pith. sign in

arxiv: 2606.28213 · v1 · pith:ZWKT3N5Unew · submitted 2026-06-26 · 🧮 math.MG · math.FA

Wasserstein Barycenter Convexity Detects Hilbertian Geometry

Bang-Xian Han , Deng-Yu Liu This is my paper

Pith reviewed 2026-06-29 01:28 UTC · model grok-4.3

classification 🧮 math.MG math.FA
keywords Wasserstein barycenterentropy convexityHilbertian geometryBanach spacesoptimal transportFinsler manifoldsparallelogram identity
0
0 comments X

The pith

If the entropy satisfies Jensen's inequality at Wasserstein barycenters of arbitrary finite measures on a finite-dimensional normed space with Lebesgue measure, then the norm must be induced by an inner product.

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

The paper establishes that convexity of Boltzmann entropy at Wasserstein barycenters is selective enough to force finite-dimensional normed spaces to be Hilbertian. This gives an optimal-transport criterion that rules out all non-Hilbertian norms, in contrast to curvature conditions satisfied by every finite-dimensional space. The argument applies without assuming strict convexity and uses the inequality holding for arbitrary finite families of measures. As a direct consequence, smooth reversible Finsler manifolds obeying the corresponding barycentric condition must have Riemannian tangent norms.

Core claim

We prove that convexity of the Boltzmann entropy at Wasserstein barycenters is strong enough to distinguish Hilbert spaces from general Banach spaces. Thus Wasserstein barycenters provide an intrinsic optimal-transport test for Hilbertian geometry. More precisely, we show that if a finite-dimensional normed vector space, equipped with Lebesgue measure, satisfies the Wasserstein Jensen's inequality for the entropy at barycenters of arbitrary finite families of probability measures, then its norm must be induced by an inner product.

What carries the argument

The Wasserstein Jensen's inequality for the entropy at barycenters of arbitrary finite families, established via a rank-one polarization argument yielding the dual parallelogram identity and a maximal-face trapping argument ruling out flat faces of the unit ball.

If this is right

  • Smooth reversible Finsler manifolds satisfying the barycentric curvature-dimension condition must have Riemannian tangent norms.
  • Barycenter convexity of entropy excludes every non-Hilbertian norm, whereas the Lott-Sturm-Villani nonnegative Ricci condition holds in all finite-dimensional spaces.
  • The conclusion requires no assumption of strict convexity on the norm.
  • The rank-one polarization and maximal-face trapping arguments suffice to derive the dual parallelogram identity and exclude flat faces.

Where Pith is reading between the lines

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

  • The same barycenter test might detect Hilbertian structure in discrete or infinite-dimensional settings if the inequality can be suitably formulated.
  • Numerical checks in low-dimensional non-Euclidean norms could quantify how close the inequality comes to holding before the geometry forces a contradiction.
  • The technique supplies a new way to characterize inner-product norms through optimal-transport functionals without reference to curvature.

Load-bearing premise

The inequality must hold for arbitrary finite families of probability measures on the space equipped with Lebesgue measure, and the space must be finite-dimensional.

What would settle it

Exhibiting a single non-Hilbertian finite-dimensional normed space equipped with Lebesgue measure in which the entropy satisfies the Jensen inequality at every Wasserstein barycenter of every finite collection of measures would falsify the claim.

read the original abstract

We prove that convexity of the Boltzmann entropy at Wasserstein barycenters is strong enough to distinguish Hilbert spaces from general Banach spaces. Thus Wasserstein barycenters provide an intrinsic optimal-transport test for Hilbertian geometry. More precisely, we show that if a finite-dimensional normed vector space, equipped with Lebesgue measure, satisfies the Wasserstein Jensen's inequality for the entropy at barycenters of arbitrary finite families of probability measures, then its norm must be induced by an inner product. This contrasts sharply with a well-known result: every finite-dimensional normed vector space satisfies the nonnegative Ricci curvature condition in the sense of Lott--Sturm--Villani, whereas barycenter convexity excludes all non-Hilbertian norms. As a consequence, smooth reversible Finsler manifolds satisfying the corresponding barycentric curvature-dimension condition have Riemannian tangent norms. The proof does not assume strict convexity of the norm. Its two main ingredients are a rank-one polarization argument, which yields the dual parallelogram identity in the strictly convex case, and a maximal-face trapping argument, which rules out flat faces of the unit ball.

