McKean-Vlasov stochastic equations with H\"older coefficients

Alessio Rondelli; Andrea Pascucci

REVIEW 2 major objections 3 minor 10 references

McKean-Vlasov stochastic equations with H\"older coefficients

T0 review · 2 major / 3 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read McKean–Vlasov equations with Hölder coefficients admit a unique weak solution for every initial law.

desk verdict A clean fixed-point proof of weak well-posedness for McKean–Vlasov SDEs with Hölder interaction kernels, but the contraction display (3.3) is miswritten and must be corrected. read the letter →

arxiv 2412.00834 v1 pith:YA46BRFD submitted 2024-12-01 math.PR

Andrea Pascucci , Alessio Rondelli This is my paper · ORCID

classification math.PR MSC 60H1060J60

keywords McKean–VlasovequationsHöldercontinuouscoefficientsweakwell-posednesscontractionmappingdual-HöldermetricGaussianestimatesstochasticdifferentialMarkovprocesses

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 establishes that a McKean–Vlasov stochastic differential equation with bounded Hölder-continuous coefficients and a uniformly non-degenerate diffusion has exactly one weak solution for any prescribed initial law. The proof works by freezing the flow of marginal laws, solving the resulting linear diffusion, and showing that the map from a frozen flow to its output flow is a contraction in a dual-Hölder metric on continuous paths of probability measures. This yields a fixed point whose associated diffusion is the unique weak solution. The argument avoids derivatives with respect to the measure argument, relying only on standard Gaussian estimates for uniformly parabolic equations, and the authors indicate the same strategy should extend to degenerate, hypoelliptic settings.

What carries the argument

The central object is the inversion lemma (Lemma 2.1), which expresses the difference of two push-forward semigroups acting on an initial law as the integral $\int \bar{\mu}_0(dx)\int_0^s \vec{P}^{\mu}_{0,t}(A^\mu_{t,\cdot}-A^\nu_{t,\cdot})\vec{P}^{\nu}_{t,s}f(x)\,dt$. This identity turns the contraction estimate into a bound on the difference of the infinitesimal generators, controlled by $M_\alpha(\mu,\nu)$, together with a bound on spatial derivatives of $\vec{P}^{\nu}_{t,s}f$, controlled by Gaussian estimates for the fundamental solution of a uniformly parabolic operator with Hölder coefficients. The dual-Hölder metric $m_\alpha$ is the norm induced on probability measures by test functions of bounded $\alpha$-Hölder norm; it makes the space of flows of marginals complete and yields the preliminary estimate $\|C^\mu-C^\nu\|_{\infty}+\|B^\mu-B^\nu\|_{\infty}\le cM_\alpha(\mu,\nu)$.

What would settle it

Find one pair of coefficients satisfying Assumptions 1.1 and 1.2 for which the required estimate $\|\partial_{x_ix_j}\vec{P}^{\nu}_{t,s}f\|_\infty\le c(s-t)^{-(1-\alpha/2)}[f]_{C^\alpha}$ fails for the frozen-flow transition density; then Lemma 2.1 and the contraction bound (3.4) would collapse, and Theorem 1.3 as proved would be false.

Watch

Extended reading notes

Core claim

Theorem 1.3 states that under Assumptions 1.1 and 1.2, for every $T>0$ and every initial law $\bar{\mu}_0\in\mathcal{P}(\mathbb{R}^d)$, there exists a unique weak solution of the McKean–Vlasov equation (1.1). The construction is a contraction mapping on $C([0,T];\mathcal{P}(\mathbb{R}^d))$ equipped with the metric $M_\alpha(\mu,\nu)=\max_{t\in[0,T]} m_\alpha(\mu_t,\nu_t)$, where $m_\alpha$ is the dual-Hölder distance between probability measures. For each frozen flow $\mu$, the linearized equation has a transition density $p^\mu$, and the map $\mu\mapsto [X^\mu_\cdot]$ is shown to satisfy $M_\alpha(\mu,\nu)\le c T^{\alpha/2} M_\alpha(\mu,\nu)$, giving a fixed point for small time horizons and then for all horizons by continuation. As a corollary, if $\sigma$ is additionally Lipschitz continuous in the spatial variable, the unique weak solution is in fact a unique strong solution.

