio Carboni. Matrices, relations, and group repre sentations. J. Algebra, 136(2):497–529, 1991. ↑ 29

    1991

  3. [11]

    N. N. ˇCencov. The categories of mathematical statistics. Dokl. Akad. Nauk SSSR , 164:511–514, 1965. ↑ 8, 18

    1965

  4. [12]

    N. N. ˇCencov. Algebraic foundation of mathematical statistics. Math. Operationsforsch. Statist. Ser. Statist., 9(2):267–276, 1978. ↑ 8

    1978

  5. [13]

    N. N. ˇCencov. Statistical decision rules and optimal inference , volume 53 of Translations of Mathematical Monographs. American Mathematical Society, Providence, R.I., 1982. T ranslation from the Russian edited by Lev J. Leifman. ↑ 8, 18

    1982

  6. [14]

    Probabilistic theories with purification

    Giulio Chiribella, Mauro D’Ariano, and Paolo Perinott i. Probabilistic theories with purification. Phys. Rev. A , 81:062348, 2010. arXiv:0908.1583. ↑ 54

    2010 arXiv

  7. [15]

    Disintegration and Bayesian inversion via string diagrams

    Kenta Cho and Bart Jacobs. Disintegration and Bayesian inversion via string diagrams. Math. Structures Comput. Sci. , 29:938–971, 2019. arXiv:1709.00322. ↑ 7, 9, 12, 13, 18, 19, 29, 39, 42, 53, 56, 57, 58, 64, 71, 73, 74, 75

    2019 arXiv

  8. [16]

    An introduction to effectus theory

    Kenta Cho, Bart Jacobs, Bas Westerbaan, and Abraham Wes terbaan. An introduction to effectus theory. arXiv:1512.05813. ↑ 7

    arXiv

  9. [17]

    Probability theory

    Yuan Shih Chow and Henry Teicher. Probability theory. Springer Texts in Statistics. Springer-Verlag, New York, third edition, 1997. Independence, interchangea bility, martingales. ↑ 59

    1997

  10. [18]

    Terminality implies non-signalling

    Bob Coecke. Terminality implies non-signalling. In Proceedings of the 11th Workshop on Quantum Physics and Logic , volume 172 of Electron. Proc. Theor. Comput. Sci. (EPTCS) , pages 27–35, 2014. arXiv:1405.3681. ↑ 13

    2014 arXiv

  11. [19]

    Causal categories: relativ istically interacting processes

    Bob Coecke and Raymond Lal. Causal categories: relativ istically interacting processes. Found. Phys. , 43(4):458–501, 2013. arXiv:1107.6019. ↑ 54

    2013 arXiv

  12. [20]

    Classical and qua ntum structuralism

    Bob Coecke, ´Eric Oliver Paquette, and Dusko Pavlovic. Classical and qua ntum structuralism. In Semantic techniques in quantum computation , pages 29–69. Cambridge Univ. Press, Cambridge, 2010. ↑ 30

    2010

  13. [21]

    Spekkens

    Bob Coecke and Robert W. Spekkens. Picturing classical and quantum Bayesian inference. Synthese, 186(3):651–696, 2012. arXiv:1102.2368. ↑ 8, 12, 57, 64, 71

    2012 arXiv

  14. [22]

    Purity through fact orisation

    Oscar Cunningham and Chris Heunen. Purity through fact orisation. In Proceedings 14th International Conference on Quantum Physics and Logic , volume 266 of Electron. Proc. Theor. Comput. Sci. (EPTCS) , pages 315–328. EPTCS, 2018. arXiv:1705.07652. ↑ 49

    2018 arXiv

  15. [23]

    Borel kernels and their approxi- mation, categorically

    Fredrik Dahlqvist, Alexandra Silva, Vincent Danos, an d Ilias Garnier. Borel kernels and their approxi- mation, categorically. In Proceedings of the Thirty-Fourth Conference on the Mathema tical Foundations 94 TOBIAS FRITZ of Programming Semantics (MFPS XXXIV) , volume 341 of ...