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 proves that if a finite-dimensional normed vector space equipped with Lebesgue measure satisfies the Wasserstein Jensen inequality for Boltzmann entropy at barycenters of arbitrary finite families of probability measures, then the norm must arise from an inner product. The argument splits into a rank-one polarization step yielding the dual parallelogram identity under strict convexity and a maximal-face trapping step that rules out flat faces of the unit ball in the general case. This yields a sharp distinction from the Lott–Sturm–Villani curvature condition, which holds for every finite-dimensional normed space, and implies that smooth reversible Finsler manifolds obeying the corresponding barycentric curvature-dimension condition must have Riemannian tangent norms.

Significance. If correct, the result supplies an intrinsic optimal-transport characterization of Hilbertian geometry that is strictly stronger than LSV curvature. The proof’s explicit handling of non-strict convexity and its applicability to arbitrary finite families are notable strengths; the manuscript also ships a direct implication from the stated inequality to the inner-product property without additional ad-hoc assumptions.

major comments (2)
  1. [§3] §3 (rank-one polarization): the derivation of the dual parallelogram identity from the entropy inequality at two-point barycenters appears to rely on the polarization identity holding for rank-one measures; a concrete verification that the entropy functional’s second variation produces exactly the required quadratic form without hidden strict-convexity assumptions would strengthen the step.
  2. [§4] §4 (maximal-face trapping): the argument that a flat face of the unit ball would produce a family of measures whose barycenter entropy violates convexity is load-bearing for the non-strictly-convex case; an explicit construction of the violating family (or a reference to the precise trapping lemma) should be supplied so that the reduction to the strictly convex case is fully transparent.
minor comments (2)
  1. [Abstract / Theorem 1.1] The abstract states the result for “arbitrary finite families”; the main theorem statement should repeat this quantifier verbatim to avoid any ambiguity about whether the inequality is assumed only for two-point or for n-point barycenters.
  2. [§2] Notation for the Wasserstein barycenter and the entropy functional is introduced without a dedicated preliminary subsection; a short paragraph collecting the definitions (including the precise normalization of Lebesgue measure) would improve readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading, positive evaluation, and constructive suggestions. The comments highlight opportunities to improve clarity in the two main steps of the argument. We address each point below and will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [§3] §3 (rank-one polarization): the derivation of the dual parallelogram identity from the entropy inequality at two-point barycenters appears to rely on the polarization identity holding for rank-one measures; a concrete verification that the entropy functional’s second variation produces exactly the required quadratic form without hidden strict-convexity assumptions would strengthen the step.

    Authors: The derivation in §3 applies the given barycenter convexity inequality to two-point measures whose supports lie along rank-one lines and then extracts the second variation of the entropy functional at the barycenter. This second variation is computed directly from the definition of Boltzmann entropy (with respect to Lebesgue measure) and the explicit form of the Wasserstein distance in the underlying norm; the resulting quadratic form is precisely the dual parallelogram identity. No additional strict-convexity hypothesis on the norm is used beyond the standing assumption that the inequality holds for all finite families. To make this computation fully transparent, we will insert an expanded paragraph in the revised §3 that carries out the second-variation calculation line by line. revision: yes

  2. Referee: [§4] §4 (maximal-face trapping): the argument that a flat face of the unit ball would produce a family of measures whose barycenter entropy violates convexity is load-bearing for the non-strictly-convex case; an explicit construction of the violating family (or a reference to the precise trapping lemma) should be supplied so that the reduction to the strictly convex case is fully transparent.