Load-bearing premise

The proof assumes that classical parabolic theory delivers Gaussian upper bounds and derivative estimates for the transition densities of every linearized equation, uniformly over all frozen flows of measures, even when the coefficients are merely measurable in time.

Editorial extensions

If this is right

If the theorem is correct, every non-degenerate McKean–Vlasov equation with bounded $\alpha$-Hölder coefficients has a well-defined nonlinear flow of marginals, namely the unique fixed point of $\mu\mapsto [X^\mu_{\cdot}]$.
Adding Lipschitz continuity of $\sigma$ in the spatial variable upgrades the result from existence and uniqueness of weak solutions to existence and uniqueness of strong solutions.
The result holds for coefficients of the form $B(t,x,\mu)=\int b(t,x,y)\mu(dy)$ and $\Sigma(t,x,\mu)=\int \sigma(t,x,y)\mu(dy)$, and Remark 1.5 extends it to any bounded coefficients satisfying a joint Hölder condition in $(x,\mu)$ with respect to the dual-Hölder metric.
Because the proof relies only on Gaussian bounds for linearized parabolic operators, the same contraction scheme is expected to work for degenerate kinetic equations, such as Langevin-type systems, whenever matching Gaussian estimates are available for the hypoelliptic fundamental solution.
The coefficients are allowed to be merely measurable in time, not continuous, as long as they are bounded and Hölder continuous in the spatial variables.

Reading between the lines

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

A natural test of the method would be to check whether the same contraction estimate survives when the linearized equation is degenerate but hypoelliptic, since the paper's own extension plan points exactly to kinetic Langevin systems with rough coefficients.
Because the dual-Hölder metric is weaker than total variation or Wasserstein metrics with higher moments, the uniqueness statement is tied to this particular metric; switching to a stronger metric could break the contraction argument without necessarily breaking well-posedness in another sense.
The contraction constant scales like $cT^{\alpha/2}$, so for very small Hölder exponents the admissible time step shrinks; numerical or analytic continuation to a fixed horizon would then require many small steps, a practical issue the paper does not address.
The paper establishes weak well-posedness but only proves pathwise uniqueness when $\sigma$ is Lipschitz; an interesting open direction would be to determine whether merely Hölder $\sigma$ can admit more than one strong solution in this McKean–Vlasov setting.

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.

Desk Editor's Note

read the letter

The genuine novelty here is Theorem 1.3: weak well-posedness for non-degenerate McKean–Vlasov equations whose drift and diffusion are of interaction form, B=∫b dμ and Σ=∫σ dμ, with b,σ only α-Hölder in (x,y), and for any initial law. That is a real extension of Chaudru de Raynal, whose drift is β(t,x,∫φ dμ) and requires differentiability in the third argument. The proof strategy is right: freeze the measure, solve the linearized SDE classically, then invert the flow via the identity in Lemma 2.1, which avoids any derivative with respect to the measure. The Schauder-type estimate (3.4) then gives a contraction on the space of continuous flows of marginals. This is a short, elegant note and the central idea is sound.

The soft spots are real but mostly cosmetic. The most consequential is display (3.3): the right side, max_s sup_f I_s^{μ,ν}(f), is exactly M_α(Φ(μ),Φ(ν)), the distance between the two flows of laws, not M_α(μ,ν). As printed, combining (3.3) with (3.4) yields the circular M_α(μ,ν) ≤ c T^{α/2} M_α(μ,ν), which proves nothing. The fix is obvious: replace the left side by M_α(Φ(μ),Φ(ν)), and the same estimates give the genuine contraction for small T. This is a typographical error, not a hidden mathematical gap, but a referee must catch it.

Corollary 1.4 is asserted without proof; it likely follows from standard Yamada–Watanabe logic, but needs a few lines. The abstract and introduction promise applicability to hypoelliptic equations, but the body only says that is future work. That oversells slightly. The analytic inputs — Gaussian bounds, derivative estimates, potential estimates — are imported from Friedman and Pascucci's book rather than proved here; acceptable for a note, though the measurable-in-time footnote deserves scrutiny.