    2018 arXiv

  16. [24]

    Philip Dawid

    A. Philip Dawid. Conditional independence in statisti cal theory. J. Roy. Statist. Soc. Ser. B , 41(1):1–31,

  17. [25]

    Philip Dawid

    A. Philip Dawid. Conditional independence for statist ical operations. Ann. Statist. , 8(3):598–617, 1980. ↑ 7, 22, 84, 87

    1980

  18. [26]

    Philip Dawid

    A. Philip Dawid. Separoids: a mathematical framework f or conditional independence and irrelevance. Ann. Math. Artif. Intell. , 32(1-4):335–372, 2001. Representations of uncertainty. ↑ 7

    2001

  19. [27]

    Philip Dawid and Milan Studen´ y

    A. Philip Dawid and Milan Studen´ y. Conditional produc ts: An alternative approach to conditional independence. In Artificial Intelligence and Statistics ’99 , pages 32–40. 1999. ftp://ftp.utia.cas.cz/pub/staff/studeny/daw-stu-99-p df.pdf. ↑ 7, 58, 61

    1999

  20. [28]

    J. L. Doob. Stochastic processes. Wiley Classics Library. John Wiley & Sons, Inc., New York, 1 990. Reprint of the 1953 original, A Wiley-Interscience Publica tion. ↑ 49

    1953

  21. [29]

    Druzdzel and Herbert A

    Marek J. Druzdzel and Herbert A. Simon. Causality in bay esian belief networks. In Uncertainty in artificial intelligence (Washington, DC, 1993) , pages 3–11. Morgan Kaufmann Publishers, 1993. arXiv:1303.1454. ↑ 49

    1993 arXiv

  22. [30]

    Morris L. Eaton. Multivariate statistics, volume 53 of Institute of Mathematical Statistics Lecture Notes/emdash.cyr Monograph Series. Institute of Mathematical Statistics, Beachwood, OH, 200 7. A vector space approach, Reprint of the 1983 original. ↑ 43

    1983

  23. [31]

    G. A. Edgar. Measurable weak sections. Illinois J. Math. , 20(4):630–646, 1976. ↑ 74

    1976

  24. [32]

    Arnold M. Faden. The existence of regular conditional p robabilities: necessary and sufficient conditions. Ann. Probab., 13(1):288–298, 1985. ↑ 40

    1985

  25. [33]

    Compositories and glea ves

    Cecilia Flori and Tobias Fritz. Compositories and glea ves. Theory Appl. Categ., 31:Paper No. 33, 928–988,

  26. [34]

    Causal theories: A categorical perspect ive on Bayesian networks

    Brendan Fong. Causal theories: A categorical perspect ive on Bayesian networks. Master’s thesis, Univer- sity of Oxford, 2012. arXiv:1301.6201. ↑ 7, 9, 30

    2012 arXiv

  27. [35]

    The Algebra of Open and Interconnected Systems

    Brendan Fong. The Algebra of Open and Interconnected Systems . PhD thesis, University of Oxford, 2016. arXiv:1609.05382. ↑ 29, 33, 53

    2016 arXiv

  28. [36]

    Supplying bells and whis tles in symmetric monoidal categories

    Brendan Fong and David Spivak. Supplying bells and whis tles in symmetric monoidal categories. arXiv:1908.02633. ↑ 33, 35, 37

    1908 arXiv

  29. [37]

    Hypergraph categories

    Brendan Fong and David Spivak. Hypergraph categories. J. Pure Appl. Algebra , 223:4746–4777, 2019. arXiv:1806.08304. ↑ 29, 33, 37, 53

    2019 arXiv

  30. [38]

    What is stochastic independence? In Non-commutativity, infinite-dimensionality and prob- ability at the crossroads , volume 16 of QP–PQ: Quantum Probab

    Uwe Franz. What is stochastic independence? In Non-commutativity, infinite-dimensionality and prob- ability at the crossroads , volume 16 of QP–PQ: Quantum Probab. White Noise Anal. , pages 254–274. World Sci. Publ., River Edge, NJ, 2002. arXiv:math/0206017. ↑ 7, 13, 72

    2002 arXiv

  31. [39]

    D. H. Fremlin. Measure theory. Vol. 3 . Torres Fremlin, Colchester, 2004. Measure algebras, Corr ected second printing of the 2002 original. essex.ac.uk/maths/people/fremlin/mt.htm. ↑ 44, 74

    2004

  32. [40]

    D. H. Fremlin. Measure theory. Vol. 4. Torres Fremlin, Colchester, 2006. Topological measure sp aces. Part I, II, Corrected second printing of the 2003 original. essex.ac.uk/maths/people/fremlin/mt.htm. ↑ 41

    2006

  33. [41]

