Pith. sign in

REVIEW 5 major objections 5 minor 27 references

Towards a category-theoretic foundation of Classical and Quantum Information Geometry

T0 review · 5 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read This paper proposes a common categorical stage for classical and quantum statistical models and argues that classifying its covariance functors subsumes both Čencov's and Petz's uniqueness theorems.

desk verdict A clear, honest research announcement whose two headline theorems are both explicitly deferred; worth reading for the framework, but not yet a proof-carrying paper. read the letter →

arxiv 2509.10262 v1 pith:26ZJUJLM submitted 2025-09-12 math-ph math.MPquant-ph

Florio M. Ciaglia , Fabio Di Cosmo , Laura González-Bravo This is my paper · ORCID
classification math-phmath.MPquant-ph MSC 53B1246L3046L1018B99
keywords informationgeometryW*-algebrasnormalstatescompletelypositivemapsGNSconstructionfieldsofcovariancesCencovtheoremPetzclassification
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

The paper is a programmatic proposal: it builds a single category, NCP, whose objects are pairs (A, ρ) consisting of a W*-algebra and a normal state, and whose morphisms are state-preserving unital completely positive maps, so that classical probability spaces and quantum state spaces live in the same structure. Its central claim is that the right invariant-geometry problem in this setting is the classification of fields of covariances, defined as contravariant functors from NCP to Hilbert spaces that share the underlying vector space of the GNS Hilbert space at every object. The authors state that this classification entails both Čencov's uniqueness of the Fisher-Rao metric tensor and Petz's classification of monotone quantum metric tensors as particular cases, and that on tracial states over finite-dimensional algebras the GNS functor is the only such functor up to two positive constants. They also show how statistical models, including Lie-group transformation models and coherent states, embed as subcategories of NCP, so that a field of covariance can be pulled back to a Riemannian metric on a model, recovering Fisher-Rao, Fubini-Study, and Bures-Helstrom metrics in special cases. The proofs are deferred to forthcoming works; the present contribution is the categorical reformulation that makes the unification conceivable.

What carries the argument

The load-bearing object is the category NCP together with the notion of a field of covariances: a contravariant functor C: NCP → Hilb whose underlying vector spaces coincide with those of the GNS functor G through the condition F∘C = F∘G. The GNS construction supplies the reference Hilbert space H_ρ, obtained by completing the algebra modulo the left ideal of elements with zero ρ-norm, with inner product ⟨x|y⟩_ρ = ρ(x*y), and the GNS functor sends each state to this space and each CPU map to the induced contraction. A field of covariances keeps the same underlying spaces and maps but may replace the inner product, and the inequality imposed by functoriality is precisely the monotonicity property used in Petz's classification. Statistical models enter as subcategories of NCP, built for instance from action groupoids for Lie-group models such as univariate normal distributions, and a field of covariance then pulls back to a Riemannian metric tensor on the model's manifold.

What would settle it

Take the two-by-two matrix algebra with a faithful state and check whether every Petz monotone metric can be written as the pullback of a field of covariances satisfying F∘C = F∘G; if any operator monotone function yields a metric that no such functor can produce, the announced subsumption of Petz's classification fails. Equivalently, on the tracial subcategory fNCT, finding a field of covariances that is not a positive constant times the GNS inner product would disprove the claimed uniqueness.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that a single functorial object, a field of covariances, can carry the information-geometric content of both classical and quantum statistics. A field of covariances is a contravariant functor C: NCP → Hilb satisfying F∘C = F∘G, where G is the GNS functor; concretely, at each state ρ the functor chooses a new inner product C_ρ on the GNS Hilbert space H_ρ, and on every state-preserving CPU map it acts by the same contraction as G. Functoriality then enforces the monotonicity inequality that defines Petz's quantum metrics, while the GNS inner product reduces to classical statistical covariance on commutative algebras, whose inverse is the Fisher-Rao metric tensor. The paper asserts, with proofs reserved for later work, that classifying these functors recovers Čencov's and Petz's classifications, that on the tracial finite-dimensional subcategory the GNS functor is unique up to two positive constants, and that pullback of fields of covariances along statistical subcategories yields the Fisher-Rao, Bures-Helstrom, and Fubini-Study metrics in the appropriate limits.

Load-bearing premise

The argument depends on the stipulated condition that every admissible geometry is a field of covariances, a functor that reuses the GNS Hilbert space at every state and varies only its inner product; if some legitimate monotone metric escapes this representation or the condition admits non-geometric functors, the announced reduction of Čencov's and Petz's theorems to one classification problem collapses.