Who is this for? People working on McKean–Vlasov equations with rough coefficients, and anyone who wants a concise example of the fixed-point method for flows of marginals. It deserves a serious referee and, after fixing (3.3) and adding a word on Corollary 1.4, should be published.

Referee Report

2 major / 3 minor

Summary. The paper claims to give a streamlined proof of weak well-posedness for non-degenerate McKean-Vlasov SDEs with Hölder continuous coefficients (Assumptions 1.1 and 1.2), via a contraction argument on the space C([0,T];P(R^d)) equipped with the dual-Hölder metric M_α. For each frozen flow μ, the linearized SDE (2.1) has a transition density p^μ; Lemma 2.1 expresses the difference between two such flows through the inversion formula (2.3). Section 3 aims to show that the fixed-point map is a contraction for small T, using estimates (3.1)-(3.4). The intended argument is elegant, but the central identity (3.3) is misstated and, as printed, the proof does not establish the contraction.

Significance. If the proof is repaired, the paper provides a genuinely useful simplification of Chaudru de Raynal's well-posedness theorem: it avoids derivatives with respect to the measure variable, relies only on Gaussian estimates for uniformly parabolic PDEs, and is designed to extend to hypoelliptic settings. The inversion lemma is a nice adaptation of a formula of Kolokoltsov, and the contraction scheme has no fitted constants or normalization tricks. These strengths make the paper worth pursuing, but the misstated identity in Section 3 is a load-bearing error that must be corrected before the main theorem is established.

major comments (2)

[Section 3, Eq. (3.3)] The displayed identity is incorrect: the right-hand side max_{s∈[0,T]} sup_{‖f‖_{bC^α}≤1} I_s^{μ,ν}(f) equals M_α(Φ(μ),Φ(ν)), where Φ(μ)_t = [X_t^μ] and Φ(ν)_t = [X_t^ν], not M_α(μ,ν). As printed, combining (3.3) with (3.4) yields the circular inequality M_α(μ,ν) ≤ cT^{α/2}M_α(μ,ν), which proves nothing. The intended argument is recovered by replacing the left-hand side of (3.3) with M_α(Φ(μ),Φ(ν)) and then concluding from (3.4) that M_α(Φ(μ),Φ(ν)) ≤ cT^{α/2}M_α(μ,ν), which gives a genuine contraction for T small. The final sentence of Section 3 should also be revised to state this corrected contraction inequality explicitly.
[Section 2, paragraph after (2.2) and footnote 1] The proof relies on a package of parabolic estimates for the operator ∂_t + A^μ_{t,x} with coefficients B^μ, C^μ that are only bounded and α-Hölder in x and merely measurable in t, since Assumption 1.1 is in L∞([0,T]; bC^α). The needed estimates are: existence of a fundamental solution with Gaussian upper bounds, derivative bounds such as ‖∂_{x_i x_j} P^ν_{t,s} f‖_∞ ≤ c(s-t)^{-(1-α/2)}[f]_{C^α}, and potential estimates justifying the interchanges in Lemma 2.1, all uniformly over frozen flows μ. The manuscript cites [4] and [8] for these facts in the time-continuous case and relegates the measurable-in-time case to a footnote saying the proof 'proceeds in a similar manner'. This is load-bearing: if these uniform estimates fail, both Lemma 2.1 and the contraction bound (3.4) collapse. The authors should state a precise theorem covering the measurable-in-time case with uniform constants, give a reference that explicitly covers it, or modify Assumption 1.1 to include time-continuity.

minor comments (3)