    Authors: Section 4 proceeds by contradiction: if the unit ball possesses a maximal flat face of positive dimension, one can construct a finite family of probability measures supported on parallel translates of that face whose Wasserstein barycenter lies in the relative interior of the face; the entropy convexity inequality then fails because the entropy functional is strictly concave along the flat directions. The construction is canonical (uniform measures on small balls centered at points separated along the face) and does not rely on any external lemma. In the revision we will spell out this family explicitly, compute the entropy values at the measures and at their barycenter, and verify the violation, thereby rendering the reduction to the strictly convex case self-contained. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The derivation proceeds by establishing the dual parallelogram identity via rank-one polarization (in the strictly convex case) and then applying maximal-face trapping to rule out flat faces of the unit ball, directly yielding the known finite-dimensional characterizations of inner-product norms. These steps rely on the stated Wasserstein Jensen inequality for arbitrary finite families and Lebesgue measure, without reducing to fitted parameters, self-definitional equivalences, or load-bearing self-citations. External results such as Lott-Sturm-Villani nonnegative Ricci curvature are cited only for contrast and are independent of the present argument. The proof is self-contained.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on standard properties of Wasserstein distance, Boltzmann entropy, and Lebesgue measure in finite-dimensional normed spaces; no free parameters or invented entities are introduced.

axioms (2)
  • standard math Standard properties of Wasserstein distance and Boltzmann entropy on normed spaces
    Invoked throughout the statement of the Jensen inequality and barycenter construction.
  • domain assumption Finite-dimensionality of the normed vector space
    Explicitly required for the characterization to hold.

pith-pipeline@v0.9.1-grok · 5721 in / 1250 out tokens · 31337 ms · 2026-06-29T01:28:06.568000+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

25 extracted references

  1. [1]

    Agueh and G

    M. Agueh and G. Carlier,Barycenters in the Wasserstein space, SIAM J. Math. Anal. 43 (2011), no. 2, 904–924

    2011

  2. [2]

    Ambrosio, N

    L. Ambrosio, N. Gigli, and G. Savar´ e,Metric measure spaces with Riemannian Ricci curvature bounded from below, Duke Math. J. 163 (2014), no. 7, 1405–1490

    2014

  3. [3]

    Amir,Characterizations of Inner Product Spaces, Operator Theory: Advances and Applications 20, Birkh¨ auser, Basel, 1986

    D. Amir,Characterizations of Inner Product Spaces, Operator Theory: Advances and Applications 20, Birkh¨ auser, Basel, 1986. 16

    1986

  4. [4]

    Bao, S.-S

    D. Bao, S.-S. Chern, and Z. Shen,An Introduction to Riemann–Finsler Geometry, Graduate Texts in Mathematics 200, Springer, New York, 2000

    2000

  5. [5]

    Carlier and I

    G. Carlier and I. Ekeland,Matching for teams, Econom. Theory 42 (2010), no. 2, 397–418

    2010

  6. [6]

    Cordero-Erausquin, R

    D. Cordero-Erausquin, R. J. McCann, and M. Schmuckenschl¨ ager,A Riemannian in- terpolation inequality ` a la Borell, Brascamp and Lieb, Invent. Math. 146 (2001), no. 2, 219–257

    2001

  7. [7]

    Cordero-Erausquin, R

    D. Cordero-Erausquin, R. J. McCann, and M. Schmuckenschl¨ ager,Pr´ ekopa–Leindler type inequalities on Riemannian manifolds, Jacobi fields, and optimal transport, Ann. Fac. Sci. Toulouse Math. (6) 15 (2006), no. 4, 613–635

    2006

  8. [8]

    Gigli,On the differential structure of metric measure spaces and applications, Mem

    N. Gigli,On the differential structure of metric measure spaces and applications, Mem. Amer. Math. Soc. 236 (2015), no. 1113

    2015

  9. [9]