    Beyond Bell’s theorem II: scenarios with arbitrary causal structure

    Tobias Fritz. Beyond Bell’s theorem II: scenarios with arbitrary causal structure. Comm. Math. Phys. , 341(2):391–434, 2016. arXiv:1404.4812. ↑ 49

    2016 arXiv

  34. [42]

    Bimonoidal structure o f probability monads

    Tobias Fritz and Paolo Perrone. Bimonoidal structure o f probability monads. In Proceedings of the Thirty- Fourth Conference on the Mathematical Foundations of Progr amming Semantics (MFPS XXXIV) , vol- ume 341 of Electron. Notes Theor. Comput. Sci. , pages 121–149. Elsevier S...

    2018 arXiv

  35. [43]

    Pr obability, valuations, hyperspaces: three monads on Top and the support as a morphism

    Tobias Fritz, Paolo Perrone, and Sharwin Rezagholi. Pr obability, valuations, hyperspaces: three monads on Top and the support as a morphism. arXiv:1910.03752. ↑ 16

    1910 arXiv

  36. [44]

    The zero-one laws of Kolmogorov and Hewitt–Savage in cate- gorical probability

    Tobias Fritz and Eigil Fjeldgren Rischel. The zero-one laws of Kolmogorov and Hewitt–Savage in cate- gorical probability. arXiv:1912.02769. ↑ 21 A SYNTHETIC APPROACH TO MARKOV KERNELS AND STATISTICS 95

    1912 arXiv

  37. [45]

    Varieties of effects

    Carsten F¨ uhrmann. Varieties of effects. In Foundations of software science and computation structure s (Grenoble, 2002) , volume 2303 of Lecture Notes in Comput. Sci. , pages 144–158. Springer, Berlin, 2002. ↑ 30

    2002

  38. [46]

    Categorical Duality in Probability and Quantum Foundation s

    Robert Furber. Categorical Duality in Probability and Quantum Foundation s. PhD thesis, Radboud Uni- versity Nijmegen, 2017. people.cs.aau.dk/ furber/phdthesis.pdf . ↑ 23

    2017

  39. [47]

    From Kleisli categories to commutative C*-algebras: Probabilistic Gelfand duality

    Robert Furber and Bart Jacobs. From Kleisli categories to commutative C*-algebras: Probabilistic Gelfand duality. Log. Methods Comput. Sci. , 11:1–28, 2015. arXiv:1303.1115. ↑ 23

    2015 arXiv

  40. [48]

    Categorial L´ evy processes

    Malte Gerhold, Stephanie Lachs, and Michael Sch¨ urman n. Categorial L´ evy processes. arXiv:1612.05139. ↑ 7, 13, 72

    arXiv

  41. [49]

    A categorical approach to probability t heory

    Mich` ele Giry. A categorical approach to probability t heory. In Categorical aspects of topology and analysis (Ottawa, Ont., 1980) , volume 915 of Lecture Notes in Math. , pages 68–85. Springer, 1982. ↑ 7, 18, 19

    1980

  42. [50]

    P. V. Golubtsov. Relative informativity and a priori in formation in the category of linear information transformers. Problemy Peredachi Informatsii , 31(3):3–23, 1995. In Russian. English translation in Prob- lems Inform. Transmission 31(3):195–215. ↑ 43

    1995

  43. [51]

    Golubtsov

    Peter V. Golubtsov. Informativity in the category of li near measurement systems. Problemy Peredachi In- formatsii, 28(2):30–46, 1992. In Russian. English translation in Problems Inform Transmission 28(2):125– 140 (1992). ↑ 25

    1992

  44. [52]

    Golubtsov

    Peter V. Golubtsov. Theory of fuzzy sets as a theory of un certainty and decision-making problems in fuzzy experiments. Problemy Peredachi Informatsii , 30(3):47–67, 1994. In Russian. English translation in Problems Inform. Transmission 30(3):232–250. ↑ 39, 42

    1994

  45. [53]

    Golubtsov

    Peter V. Golubtsov. Informativity in the category of mu ltivalued information transformers. Problemy Peredachi Informatsii, 34(3):60–80, 1998. In Russian. English translation in Problems Inform. Transmis- sion 34(3):259–276 (1998). ↑ 15

    1998

  46. [54]

    Golubtsov

    Peter V. Golubtsov. Axiomatic description of categori es of information converters. Problemy Peredachi In- formatsii, 35(3):80–98, 1999. In Russian. English translation in Problems Inform. Transmission 35(3):259– 274, 1999. ↑ 7, 84

    1999

  47. [55]

    Golubtsov

    Peter V. Golubtsov. Monoidal Kleisli category as a back ground for information transformers theory. РҶР„СҺР/guillemotleft.cyrСҐРёРөСҘРҷР/guillemotleft.cyrЄЄСҜРҳ Р/guillemotright.cyrСҐР/guillemotleft.cyrСҘРҳСҒСҒСҜ (Information Theory and Information Processing), 2:62–84, 2002...

