Pith. sign in

REVIEW 4 minor 27 references

Stronger convexity notions make the Wasserstein proximal map non-expansive; ordinary generalized geodesic convexity yields only local half-Hölder continuity.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-10 11:17 UTC pith:KATKUQQC

load-bearing objection Solid, self-contained progress on Wasserstein proximal stability: non-expansivity under total/2-base convexity plus a general local 1/2-Hölder for generalized-geodesically convex functionals.

arxiv 2607.08209 v1 pith:KATKUQQC submitted 2026-07-09 math.OC math.APmath.FA

On the stability of proximal operators in Wasserstein spaces under different notions of convexity

Simone Di Marino , Sara Farinelli , Emanuele Naldi This is my paper
classification math.OC math.APmath.FA MSC 49J4049Q2290C2546G05
keywords Wasserstein proximal operatornon-expansivitytotal convexitygeneralized geodesics2-base generalized geodesicsconvex orderMoreau envelopelocal Hölder continuity
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

In Hilbert space the proximal map of a convex functional is always 1-Lipschitz. The same statement for the Wasserstein proximal map on probability measures has been open for years. This paper settles how much convexity is needed. Total convexity (convexity along every transport plan) and a new intermediate notion called convexity along 2-base generalized geodesics both force the proximal map to be non-expansive. Merely convexity along ordinary generalized geodesics is enough only for a local half-Hölder estimate that holds on the whole space of measures with finite second moment. Along the way the authors prove that total convexity is extremely rigid: such functionals are monotone with respect to convex order, so their minimizers must be Dirac masses and they cannot approximate typical internal energies. Conditional (1+δ)-Lipschitz bounds are also obtained when one measure is close to a Dirac or a minimizer. The results clarify which contraction properties survive when one discretizes gradient flows in Wasserstein space.

Core claim

If a proper lower-semicontinuous functional on the 2-Wasserstein space is totally convex, or even only convex along 2-base generalized geodesics, then its proximal operator is non-expansive. If the functional is merely convex along outer generalized geodesics, the proximal operator is still locally ½-Hölder continuous everywhere. Total convexity further implies monotonicity with respect to convex order of measures, which excludes internal energies and prevents approximation by totally convex functionals.

What carries the argument

Three nested convexity notions—total convexity (convexity along every coupling), convexity along 2-base generalized geodesics (curves induced by 3-geodesic plans), and ordinary generalized geodesic convexity—together with the associated weak-non-expansivity identities that compare the Wasserstein distance of the images to a carefully chosen transport cost between the originals.

Load-bearing premise

The functionals must be convex along outer generalized geodesics (not merely the weaker inner version), otherwise the proximal map need not even be single-valued for every starting measure.

What would settle it

Exhibit a functional that is convex along outer generalized geodesics yet whose proximal map fails to be locally ½-Hölder continuous at some pair of measures with finite second moment, or construct a totally convex functional whose proximal map expands distances.

Watch this falsifier — get emailed when new claim-graph text bears on it.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

0 major / 4 minor

Summary. The paper studies stability (especially non-expansivity) of the Wasserstein proximal operator prox_τ^W F under several notions of convexity on P_2(R^d). For proper lsc totally convex F it proves that the proximal map is non-expansive (Proposition 2.1), via Lagrangian lifting; it further shows that total convexity implies monotonicity with respect to convex order (Theorem 2.6), which immediately excludes internal energies and rules out approximation by totally convex functionals. It introduces weak non-expansivity for geodesically convex and generalized-geodesically convex functionals (Theorems 3.4 and 3.8), defines convexity along 2-base generalized geodesics and proves non-expansivity under that assumption (Theorem 3.17), obtains conditional (1+δ)-Lipschitz estimates when one measure is near a Dirac or a minimizer (Section 3.4), and establishes local ½-Hölder continuity of the proximal map for any functional that is convex along outer generalized geodesics (Theorem 4.3).