    Han, D.-Y

    B.-X. Han, D.-Y. Liu, and Z.-N. Zhu,On the geometry of Wasserstein barycenter I, arXiv:2412.01190v2 [math.MG], 2025

    arXiv 2025

  10. [10]

    Han, D.-Y

    B.-X. Han, D.-Y. Liu, and Z.-N. Zhu,Barycenter curvature-dimension condition for extended metric measure spaces, Indag. Math. (N.S.) 37 (2026), no. 3, 892–902

    2026

  11. [11]

    Jordan and J

    P. Jordan and J. von Neumann,On inner products in linear metric spaces, Ann. of Math. (2) 36 (1935), no. 3, 719–723

    1935

  12. [12]

    Kakutani,Some characterizations of Euclidean space, Japan

    S. Kakutani,Some characterizations of Euclidean space, Japan. J. Math. 16 (1939), 93–97

    1939

  13. [13]

    Kim and B

    Y.-H. Kim and B. Pass,Wasserstein barycenters over Riemannian manifolds, Adv. Math. 307 (2017), 640–683

    2017

  14. [14]

    Lott and C

    J. Lott and C. Villani,Ricci curvature for metric-measure spaces via optimal transport, Ann. of Math. (2) 169 (2009), no. 3, 903–991

    2009

  15. [15]

    R. J. McCann,A convexity principle for interacting gases, Adv. Math. 128 (1997), no. 1, 153–179

    1997

  16. [16]

    Ohta,Finsler interpolation inequalities, Calc

    S. Ohta,Finsler interpolation inequalities, Calc. Var. Partial Differential Equations 36 (2009), no. 2, 211–249

    2009

  17. [17]

    Ohta,Ricci curvature, entropy, and optimal transport, inOptimal Transportation, London Math

    S. Ohta,Ricci curvature, entropy, and optimal transport, inOptimal Transportation, London Math. Soc. Lecture Note Ser. 413, Cambridge University Press, Cambridge, 2014, 145–199

    2014

  18. [18]

    Ohta,Nonlinear geometric analysis on Finsler manifolds, Eur

    S. Ohta,Nonlinear geometric analysis on Finsler manifolds, Eur. J. Math. 3 (2017), no. 4, 916–952

    2017

  19. [19]

    Ohta and K.-T

    S. Ohta and K.-T. Sturm,Heat flow on Finsler manifolds, Comm. Pure Appl. Math. 62 (2009), no. 10, 1386–1433

    2009

  20. [20]

    Ohta and K.-T

    S. Ohta and K.-T. Sturm,Bochner–Weitzenb¨ ock formula and Li–Yau estimates on Finsler manifolds, Adv. Math. 252 (2014), 429–448

    2014

  21. [21]

    R. T. Rockafellar,Convex Analysis, Princeton Mathematical Series 28, Princeton University Press, Princeton, 1970

    1970

  22. [22]

    Sturm,On the geometry of metric measure spaces

    K.-T. Sturm,On the geometry of metric measure spaces. I, Acta Math. 196 (2006), no. 1, 65–131. 17

    2006

  23. [23]

    Sturm,On the geometry of metric measure spaces

    K.-T. Sturm,On the geometry of metric measure spaces. II, Acta Math. 196 (2006), no. 1, 133–177

    2006

  24. [24]

    Villani,Optimal transport: old and new, Grundlehren der mathematischen Wis- senschaften 338, Springer, Berlin, 2009

    C. Villani,Optimal transport: old and new, Grundlehren der mathematischen Wis- senschaften 338, Springer, Berlin, 2009

    2009

  25. [25]

    von Renesse and K.-T

    M.-K. von Renesse and K.-T. Sturm,Transport inequalities, gradient estimates, entropy, and Ricci curvature, Comm. Pure Appl. Math. 58 (2005), no. 7, 923–940. Declarations The authors declare that they have no conflict of interest and that the manuscript has no associated data. 18

    2005