pith. sign in

arxiv: 2605.03902 · v3 · pith:JSPKBDJRnew · submitted 2026-05-05 · 🧮 math.PR

Bundles of Probability Schemes

Wai Yan Pong This is my paper

Pith reviewed 2026-05-20 23:27 UTC · model grok-4.3

classification 🧮 math.PR
keywords finite probabilitybundlesconditional expectationcategorical probabilityMarkov chainslinear regressionfiber products
0
0 comments X

The pith

Bundles of finite probability schemes yield conditional expectation through two linear functors.

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

The paper defines bundles as objects in the category of finite probability schemes and probability-preserving maps. Each bundle simultaneously encodes a quotient of the sample space, an algebra of random variables, and the family of conditional schemes over the quotient. Two natural linear functors on a bundle directly construct the conditional expectation operator and account for why it behaves as a projection. Standard results including the laws of total expectation, variance, and covariance, the weak law of large numbers, and variance decomposition in linear regression are recovered inside this structure, while fiber products encode conditional independence and discrete-time Markov chains.

Core claim

A bundle records a quotient of the sample space, an algebra of random variables, and the family of conditional schemes over the quotient. The two associated linear functors construct conditional expectation and explain its projection properties, from which the laws of total expectation, variance, and covariance, the weak law of large numbers, and the variance decomposition in simple linear regression follow, with fiber products encoding conditional independence and discrete-time Markov chains.

What carries the argument

Bundle: an object that simultaneously records a quotient of a sample space, an algebra of random variables, and the family of conditional schemes over the quotient.

If this is right

  • Conditional expectation is obtained directly as the output of the two linear functors without separate definition.
  • The law of total expectation and the variance decomposition in linear regression follow immediately from the functor construction.
  • Fiber products in the category encode conditional independence relations.
  • Discrete-time Markov chains arise naturally from successive fiber products of bundles.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same functor-based construction might be tested on small finite examples to verify recovery of the weak law of large numbers without measure-theoretic tools.
  • Fiber-product encoding of Markov chains suggests the framework could organize chains of conditional schemes in a single diagram.

Load-bearing premise

The category of finite probability schemes equipped with probability-preserving maps is expressive enough to encode all relevant quotients, random variable algebras, and conditional schemes needed to recover the standard laws and relations.

What would settle it

A concrete probability law such as the law of total expectation that cannot be obtained as a direct consequence of the two linear functors on any bundle would show the construction does not fully capture the required structure.

read the original abstract

We study finite probability theory through a category of finite probability schemes and probability-preserving maps, called \emph{bundles}. A bundle simultaneously records a quotient of a sample space, an algebra of random variables, and the family of conditional schemes over the quotient. The two natural linear functors associated with a bundle give a compact construction of conditional expectation and explain its projection properties. Within this framework we recover the laws of total expectation, variance, and covariance, the weak law of large numbers, and the variance decomposition behind simple linear regression. Fiber products then encode conditional independence and discrete-time Markov chains.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper introduces a category of finite probability schemes equipped with probability-preserving maps, termed bundles. Each bundle encodes a quotient of the sample space, an algebra of random variables, and the family of conditional schemes over the quotient. The two natural linear functors on a bundle are used to construct conditional expectation and recover its projection properties. The framework is then applied to recover the laws of total expectation, total variance, and total covariance, the weak law of large numbers, and the variance decomposition in simple linear regression; fiber products are used to model conditional independence and discrete-time Markov chains.

Significance. If the central constructions hold, the work supplies a compact categorical account of conditional expectation and several classical limit and decomposition laws in finite probability. The functorial treatment of projection properties and the use of fiber products for independence relations could streamline proofs and clarify structural relationships that are usually treated separately.

major comments (2)
  1. [Definition of bundles and the two natural linear functors] The central claim that bundles simultaneously encode quotients, random-variable algebras, and all conditional schemes rests on the morphisms being restricted to probability-preserving maps. This restriction appears insufficient to capture conditional schemes in which random variables are not constant on fibers or when non-surjective maps are required, which directly affects the recovery of general conditional independence and the Markov property for chains.
  2. [Recovery of laws of total expectation, variance, and covariance] The recovery of the law of total covariance and the variance decomposition in linear regression is asserted via the bundle functors, yet the manuscript does not explicitly verify that the constructions remain valid when some atoms have zero probability; this is load-bearing for the claim that the framework recovers the classical laws without post-hoc restrictions.
minor comments (2)
  1. [Section introducing the functors] Notation for the two linear functors is introduced without a clear diagram or explicit matrix representation in the finite case, making it harder to verify the claimed projection properties by direct computation.
  2. [Fiber products and conditional independence] The abstract states that fiber products encode conditional independence, but the manuscript should include a short table comparing the categorical definition with the usual probabilistic definition to confirm equivalence on finite spaces.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their careful reading of the manuscript and for the positive assessment of its potential significance. We respond to each major comment below, indicating where we agree and where we maintain the original approach while offering clarifications.