Significance. The work cleanly organizes the hierarchy of convexity notions that control the regularity of Wasserstein proximal maps and settles several questions left open by Ambrosio–Gigli–Savaré, Carlen–Craig and Adve–Mészáros. The non-expansivity results under total convexity and under the new 2-base notion, the rigidity theorem with respect to convex order, and the unconditional local ½-Hölder estimate (Theorem 4.3) are all new and of direct interest for the analysis of JKO schemes and proximal algorithms in Wasserstein space. Proofs rely only on standard gluing, first-variation inequalities and Lagrangian lifting; they are complete and self-contained. The outer-versus-inner distinction is correctly flagged and used only where uniqueness is needed.

minor comments (4)
  1. The discussion after Example 3.18 on the size of the class of 2-base convex functionals is incomplete; a short remark clarifying whether the notion is strictly weaker than total convexity (or an explicit open question) would help the reader.
  2. Several typographical slips appear (e.g., “constext”, “aprroximated”, “funcionals”, “gneralized”, “costrained”, “techinques”). A careful proof-reading pass is needed.
  3. In Definition 3.13 the coordinates of the 3-geodesic plan are written (x,y,˜x,˜y) while earlier sections use (x,˜x,˜y,y); a uniform convention would improve readability.
  4. Lemma 1.6 and Remark 1.7 correctly emphasize the necessity of outer generalized geodesic convexity for uniqueness; a one-sentence pointer in the introduction would make this hypothesis more visible to non-specialists.

Circularity Check

0 steps flagged

No circularity: pure mathematical derivations of proximal stability under stated convexity notions, self-contained against external benchmarks.

full rationale

The paper is a pure math.OC work deriving non-expansivity of the Wasserstein proximal map under total convexity (Prop. 2.1 via Lagrangian lifting to Hilbert-space prox non-expansivity) or under the new 2-base generalized geodesic convexity (Thm. 3.17 via 3-geodesic plans and first-order expansions of the Moreau functional), and local 1/2-Hölder continuity under outer generalized geodesic convexity (Thm. 4.3 via generalized geodesics of the Moreau envelope plus local boundedness of C_τ^μ from Lemma 4.2). All steps are explicit inequalities obtained from gluing of optimal plans, convexity inequalities along the resulting curves, and elementary rearrangements (e.g., 2c⟨v,w⟩ = ∥v∥² + c²∥w∥² − ∥v−cw∥²); none is definitional of the target claim, none fits a free parameter to data, and none imports a uniqueness theorem solely from overlapping authors as an external fact that forces the result. Self-citations (e.g., to Cavagnari–Savaré–Sodini or the authors’ own prior work on total convexity) supply independent lemmas used as black boxes; the outer-vs-inner distinction (Rem. 1.7) is an explicit hypothesis needed for uniqueness, not a circular reduction. The rigidity results of §2.2 correctly delimit the class of totally convex functionals via monotonicity w.r.t. convex order and do not feed back into the stability statements. Score 0 is therefore the honest finding.

Axiom & Free-Parameter Ledger

0 free parameters · 3 axioms · 2 invented entities

The paper works entirely inside the standard axiomatic framework of Wasserstein geometry and convex analysis; the only new objects are definitional (2-base generalized geodesics, weak non-expansivity). No free parameters or unproved physical hypotheses appear.

axioms (3)
  • standard math P_2(R^d) is a geodesic metric space and the usual gluing lemma for optimal plans holds.
    Used throughout Sections 1–4; classical from Ambrosio–Gigli–Savaré.
  • domain assumption A proper lsc functional that is λ-convex along outer generalized geodesics with λτ > −1 has a unique proximal point for every measure.
    Lemma 1.6; the outer-versus-inner distinction is essential and is illustrated by a counter-example (Remark 1.7).
  • standard math Total convexity is equivalent to convexity of the Lagrangian lift to L^2(Ω;R^d).
    Taken from Cavagnari–Savaré–Sodini and used in the proof of non-expansivity (Proposition 2.1).
invented entities (2)
  • 2-base generalized geodesic / convexity along 2-base generalized geodesics no independent evidence
    purpose: Intermediate convexity notion strictly between total convexity and ordinary generalized geodesic convexity that still guarantees non-expansivity of the proximal map.
    Defined in Section 3.3; no independent evidence outside the paper is claimed.
  • weak non-expansivity (generalized) no independent evidence
    purpose: Relaxed contraction property that holds under mere geodesic or generalized geodesic convexity.
    Extends the definition of Adve–Mészáros; used to obtain conditional (1+δ)-Lipschitz estimates.