    2002

  48. [56]

    Golubtsov

    Peter V. Golubtsov. Information transformers: catego ry-theoretical structure, informativeness, decision- making problems. Hadronic J. Suppl. , 19(3):375–424, 2004. ↑ 7

    2004

  49. [57]

    Golubtsov and Stepan S

    Peter V. Golubtsov and Stepan S. Moskaliuk. Method of ad ditional structures on the objects of a monoidal Kleisli category as a background for information transform ers theory. Hadronic J. , 25(2):179–237, 2002. arXiv:math-ph/0211067. ↑ 7

    2002 arXiv

  50. [58]

    Algebras of th e extended probabilistic powerdomain monad

    Jean Goubault-Larrecq and Xiaodong Jia. Algebras of th e extended probabilistic powerdomain monad. arXiv:1903.07472. ↑ 16

    1903 arXiv

  51. [59]

    Sur quelques points d’alg` eb re homologique

    Alexander Grothendieck. Sur quelques points d’alg` eb re homologique. Tˆ ohoku Math. J. (2), 9:119–221,

  52. [60]

    Tenseurs et machines

    Ren´ e Guitart. Tenseurs et machines. Cahiers Topologie G´ eom. Diff´ erentielle, 21(1):5–62, 1980. ↑ 17

    1980

  53. [61]

    Halmos and L

    Paul R. Halmos and L. J. Savage. Application of the Radon -Nikodym theorem to the theory of sufficient statistics. Ann. Math. Statistics , 20:225–241, 1949. ↑ 87

    1949

  54. [62]

    Spaces of valuations

    Reinhold Heckmann. Spaces of valuations. In Papers on general topology and applications (Gorham, ME, 1995), volume 806 of Ann. New York Acad. Sci. , pages 174–200. New York Acad. Sci., New York, 1996. ↑ 16

    1995

  55. [63]

    A convenient category for higher-order probability theory

    Chris Heunen, Ohad Kammar, Sam Staton, and Hongseok Yan g. A convenient category for higher-order probability theory. In 2017 32nd Annual ACM/IEEE Symposium on Logic in Computer Sci ence (LICS), page 12. IEEE, [Piscataway], NJ, 2017. arXiv:1701.02547. ↑ 7

    2017 arXiv

  56. [64]

    Categories for Quantum Theory: An Introduction

    Chris Heunen and Jamie Vicary. Categories for Quantum Theory: An Introduction . Oxford University Press, 2020. ↑ 35 96 TOBIAS FRITZ

    2020

  57. [65]

    Symmetric monoidal sketc hes

    Martin Hyland and John Power. Symmetric monoidal sketc hes. In Second International Conference on Principles and Practice of Declarative Progr amming, pages 280–288. 2000. https://webdpmms.maths.cam.ac.uk/∼ martin/Research/Pub91-00/hp00.pdf. ↑ 38

    2000

  58. [66]

    Semantics of weakening and contraction

    Bart Jacobs. Semantics of weakening and contraction. Ann. Pure Appl. Logic , 69(1):73–106, 1994. ↑ 13

    1994

  59. [67]

    Structured probabilistic reasoning, 201 9

    Bart Jacobs. Structured probabilistic reasoning, 201 9. Forthcoming book, available at cs.ru.nl/B.Jacobs/PAPERS/ProbabilisticReasoning.pdf. ↑ 7

  60. [68]

    Categoric al semantics for arrows

    Bart Jacobs, Chris Heunen, and Ichiro Hasuo. Categoric al semantics for arrows. J. Funct. Programming, 19(3-4):403–438, 2009. ↑ 38

    2009

  61. [69]

    A predicate/state transf ormer semantics for Bayesian learning

    Bart Jacobs and Fabio Zanasi. A predicate/state transf ormer semantics for Bayesian learning. In The Thirty-second Conference on the Mathematical Foundations of Programming Semantics (MFPS XXXII) , volume 325 of Electron. Notes Theor. Comput. Sci. , pages 185–200. Elsevier Sci...