read point-by-point responses

  1. Referee: The central claim that bundles simultaneously encode quotients, random-variable algebras, and all conditional schemes rests on the morphisms being restricted to probability-preserving maps. This restriction appears insufficient to capture conditional schemes in which random variables are not constant on fibers or when non-surjective maps are required, which directly affects the recovery of general conditional independence and the Markov property for chains.

    Authors: We maintain that the restriction to probability-preserving maps is the appropriate choice of morphisms and suffices for the claimed encodings. In the definition of a bundle, the quotient map induces an algebra consisting precisely of those random variables that are constant on the fibers; this is the standard correspondence between quotients and generated sigma-algebras in finite spaces. Conditional schemes are obtained by restricting the measure to each fiber of the quotient. Probability-preserving maps between bundles are required to commute with the quotients and to push measures forward correctly, which ensures that fiber-product constructions correctly capture conditional independence and the Markov property for discrete-time chains as developed in the applications. Non-surjective maps are admissible in the category; the image carries the pushed-forward measure, and all subsequent constructions (conditional expectation, fiber products) are performed relative to that image. We will add a short clarifying paragraph after the definition of bundle morphisms to emphasize these points. revision: partial

  2. Referee: The recovery of the law of total covariance and the variance decomposition in linear regression is asserted via the bundle functors, yet the manuscript does not explicitly verify that the constructions remain valid when some atoms have zero probability; this is load-bearing for the claim that the framework recovers the classical laws without post-hoc restrictions.

    Authors: The referee correctly notes that the manuscript does not contain an explicit verification for the zero-probability case. While the classical identities continue to hold formally when zero-probability atoms are simply omitted (as they contribute nothing to the relevant sums), we agree that an explicit check strengthens the claim that the functorial constructions recover the laws directly. In the revised manuscript we will insert a brief remark (or short appendix paragraph) confirming that the two linear functors and the resulting covariance and regression decompositions remain valid when some atoms receive zero mass, either by direct substitution into the finite sums or by observing that the expressions are continuous in the probability weights. revision: yes

Circularity Check

0 steps flagged

No circularity: new categorical framework recovers known laws via explicit constructions

full rationale

The paper introduces bundles as a category of finite probability schemes with probability-preserving maps, then uses two natural linear functors to construct conditional expectation and its projection properties directly from the definitions. It recovers the laws of total expectation, variance, covariance, the weak law of large numbers, variance decomposition in regression, and conditional independence via fiber products through explicit categorical constructions. These steps build from the bundle structure to the target results without reducing any equation to a fitted parameter, self-citation chain, or input by definition; the recoveries serve as consistency checks rather than tautological rederivations. The framework is self-contained against external benchmarks because the recovered statements match independently known probability theory outside the paper's fitted values or prior author work.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

The paper introduces the bundle as a new defined object rather than deriving it from prior structures. It relies on the standard axioms of category theory and the domain assumption that all sample spaces are finite. No numerical free parameters appear; the only invented entity is the bundle itself, which has no independent empirical test outside the framework.

axioms (2)
  • standard math Standard axioms of category theory (objects, morphisms, composition, identities).
    The paper constructs a category of probability schemes and maps, invoking the usual category axioms.
  • domain assumption All sample spaces under consideration are finite.
    The entire development is restricted to finite probability theory as stated in the abstract.
invented entities (1)
  • Bundle no independent evidence
    purpose: A single structure that records a quotient of the sample space, an algebra of random variables, and the family of conditional schemes over the quotient.
    This is the central new mathematical object defined in the paper.

pith-pipeline@v0.9.0 · 5604 in / 1560 out tokens · 70288 ms · 2026-05-20T23:27:33.679736+00:00 · methodology

Review history (2 revisions) →

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

21 extracted references · 21 canonical work pages

  1. [1]

    Nelson, Edward , TITLE =. Bull. Amer. Math. Soc. , FJOURNAL =. 1977 , NUMBER =. doi:10.1090/S0002-9904-1977-14398-X , URL =

    work page doi:10.1090/s0002-9904-1977-14398-x 1977

  2. [2]