[Lemma 2.1, proof] In the sentence defining test functions, 'for any test function ϕ ∈ C∞_0 ∈ ]t, T] × R^d' should read 'for any test function ϕ ∈ C∞_0(]t, T] × R^d)'.
[Remark 1.6 and Appendix] The symbol m_α is used for two different metrics: in (1.5) it is defined with the bC^α norm, while Remark 1.6 and the Appendix introduce a variant using only the C^α seminorm. This reuse of notation is confusing; please use a different symbol for the second metric, such as ṁ_α.
[Corollary 1.4] The corollary states strong well-posedness when σ is Lipschitz in x, but no proof or reference is given. Since B remains only Hölder in x, pathwise uniqueness is not a formal consequence of Theorem 1.3 alone; please add a precise statement or citation (for example, to Veretennikov's pathwise-uniqueness results) explaining how strong well-posedness follows.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the fixed-point proof rests on independent parabolic estimates; the misprinted equality in (3.3) is a correctness flaw, not a circular reduction.

full rationale

The argument is not circular. Theorem 1.3 is obtained by freezing the measure argument: for a flow μ one solves the linear SDE (2.1), obtaining a new flow Φ(μ)_s = [X^μ_s], and then proves that Φ is a contraction in the dual-Hölder metric M_α. The only identity that has a definitional flavor is (3.3), but as printed it is miswritten: its left side must be M_α(Φ(μ), Φ(ν)), because I_s^{μ,ν}(f) is defined in (2.3) as ∫ f d([X^μ_s] − [X^ν_s]). With that correction, the identity is exactly the definition of M_α applied to the pushed-forward flows, while the bound (3.4) is a separate estimate obtained from (3.1) and Gaussian derivative bounds. The written combination of the incorrect (3.3) with (3.4) gives only the vacuous inequality M_α(μ,ν) ≤ cT^{α/2}M_α(μ,ν); this is a correction/rigor problem in the exposition, not a case where the theorem's conclusion is assumed or fitted. The estimated quantities involve no fitted parameters, no data, and no normalization constants. The cited Gaussian estimates are attributed to Friedman [4] and to the textbook [8]; even though [8] is by one of the authors, the same results are also cited to [4], so the load-bearing analysis does not reduce to a self-citation. The inversion lemma draws inspiration from Kolokoltsov [5], but the formula (2.3) is proved in the paper from Kolmogorov equations and potential estimates. Footnote 1 defers the merely measurable-in-time case with a remark that the proof 'proceeds in a similar manner'; this is a possible technical gap, but it does not make the derivation circular. The self-citation [9] concerns future degenerate work and is not load-bearing for Theorem 1.3.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

The paper introduces no free parameters and no invented entities. The central claim rests entirely on four classical background inputs: uniform parabolic theory with Gaussian estimates (the main one), well-posedness of the linearized SDE, potential estimates for the inversion lemma, and completeness of the measure spaces. All are standard, but the first is under-specified in the paper, which relies on citations to [4] and [8] rather than stating the estimates.

assumptions (4)

standard math Uniformly parabolic PDE theory: the operator ∂_t + A^μ_{t,x}, with B^μ and C^μ bounded and α-Hölder in x (measurable in t), has a fundamental solution p^μ with two-sided Gaussian bounds, C^{1,2} regularity, and derivative estimates, uniform in the frozen flow μ.
Invoked at the start of Section 2 ('By the classical theory of uniformly parabolic PDEs...', citing [4] and [8]) and again in the bound (3.4). This is the main external input and is not stated precisely.
standard math The linearized (frozen) SDE (2.1) admits a unique weak solution X^μ, which is Markov and has a transition density p^μ, for bounded measurable drift B^μ and uniformly elliptic diffusion C^μ.
Used in Section 2 to define X^μ and p^μ; classical SDE and parabolic theory cited indirectly through [4] and [8].
standard math Potential estimates and Fubini-type interchanges for products of fundamental solutions, justified by Gaussian decay, allow differentiating ∫ p^μ(0,x;t,z)p^ν(t,z;s,y) dz under the integral sign in t.
Used throughout the proof of Lemma 2.1 ('by classical potential estimates', citing Friedman [4] Chapter 1 Section 3 and [8] Proposition 20.3.9).
standard math The spaces (P(R^d), m_α) and (P_α(R^d), m_α) are complete, so (C([0,T]; P(R^d)), M_α) is complete for the fixed point argument.
Proposition 4.1 in the appendix, citing Bogachev [1] and Villani [10]; the proof identifies m_α with a bounded-Lipschitz distance and m_α with W_{1,d_α}. This is needed for the Banach fixed point theorem in Section 3.