    2016

  62. [70]

    Probabilistic mappings and Bayesian nonparametrics

    J¨ urgen Jost, Hˆ ong Vˆ an Lˆ e, Duc Hoang Luu, and Tat Dat T ran. Probabilistic mappings and Bayesian nonparametrics. arXiv:1905.11448. ↑ 18, 49, 83, 85

    1905 arXiv

  63. [71]

    The geometry of tensor cal culus

    Andr´ e Joyal and Ross Street. The geometry of tensor cal culus. I. Adv. Math. , 88(1):55–112, 1991. ↑ 38

    1991

  64. [72]

    Ergodic theory of random transformations , volume 10 of Progress in Probability and Statistics

    Yuri Kifer. Ergodic theory of random transformations , volume 10 of Progress in Probability and Statistics . Birkh¨ auser Boston, Inc., Boston, MA, 1986. ↑ 49

    1986

  65. [73]

    Finite matrices are complete for (dag ger-)hypergraph categories

    Aleks Kissinger. Finite matrices are complete for (dag ger-)hypergraph categories. arXiv:1406.5942. ↑ 29, 33, 53

    arXiv

  66. [74]

    Equivale nce of relativistic causal structure and process terminality

    Aleks Kissinger, Matty Hoban, and Bob Coecke. Equivale nce of relativistic causal structure and process terminality. arXiv:1708.04118. ↑ 13

    arXiv

  67. [75]

    H. Kleisli. Every standard construction is induced by a pair of adjoint functors. Proc. Amer. Math. Soc. , 16:544–546, 1965. ↑ 18

    1965

  68. [76]

    Probability theory

    Achim Klenke. Probability theory. Universitext. Springer, London, second edition, 2014. A c omprehensive course. ↑ 8, 40, 41

    2014

  69. [77]

    Lauritzen

    Steffen L. Lauritzen. Extremal families and systems of sufficient statistics , volume 49 of Lecture Notes in Statistics. Springer-Verlag, New York, 1988. ↑ 28, 83, 92

    1988

  70. [78]

    Lauritzen

    Steffen L. Lauritzen. Graphical models, volume 17 of Oxford Statistical Science Series . The Clarendon Press, Oxford University Press, New York, 1996. Oxford Scie nce Publications. ↑ 56, 72

    1996

  71. [79]

    Lauritzen and Frank Jensen

    Steffen L. Lauritzen and Frank Jensen. Stable local comp utation with conditional Gaussian distributions. Stat. Comput. , 11(2):191–203, 2001. ↑ 7, 25, 26, 43

    2001

  72. [80]

    The category of probabilistic mappings, 1962

    Bill Lawvere. The category of probabilistic mappings, 1962. ncatlab.org/nlab/files/lawvereprobability1962.pdf . ↑ 2, 7, 18

    1962

  73. [81]

    E. L. Lehmann. Selected works of E. L. Lehmann . Selected Works in Probability and Statistics. Springer, New York, 2012. Edited by Javier Rojo. ↑ 96

    2012

  74. [82]

    E. L. Lehmann and Henry Scheff´ e. Completeness, similar regions, and unbiased estimation. I. Sankhy¯ a, 10:305–340, 1950. Reprinted at [ 81, p. 239]. ↑ 87

    1950

  75. [83]

    Monoidal categories with projections, 2 016

    Tom Leinster. Monoidal categories with projections, 2 016. n-Category Caf´ e blog post. golem.ph.utexas.edu/category/2016/08/monoidal_categories_with_proje.html . ↑ 13

    2016

  76. [84]

    Monoidal categories with projections, 2 016

    Tom Leinster. Monoidal categories with projections, 2 016. n-Category Caf´ e comment. golem.ph.utexas.edu/category/2016/08/monoidal_categories_with_proje.html#c051133 . ↑ 13

    2016

  77. [85]

    Modell ing environments in call-by-value programming languages

    Paul Blain Levy, John Power, and Hayo Thielecke. Modell ing environments in call-by-value programming languages. Inform. and Comput. , 185(2):182–210, 2003. ↑ 38

    2003

  78. [86]