    1987 , PAGES =

    Nelson, Edward , TITLE =. 1987 , PAGES =. doi:10.1515/9781400882144 , URL =

    work page doi:10.1515/9781400882144 1987

  3. [3]

    and McLeish, D

    Small, Christopher G. and McLeish, D. L. , TITLE =. 1994 , PAGES =. doi:10.1002/9781118165522 , URL =

    work page doi:10.1002/9781118165522 1994

  4. [4]

    , TITLE =

    Sternberg, Shlomo Z. , TITLE =. 2014 , URL =

    work page 2014

  5. [5]

    Hilbert space methods in classical problems of mathematical statistics , BOOKTITLE =

    Gerle. Hilbert space methods in classical problems of mathematical statistics , BOOKTITLE =. 1976 , MRCLASS =

    work page 1976

  6. [6]

    1995 , PAGES =

    Malliavin, Paul , TITLE =. 1995 , PAGES =. doi:10.1007/978-1-4612-4202-4 , URL =

    work page doi:10.1007/978-1-4612-4202-4 1995

  7. [7]

    , TITLE =

    Geyer, Charles J. , TITLE =. 2007 , URL =

    work page 2007

  8. [8]

    2004 , PAGES =

    Berlinet, Alain and Thomas-Agnan, Christine , TITLE =. 2004 , PAGES =. doi:10.1007/978-1-4419-9096-9 , URL =

    work page doi:10.1007/978-1-4419-9096-9 2004

  9. [9]

    Jerome , TITLE =

    Fajardo, Sergio and Keisler, H. Jerome , TITLE =. 2002 , PAGES =

    work page 2002

  10. [10]

    Jerome , TITLE =

    Keisler, H. Jerome , TITLE =. Nonstandard analysis and its applications (. 1988 , ISBN =

    work page 1988

  11. [11]

    Nonstandard analysis for the working mathematician , PAGES =

    Osswald, Horst , TITLE =. Nonstandard analysis for the working mathematician , PAGES =. 2015 , ISBN =

    work page 2015

  12. [12]

    Nonstandard analysis for the working mathematician , PAGES =

    Sun, Yeneng , TITLE =. Nonstandard analysis for the working mathematician , PAGES =. 2015 , ISBN =

    work page 2015

  13. [13]

    2015 , PAGES =

    Nonstandard analysis for the working mathematician , EDITOR =. 2015 , PAGES =. doi:10.1007/978-94-017-7327-0 , URL =

    work page doi:10.1007/978-94-017-7327-0 2015

  14. [14]

    1995 , PAGES =

    Nonstandard analysis in practice , SERIES =. 1995 , PAGES =. doi:10.1007/978-3-642-57758-1 , URL =

    work page doi:10.1007/978-3-642-57758-1 1995

  15. [15]

    van den Berg, I. P. , TITLE =. Nonstandard analysis in practice , SERIES =. 1995 , ISBN =

    work page 1995

  16. [16]

    Nonstandard analysis in practice , SERIES =

    Beno\^it, Eric , TITLE =. Nonstandard analysis in practice , SERIES =. 1995 , ISBN =

    work page 1995

  17. [17]

    Georgian Math

    Babilua, Petre and Nadaraya, Elizbar and Sokhadze, Grigol , TITLE =. Georgian Math. J. , FJOURNAL =. 2015 , NUMBER =. doi:10.1515/gmj-2015-0006 , URL =

    work page doi:10.1515/gmj-2015-0006 2015

  18. [18]

    A Hilbert Space Embedding for Distributions

    Smola, Alex and Gretton, Arthur and Song, Le and Sch \"o lkopf, Bernhard. A Hilbert Space Embedding for Distributions. Algorithmic Learning Theory. 2007

    work page 2007

  19. [19]

    2016 , PAGES =

    Riehl, Emily , TITLE =. 2016 , PAGES =

    work page 2016

  20. [20]

    Adachi, Takanori and Ryu, Yoshihiro , TITLE =. J. Math. Sci. Univ. Tokyo , FJOURNAL =. 2019 , NUMBER =

    work page 2019

  21. [21]

    Proceedings of the 12th International Workshop on Quantum Physics and Logic , SERIES =

    Furber, Robert and Jacobs, Bart , TITLE =. Proceedings of the 12th International Workshop on Quantum Physics and Logic , SERIES =. 2015 , DOI =

    work page 2015