Editorial extensions

If this is right

  • A classification of fields of covariances on NCP would simultaneously recover Čencov's uniqueness of the Fisher-Rao metric and Petz's list of monotone quantum metrics, showing that the two theorems are two faces of one functorial statement.
  • Because NCP is built from W*-algebras and normal states, the same formalism would cover infinite-dimensional algebras and non-faithful states, the two cases that Čencov's and Petz's original theorems leave out.
  • On tracial states over finite-dimensional algebras, the claimed uniqueness of the GNS functor up to two positive constants would give a non-commutative Čencov theorem for states that do not see non-commutativity.
  • Statistical models realized as subcategories of NCP would inherit their geometry from the ambient functor: pullback of the GNS field of covariances gives the Fisher-Rao metric on commutative models, the Bures-Helstrom metric on faithful finite-dimensional quantum states, and the Fubini-Study metric on pure states.
  • Monotonicity under quantum channels becomes functoriality, so the same categorical condition enforces both Čencov's congruent-embedding invariance and Petz's contraction requirement.

Reading between the lines

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

  • A natural next test is to compute, for the two-by-two matrix algebra, whether every Petz monotone metric arises from a field of covariances or only a subfamily; the paper does not give the correspondence between operator monotone functions and specific functors.
  • The uniqueness statement on tracial states suggests a possible hierarchy in which the space of admissible fields of covariances grows as non-commutativity increases; this is an extrapolation, not a paper claim.
  • One could turn the pullback construction into an inverse problem: given a Riemannian metric on a statistical model, decide whether it is the pullback of some field of covariances, which would give a new admissibility criterion beyond monotonicity.
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

5 major / 5 minor

Summary. The manuscript introduces a category NCP whose objects are pairs (A, ρ) of a W*-algebra and a normal state, and whose morphisms are state-preserving unital completely positive maps. It defines 'fields of covariances' as contravariant functors C: NCP → Hilb satisfying the equality F∘C = F∘G with the GNS functor G (Eq. (1)), and claims that classifying such functors entails both Cencov's uniqueness of the Fisher-Rao metric and Petz's classification of monotone quantum metrics. It further claims that on the full subcategory fNCT of finite-dimensional tracial states the GNS functor is the only possible functor up to two positive constants. Section 3 constructs statistical subcategories of NCP, including an action groupoid for univariate normal distributions, and states that a field of covariances will induce Riemannian metrics on such subcategories, recovering Fisher-Rao and Bures-Helstrom metrics. All three main claims are explicitly deferred to forthcoming works, and no proofs are contained in the manuscript.

Significance. If the announced results were proved, the framework would offer a genuine unification: a single classification problem over W*-algebras and normal states would subsume the classical Cencov theorem and the quantum Petz classification, extend to infinite dimensions and non-faithful states, and make the geometry of statistical models functorial. The paper is clearly written and the definition of NCP together with the GNS functor provides a plausible categorical setup; the Gaussian action-groupoid example is a useful illustration. However, the manuscript as submitted is a research announcement, not a proof-bearing paper: the central statements about classification, uniqueness, and the pullback construction are unsupported. The significance of the framework therefore cannot be assessed from the present text.