pith-pipeline@v1.1.0-grok45 · 35636 in / 2300 out tokens · 27526 ms · 2026-07-10T11:17:06.276649+00:00 · methodology

0 comments
read the original abstract

The proximal operator is a fundamental tool in variational analysis and optimization. In the setting of a Hilbert space, given a proper, lower semicontinuous convex functional, its proximal operator is non-expansive, that is, 1-Lipschitz continuous. In the Wasserstein setting, the contraction properties of this operator have been investigated from different perspectives by Carlen and Craig and Adve and M\'esz\'aros, among others, and are not completely understood. In this paper, we study the stability properties of proximal maps, with a particular focus on non-expansivity, under various notions of convexity of the functional that can be considered in the Wasserstein space.

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

27 extracted references · 27 canonical work pages · 4 internal anchors

  1. [1]

    Adve and A

    A. Adve and A. R. Mészáros. On nonexpansiveness of metric projection operators on Wasserstein spaces.Advances in Calculus of Variations, 18(4):1207–1222, 2025

    work page 2025

  2. [2]

    Alfonsi and B

    A. Alfonsi and B. Jourdain. Wasserstein projections in the convex order: regularity and characteri- zation in the quadratic Gaussian case.Electronic Journal of Probability, 31:1–29, 2026

    work page 2026

  3. [3]

    Lectures in Mathematics ETH Zürich

    L.Ambrosio, N.Gigli, andG.Savaré.Gradient Flows in Metric Spaces and in the Space of Probability Measures. Lectures in Mathematics ETH Zürich. Birkhäuser, 2. ed edition, 2008

    work page 2008

  4. [4]

    Ballantine

    C. Ballantine. Products of positive definite matrices. i.Pacific Journal of Mathematics, 23:427–433, 1967

    work page 1967

  5. [5]

    ABreniertheoremon(P 2(...P2(h)...),w2)andapplications to adapted transport.arXiv preprint arXiv:2509.03506, 2025

    M.Beiglböck, G.Pammer, andS.Schrott. ABreniertheoremon(P 2(...P2(h)...),w2)andapplications to adapted transport.arXiv preprint arXiv:2509.03506, 2025

    work page arXiv 2025

  6. [6]

    Difference of Convex Programming in the Wasserstein Space with Applications to MMD Optimization

    C. Bonet, P.-C. Aubin-Frankowski, and Y. Mroueh. Difference of convex programming in the wasser- stein space with applications to mmd optimization.arXiv preprint arXiv:2606.27767, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  7. [7]

    E. A. Carlen and K. Craig. Contraction of the proximal map and generalized convexity of the Moreau–Yosida regularization in the 2-Wasserstein metric.Mathematics and Mechanics of Complex Systems, 1(1):33–65, 2013

    work page 2013

  8. [8]

    Cavagnari, G

    G. Cavagnari, G. Savaré, and G. E. Sodini. A Lagrangian approach to totally dissipative evolutions in Wasserstein spaces.Journal of Differential Equations, 470:114395, 2026

    work page 2026

  9. [9]

    K. Craig. The exponential formula for the Wasserstein metric.ESAIM: COCV, 22(1):169–187, 2016

    work page 2016

  10. [10]

    De Philippis, A

    G. De Philippis, A. R. Mészáros, F. Santambrogio, and B. Velichkov. BV estimates in optimal transportation and applications.Archive for Rational Mechanics and Analysis, 219(2):829–860, 2016

    work page 2016

  11. [11]