    Rory B. B. Lucyshyn-Wright. Functional distribution m onads in functional-analytic contexts. Adv. Math., 322:806–860, 2017. arXiv:1701.08152. ↑ 7

    2017 arXiv

  79. [87]

    Categories for the working mathematician , volume 5 of Graduate Texts in Mathe- matics

    Saunders Mac Lane. Categories for the working mathematician , volume 5 of Graduate Texts in Mathe- matics. Springer-Verlag, New York, second edition, 1998. ↑ 36

    1998

  80. [88]

    Sheaves in geometry and logic

    Saunders Mac Lane and Ieke Moerdijk. Sheaves in geometry and logic . Universitext. Springer-Verlag, New York, 1994. A first introduction to topos theory, Correct ed reprint of the 1992 edition. ↑ 91

    1994

  81. [89]

    Ernest G. Manes. A class of fuzzy theories. J. Math. Anal. Appl. , 85(2):409–451, 1982. ↑ 58 A SYNTHETIC APPROACH TO MARKOV KERNELS AND STATISTICS 97

    1982

  82. [90]

    Sofia Massa and Steffen L

    M. Sofia Massa and Steffen L. Lauritzen. Combining statis tical models. In Algebraic methods in statistics and probability II , volume 516 of Contemp. Math. , pages 239–259. Amer. Math. Soc., Providence, RI,

  83. [91]

    What is a statistical model? Ann

    Peter McCullagh. What is a statistical model? Ann. Statist. , 30(5):1225–1310, 2002. With comments and a rejoinder by the author. ↑ 7, 83

    2002

  84. [92]

    E. A. Morozova and N. N. ˇCencov. Markov invariant geometry on state manifolds. In Current problems in mathematics. Newest results, Vol. 36 (Russian) , Itogi Nauki i Tekhniki, pages 69–102, 187. Akad. Nauk SSSR, Vsesoyuz. Inst. Nauchn. i Tekhn. Inform., Moscow , 1989. Translat...

    1989

  85. [93]

    Statistical isom orphism

    Norman Morse and Richard Sacksteder. Statistical isom orphism. Ann. Math. Statist. , 37:203–214, 1966. ↑ 18

    1966

  86. [94]

    Bases math´ ematiques du calcul des probabilit´ es

    Jacques Neveu. Bases math´ ematiques du calcul des probabilit´ es. Pr´ eface de R. Fortet. Deuxi` eme ´ edition, revue et corrig´ ee. Masson et Cie, ´Editeurs, Paris, 1970. ↑ 8

    1970

  87. [95]

    Labelled Markov processes

    Prakash Panangaden. Labelled Markov processes. Imperial College Press, London, 2009. ↑ 18, 19

    2009

  88. [96]

    Causality

    Judea Pearl. Causality. Cambridge University Press, Cambridge, second edition, 2 009. Models, reasoning, and inference. ↑ 72

  89. [97]

    Graphoids: a graph-based lo gic for reasoning about relevance relations,

    Judea Pearl and Azaria Paz. Graphoids: a graph-based lo gic for reasoning about relevance relations,

  90. [98]

    Categorical Probability and Stochastic Dominance in Metri c Spaces

    Paolo Perrone. Categorical Probability and Stochastic Dominance in Metri c Spaces . PhD thesis, Max Planck Institute for Mathematics in the Sciences, 2018. paoloperrone.org/phdthesis.pdf. ↑ 7

    2018

  91. [99]

    Caract´ erisation des cat´ egories ab´ eliennes avec g´ en´ erateurs et limites inductives exactes

    Nicolae Popesco and Pierre Gabriel. Caract´ erisation des cat´ egories ab´ eliennes avec g´ en´ erateurs et limites inductives exactes. C. R. Acad. Sci. Paris , 258:4188–4190, 1964. ↑ 3

    1964

  92. [100]

    K. P. S. Bhaskara Rao and B. V. Rao. Borel spaces. Dissertationes Math. (Rozprawy Mat.) , 190:63, 1981. eudml.org/doc/268562. ↑ 31

    1981

  93. [101]

    Cat´ egories Tannakiennes

    Neantro Saavedra Rivano. Cat´ egories Tannakiennes. Lecture Notes in Mathematics, Vol. 265. Springer- Verlag, Berlin-New York, 1972. ↑ 36

    1972

  94. [102]

    A note on statistical equivalenc e

    Richard Sacksteder. A note on statistical equivalenc e. Ann. Math. Statist. , 38:787–794, 1967. ↑ 21

    1967

  95. [103]

    Moss, Chris Heunen, and Zoubin Ghahramani

    Adam ´Scibior, Ohad Kammar, Matthijs V´ ak´ ar, Sam Staton, Hongseok Yang, Yufei Cai, Klaus Ostermann, Sean K. Moss, Chris Heunen, and Zoubin Ghahramani. Denotati onal validation of higher-order Bayesian inference. In Proc. ACM Program. Lang. , volume 2, pages 60:1–29. ACM, 20...