major comments (5)
  1. [Section 2, after Eq. (1)] The paper's central claim, that the classification of fields of covariances on NCP entails Cencov's uniqueness of the Fisher-Rao metric and Petz's classification of monotone quantum metrics, is asserted without proof. The text explicitly says the classification problem will be addressed in a series of forthcoming publications, and no statement of the conjectured classification theorem is given. Since this is the load-bearing result of the paper, the omission is not a presentation issue but an absence of the paper's main content.
  2. [Section 2, paragraph on fNCT] The uniqueness result for finite-dimensional tracial states, 'the GNS functor G is the only possible functor (up to the choice of two positive constants),' is stated without proof, without a precise definition of the two constants, and without a formal statement of what 'only possible functor' means. As one of the headline results, this unsupported assertion cannot be evaluated by the reader.
  3. [Section 3, final paragraph] The pullback construction by which a field of covariances induces a Riemannian metric tensor on a statistical subcategory is deferred to a forthcoming work. This includes the promises that the GNS functor leads to the Fisher-Rao metric on commutative statistical subcategories and to the Bures-Helstrom metric on faithful finite-dimensional states. Without this construction, the bridge from Eq. (1) to the metric classification claims is missing.
  4. [Section 2, condition (1)] Condition (1) identifies the underlying vector space of C_ρ with the GNS Hilbert space H_ρ and the morphism action with the GNS contraction, but it does not identify the relevant tangent spaces of statistical models, such as the self-adjoint traceless operators in Petz's framework, as subspaces of H_ρ. The manuscript does not show that every monotone quantum metric can be represented by such a functor, nor that every functor satisfying (1) yields a Petz-type metric. The claimed reduction of Petz's classification is therefore unsupported.
  5. [Footnote 8] Smoothness is explicitly set aside in footnote 8. Since the intended application is to Riemannian metrics on smooth statistical models, and since the pullback construction is essential to connecting fields of covariances to Fisher-Rao and Bures-Helstrom metrics, the differential-geometric content of the paper remains informal and cannot ground the announced theorems.
minor comments (5)
  1. [References] Reference [26] misspells the first author's name as 'Voicolescu'; it should be 'Voiculescu'.
  2. [References] Reference [22] writes 'Sudar' without the diacritic; the correct spelling is 'Sudár'.
  3. [Section 3, Gaussian example] In the equation Φ*_ξ(p_ξ') = p_{ξ∘ξ'}, the composition symbol ∘ on the parameter space M = R × R_+ is not defined. The group law of the affine group should be spelled out explicitly so the reader can verify the action groupoid structure.
  4. [Throughout] The transliteration 'Čencov' and 'Cencov' are used inconsistently; one standard spelling should be adopted throughout.
  5. [Section 2, covariance explanation] The sentence stating that statistical covariance coincides with the GNS scalar product should specify that this identification holds for zero-mean random variables, to avoid ambiguity.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the paper is a programmatic proposal whose headline entailments are explicitly deferred to forthcoming works.

full rationale

The paper makes no fitted-data prediction and does not reduce any claimed result to an input by construction. The central claims are all explicitly deferred: the abstract says 'two results that will appear in forthcoming works'; Section 2 states 'We will address the problem of classification of fields of covariances on NCP in a series of forthcoming publications' and introduces the fNCT uniqueness of the GNS functor as 'a result of our investigation'; Section 3 says the procedure for inducing Riemannian metrics from fields of covariances will be 'thoroughly discuss[ed] in a forthcoming work'. Condition (1), F∘C=F∘G, does stipulate that the underlying vector space of each covariance Hilbert space is the GNS Hilbert space and that the morphism action is the GNS contraction, so the monotonicity inequality is indeed built into the definition of a field of covariances. However, that is the intended categorical generalization of monotone metrics, not a circular derivation of Petz's classification: Petz's theorem remains an external classification result with independent content, and the paper does not claim to prove the classification here. The self-citations to references [6,7,8] are contextual motivation, not load-bearing support for the deferred claims. Unsupported or omitted proofs are a correctness risk but not circularity, so the appropriate score is 0.

Assumptions & free parameters 1 free parameters · 3 assumptions · 3 invented entities

The framework relies on standard operator-algebraic tools and two stipulated domain assumptions: the functorial representation of metrics via condition (1), and the representation of models as subcategories. No parameter is fitted to data; the only explicit freedom is a two-constant scaling in the announced tracial-state uniqueness. The novel entities carry no independent empirical evidence, and their value hinges on the promised proofs.

free parameters (1)
  • two positive constants in tracial-state uniqueness = unspecified positive constants
    The announced classification on fNCT identifies the GNS functor as the only field of covariances up to a choice of two positive constants. These constants are free degrees of freedom in the classification, not fitted to data, and they are not determined by the framework.
assumptions (3)
  • standard math GNS construction and standard W*-algebra theory are valid background.
    Invoked throughout, e.g., Section 1 and footnote 3, with references [4,24].
  • domain assumption Contravariant Riemannian metric tensors on a statistical model can be identified with Hilbert-structure-valued functors, with smoothness conditions neglected.
    Section 2 imposes condition (1), F∘C = F∘G, and footnote 8 explicitly sets aside the smoothness requirement. This identification is the bridge between geometry and the category NCP.
  • domain assumption Every statistical model can be represented as an essentially injective subcategory of NCP.
    Section 3 defines statistical subcategories and provides group-action examples, but models without a Lie group action are deferred to forthcoming work, so the universal representation claim is not established.
invented entities (3)
  • Category NCP of non-commutative probabilities
    purpose: Common categorical stage for classical and quantum statistical models.
    A new mathematical construction; its usefulness depends on the announced classification results, not on any independent empirical handle.
  • Field of covariances
    purpose: Functorial generalization of covariance intended to encode admissible Riemannian metrics on statistical models.
    Introduced in Section 2; there is no falsifiable prediction attached to it until the classification and pullback construction are proved.
  • Statistical subcategory
    purpose: Categorical representation of a statistical model, with symmetries encoded via action groupoids.
    Defined in Section 3; only group-action examples are worked out, and the general construction is postponed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Towards a category-theoretic foundation of Classical and Quantum Information Geometry." pith.science (2026). https://pith.science/paper/26ZJUJLM