    Delalande and S

    A. Delalande and S. Farinelli. Regularized moment measures.Potential Analysis, 64, 2025

    work page 2025

  12. [12]

    Inexact JKO and proximal-gradient algorithms in the Wasserstein space

    S. Di Marino, E. Naldi, and S. Villa. Inexact JKO and proximal-gradient algorithms in the Wasser- stein space.arXiv preprint arXiv:2505.23517, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  13. [13]

    X. Feng, L. Wang, D. Needell, and R. Lai. Learn to evolve: Self-supervised neural jko operator for wasserstein gradient flow.arXiv preprint arXiv:2601.05583, 2026

    work page arXiv 2026

  14. [14]

    Gozlan, C

    N. Gozlan, C. Roberto, P. M. Samson, and P. Tetali. Kantorovich duality for general transport costs and applications.Journal of Functional Analysis, 273(11):3327–3405, 2017

    work page 2017

  15. [15]

    Hoheisel, M

    T. Hoheisel, M. Laborde, and A. Oberman. A regularization interpretation of the proximal point method for weakly convex functions.Journal of Dynamics and Games, 7(1):79–96, 2020

    work page 2020

  16. [16]

    Jordan, D

    R. Jordan, D. Kinderlehrer, and F. Otto. The variational formulation of the Fokker–Planck equation. SIAM Journal on Mathematical Analysis, 29(1):1–17, 1998. 30

    work page 1998

  17. [17]

    J. Jost. Convex functionals and generalized harmonic maps into spaces of non positive curvature. Commentarii Mathematici Helvetici, 70:659–673, 1995

    work page 1995

  18. [18]

    J. Kim, Y. H. Kim, and A. Natale. Stability of Wasserstein projections in convex order via metric extrapolation.Electronic Communications in Probability, 31:1–12, 2026

    work page 2026

  19. [19]

    Muratori and G

    M. Muratori and G. Savaré. Gradient flows and evolution variational inequalities in metric spaces. i: structural properties.Journal of Functional Analysis, 278(4):108347, 2020

    work page 2020

  20. [20]

    Nenna and B

    L. Nenna and B. Pass. Transport type metrics on the space of probability measures involving singular base measures.Applied Mathematics & Optimization, 87(2):28, 2023

    work page 2023

  21. [21]

    F. Otto. The geometry of dissipative evolution equations: the porous medium equation.Communi- cations in Partial Differential Equations, 26:101 – 174, 2001

    work page 2001

  22. [22]

    Totally convex functions, $L^2$-Optimal transport for laws of random measures, and solution to the Monge problem

    A. Pinzi and G. Savaré. Totally convex functions,L2-optimal transport for laws of random measures, and solution to the Monge problem.arXiv preprint arXiv:2509.01768, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  23. [23]

    Roudneff

    A. Roudneff. Modelisation macroscopique de mouvements de foule.https://theses.hal.science/tel- 00678596, PhD Thesis, Dec. 2011

    work page 2011

  24. [24]

    Santambrogio.Optimal transport for applied mathematicians, volume 55

    F. Santambrogio.Optimal transport for applied mathematicians, volume 55. Birkhäuser, New York, 2015

    work page 2015

  25. [25]

    Theexistence ofprobability measures withgiven marginals.The Annals of Mathematical Statistics, 36(2):423–439, 1965

    V.Strassen. Theexistence ofprobability measures withgiven marginals.The Annals of Mathematical Statistics, 36(2):423–439, 1965

    work page 1965

  26. [26]

    K. Tanaka. Accelerated gradient descent method for functionals of probability measures by new convexity and smoothness based on transport maps.arXiv preprint arXiv:2305.05127, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  27. [27]

    Tanguy, L

    E. Tanguy, L. Chapel, and J. Delon. Sliced transport plans, 2026. 31

    work page 2026