    2017 arXiv

  96. [104]

    Gavin J. Seal. Tensors, monads and actions. Theory Appl. Categ. , 28:No. 15, 403–433, 2013. arXiv:1205.0101. ↑ 17

    2013 arXiv

  97. [105]

    Selinger

    P. Selinger. A survey of graphical languages for monoi dal categories. In New Structures for Physics , volume 813 of Lecture Notes in Phys. , pages 289–355. Springer, Heidelberg, 2011. ↑ 8

    2011

  98. [106]

    Framed bicategories and monoidal fib rations

    Michael Shulman. Framed bicategories and monoidal fib rations. Theory Appl. Categ. , 20:Paper No. 18, 650–738, 2008. arXiv:0706.1286. ↑ 15

    2008 arXiv

  99. [107]

    Probability sheaves and the Giry monad, 2017

    Alex Simpson. Probability sheaves and the Giry monad, 2017. http://drops.dagstuhl.de/opus/volltexte/2017/8051/pdf/LIPIcs-CALCO-2017-1.pdf. ↑ 7

    2017

  100. [108]

    Category-theoretic structure for inde pendence and conditional independence

    Alex Simpson. Category-theoretic structure for inde pendence and conditional independence. In The Thirty-third Conference on the Mathematical Foundations o f Programming Semantics (MFPS XXXIII) , volume 336 of Electron. Notes Theor. Comput. Sci. , pages 281–297. Elsevier Sci. ...

    2018

  101. [109]

    A ckerman, Cameron E

    Sam Staton, Dario Stein, Hongseok Yang, Nathanael L. A ckerman, Cameron E. Freer, and Daniel M. Roy. The beta-Bernoulli process and algebraic effects. In Ioanni s Chatzigiannakis, Christos Kaklamanis, D´ aniel Marx, and Donald Sannella, editors, 45th International Colloquium on...

    2018 arXiv

  102. [110]

    Multi-information and the problem of characterization of conditional independence relations

    Milan Studen´ y. Multi-information and the problem of characterization of conditional independence relations. Problems Control Inform. Theory/Problemy Upravlen. Teor. Inform., 18(1):3–16, 1989. ftp.utia.cas.cz/pub/staff/studeny/multiinformation-P CIT-89.pdf. ↑ 61 98 TOBIAS FRITZ

    1989

  103. [111]

    Probabilistic conditional independence structures

    Milan Studen´ y. Probabilistic conditional independence structures . Information Science and Statistics. Springer, London, 2005. ↑ 57, 58

    2005

  104. [112]

    Monadic functors and convexity

    Tadeusz ´Swirszcz. Monadic functors and convexity. Bull. Acad. Polon. Sci. S´ er. Sci. Math. Astronom. Phys., 22:39–42, 1974. ↑ 23

    1974

  105. [113]

    Categorical Structure of Continuation Passing Style

    Hayo Thielecke. Categorical Structure of Continuation Passing Style . PhD thesis, University of Edinburgh,

  106. [114]

    Robert M. Wald. General relativity. University of Chicago Press, Chicago, IL, 1984. ↑ 16 Perimeter Institute for Theoretical Physics, W a terloo, On tario, Canada E-mail address : tfritz@pitp.ca

    1984

  107. [1985]

    ftp.cs.ucla.edu/pub/stat_ser/r53-L.pdf

    Technical report. ftp.cs.ucla.edu/pub/stat_ser/r53-L.pdf. ↑ 57

  108. [1997]

    https://www.cs.bham.ac.uk/∼ hxt/research/THESIS-FINAL.ps. ↑ 30

  109. [2010]

    web.math.ku.dk/∼ lauritzen/papers/conm10179.pdf. ↑ 7

  110. [2016]

    ↑ 61, 62, 63

    arXiv:1308.6548. ↑ 61, 62, 63

    arXiv

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.