@misc{pith2026250910262,
  author       = {Pith},
  title        = {Pith review of: Towards a category-theoretic foundation of Classical and Quantum Information Geometry},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/26ZJUJLM}},
  note         = {Machine review of arXiv:2509.10262}
}
abstract

We introduce the category $\mathsf{NCP}$, whose objects are pairs of W$^\ast$-algebras and normal states and whose morphisms are state-preserving unital completely positive (CPU) maps, as a common stage for classical and quantum information geometry, and we formulate two results that will appear in forthcoming works. First, we recast the problem of classifying admissible Riemannian geometries on classical and quantum statistical models in terms of functors $\mathfrak{C}:\mathsf{NCP}\to\mathsf{Hilb}$.These functors provide a generalization of classical statistical covariance, and we call them fields of covariances. A prominent example being the so-called GNS functor arising from the Gelfand-Naimark-Segal (GNS) construction. The classification of fields of covariances on $\mathsf{NCP}$ entails both \v{C}encov's uniqueness of the Fisher-Rao metric tensor and Petz's classification of monotone quantum metric tensors as particular cases. Then, we show how classical and quantum statistical models can be realized as subcategories of $\mathsf{NCP}$ in a way that takes into account symmetries. In this setting, the fields of covariances determine Riemannian metric tensors on the model that reduce to the Fisher-Rao, Fubini-Study, and Bures-Helstrom metric tensor in particular cases.

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

27 extracted references · 12 canonical work pages

  1. [1]

    S. I. Amari.Information Geometry and its Application. Springer, Japan, 2016. DOI: 10.1007/978-4-431-55978-8.↓6

    doi:10.1007/978-4-431-55978-8 2016

  2. [2]

    S. I. Amari and H. Nagaoka.Methods of Information Geometry. American Mathematical Society, Providence, RI, 2000. DOI: 10.1090/mmono/191.↓6

    doi:10.1090/mmono/191 2000

  3. [3]

    N. Ay, J. Jost, H. V. Le, and L. Schwachhöfer.Information Geometry. Springer International Publishing, 2017. DOI: 10.1007/978-3-319-56478-4.↓2

    doi:10.1007/978-3-319-56478-4 2017

  4. [4]

    Blackadar.Operator Algebras: Theory ofC ∗-algebras and von Neumann Algebras

    B. Blackadar.Operator Algebras: Theory ofC ∗-algebras and von Neumann Algebras. Springer-Verlag, Berlin, 2006. DOI: 10.1007/3-540-28517-2.↓3

    doi:10.1007/3-540-28517-2 2006

  5. [5]

    M. Choi. Completely positive linear maps on complex matrices.Linear Algebra and its Applications, 10(3):285–290, 1975. DOI: 10.1016/0024-3795(75)90075-0.↓2

    doi:10.1016/0024-3795(75)90075-0 1975

  6. [6]

    Lecture Notes in Computer Science

    F.M.Ciaglia, F.DiCosmo, andL.González-Bravo.Can Čencov Meet Petz, pages363–371. Lecture Notes in Computer Science. Springer Nature Switzerland, 2023. DOI: 10.1007/978- 3-031-38299-4_38, arXiv:2305.12482 [math-ph].↓3

    arXiv 2023

  7. [7]

    F. M. Ciaglia, F. Di Nocera, J. Jost, and L. Schwachhöfer. Parametric models and information geometry onW∗-algebras.Information Geometry, 5(1):329–354, 2023. DOI: 10.1007/s41884-022-00094-6, arXiv:2207.09396 [math-ph].↓3

    work page Pith review arXiv 2023

  8. [8]

    F. M. Ciaglia, J. Jost, and L. Schwachhöfer. From the Jordan product to Riemannian geometries on classical and quantum states.Entropy, 22(06):637–27, 2020. DOI: 10.3390/e22060637, arXiv:2005.02023 [math-ph].↓3

    work page Pith review arXiv 2020

Show all 27 references
  1. [9]

    Princeton University Press, December

    Harald Cramér.Mathematical Methods of Statistics. Princeton University Press, December

  2. [10]

    Fujiwara

    A. Fujiwara. Hommage to Chentsov’s theorem.Information Geometry, 2023. DOI: 10.1007/s41884-022-00077-7.↓2

    2023 doi

  3. [11]

    Hendriks

    H. Hendriks. A Cramer-Rao Type Lower Bound for Estimators with Values in a Manifold.Journal of Multivariate Analysis, 38(2):245 – 261, 1991. DOI: 10.1016/0047- 259X(91)90044-3.↓5

    1991 doi

  4. [12]

    A. Jenčová. Affine connections, duality and divergences for a von Neumann algebra. arXiv:0311004 [math-ph], 2003.↓3 7

    2003

  5. [13]

    A. Jenčová. A construction of a nonparametric quantum information manifold.Journal of Functional Analysis, 239(1):1–20, 2006. DOI: 10.1016/j.jfa.2006.02.007, arXiv:0511065 [math-ph].↓3

    2006 doi

  6. [14]

    A. Jenčová. The exponential Orlicz space in quantum information geometry. Information Geometry, 7(S1):377–395, January 2023. DOI: 10.1007/s41884-023-00097-x, arXiv:2301.06906 [quant-ph].↓3

    2023 arXiv

  7. [15]

    K. C. Mackenzie.General theory of Lie groupoids and Lie algebroids. Cambridge University Press, 2005. DOI: 10.1017/cbo9781107325883.↓4

    2005 doi

  8. [16]

    Shiva Publishing Limited, 1980

    Peter W Michor.Manifolds of differentiable mappings. Shiva Publishing Limited, 1980. DOI: 10.1007/978-3-642-11102-0_5.↓2

    1980 doi

  9. [17]

    E. A. Morozova and N. N. Čencov. Markov maps in noncommutative probability theory and mathematical statistics. In Y. V. Prokhorov, V. A. Statulevičius, V. V. Sazonov, and B. Grigelionis, editors,Probability theory and mathematical statistics: proceedings of the Fourth Vilnius ...

    doi

  10. [18]

    E. A. Morozova and N. N. Čencov. Markov invariant geometry on manifolds of states. Journal of Soviet Mathematics, 56(5):2648–2669, 1991. DOI: 10.1007/BF01095975.↓2

    1991 doi

  11. [19]

    TheFishermetricasametriconthecotangentbundle.Information Geometry, 7(1), 2024

    H.Nagaoka. TheFishermetricasametriconthecotangentbundle.Information Geometry, 7(1), 2024. DOI: 10.1007/s41884-023-00126-9, arXiv:2310.13237 [cs.IT].↓5

    2024

  12. [20]

    M. A. Nielsen and I. L. Chuang.Quantum Computation and Quantum Information. Cambridge University Press, New York, NY, 2011. DOI: 10.1017/CBO9780511976667. ↓2

    2011 doi

  13. [21]

    D. Petz. Monotone metrics on matrix spaces.Linear Algebra and its Applications, 244:81– 96, 1996. DOI: 10.1016/0024-3795(94)00211-8.↓2

    1996 doi

  14. [22]

    Petz and C

    D. Petz and C. Sudar. Geometries of Quantum States.Journal of Mathematical Physics, 37, 1996. DOI: 10.1063/1.531535.↓2

    1996 doi

  15. [23]

    Pistone and C

    G. Pistone and C. Sempi. An infinite-dimensional geometric structure on the space of all the probability measures equivalent to a given one.The Annals of Statistics, 23(5):1543– 1561, 1995. DOI: 10.1214/aos/117632431.↓2

    1995

  16. [24]

    Takesaki.Theory of Operator Algebra I

    M. Takesaki.Theory of Operator Algebra I. Springer-Verlag, Berlin, 2002. DOI: 10.1007/978-1-4612-6188-9.↓3

    2002 doi

  17. [25]

    N. N. Čencov.Statistical Decision Rules and Optimal Inference. American Mathematical Society, Providence, RI, 1982. DOI: 10.1090/mmono/053.↓2

    1982 doi

  18. [26]

    D. V. Voicolescu, K. J. Dykema, and A. Nica.Free Random Variables. American Mathematical Society, Providence, RI, 1992. ISBN: 978-0-8218-1140-5.↓3 8

    1992

  19. [1946]

    DOI: 10.1515/9781400883868.↓2

    doi

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.