how reviews work

0 comments

Cite this review

Pith. "Pith review of McKean-Vlasov stochastic equations with H\"older coefficients." pith.science (2026). https://pith.science/paper/YA46BRFD

@misc{pith2026241200834,
  author       = {Pith},
  title        = {Pith review of: McKean-Vlasov stochastic equations with H\"older coefficients},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YA46BRFD}},
  note         = {Machine review of arXiv:2412.00834}
}

read the original abstract

This work revisits the well-posedness of non-degenerate McKean-Vlasov stochastic differential equations with H\"older continuous coefficients, recently established by Chaudru de Raynal. We provide a streamlined and direct proof that leverages standard Gaussian estimates for uniformly parabolic PDEs, bypassing the need for derivatives with respect to the measure argument and extending applicability to hypoelliptic PDEs under weaker assumptions.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

10 extracted references · 6 canonical work pages

[4]

Partial diﬀerential equations of parabolic type

Friedman, A. Partial diﬀerential equations of parabolic type . Prentice-Hall Inc., Englewood Cliﬀs, N.J., 1964

work page 1964
[8]

Probability Theory II – Stochastic Calculus , vol

Pascucci, A. Probability Theory II – Stochastic Calculus , vol. 166 of UNITEXT. Springer Cham, 2024

2024
[1]

Bogachev, V. I. Measure theory. Vol. I, II . Springer-Verlag, Berlin, 2007

work page 2007
[2]

Chaudru de Raynal, P. E. Strong well posedness of McKean-Vlasov stochastic diﬀerential e quations with H¨ older drift.Stochastic Process. Appl. 130 , 1 (2020), 79–107

2020
[3]

On a class of degenerate parabolic equations of Kolmogorov type

Di Francesco, M., and Pascucci, A. On a class of degenerate parabolic equations of Kolmogorov type. AMRX Appl. Math. Res. Express , 3 (2005), 77–116

2005
[5]

Kolokoltsov, V. N. Nonlinear diﬀusions and stable-like processes with coeﬃcients depen ding on the median or VaR. Appl. Math. Optim. 68 , 1 (2013), 85–98

work page 2013
[6]

Sur la th´ eorie du mouvement brownien

Langevin, P. Sur la th´ eorie du mouvement brownien. Compt. Rendus , 146 (1908), 530–533

work page 1908
[7]

Optimal regularity for degenerate Kolmogorov equations in non-divergence form with rough-in-time coeﬃcients

Lucertini, G., Pagliarani, S., and Pascucci, A. Optimal regularity for degenerate Kolmogorov equations in non-divergence form with rough-in-time coeﬃcients. J. Evol. Equ. 23 , 4 (2023), Paper No. 69, 37. 6

work page 2023

Show all 10 references

[9]

Pascucci, A., Rondelli, A., and Veretennikov, A. Y. Existence and uniqueness results for strongly degenerate McKean-Vlasov equations with rough coeﬃcie nts. Preprint arXiv: 2409.14451 (2024)

2024 arXiv
[10]

Optimal transport, vol

Villani, C. Optimal transport, vol. 338 of Grundlehren der mathematischen Wissenschaften [Funda- mental Principles of Mathematical Sciences] . Springer-Verlag, Berlin, 2009. Old and new. 7

2009

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [4]

Partial diﬀerential equations of parabolic type

Friedman, A. Partial diﬀerential equations of parabolic type . Prentice-Hall Inc., Englewood Cliﬀs, N.J., 1964

work page 1964

[2] [8]

Probability Theory II – Stochastic Calculus , vol

Pascucci, A. Probability Theory II – Stochastic Calculus , vol. 166 of UNITEXT. Springer Cham, 2024

2024

[3] [1]

Bogachev, V. I. Measure theory. Vol. I, II . Springer-Verlag, Berlin, 2007

work page 2007

[4] [2]

Chaudru de Raynal, P. E. Strong well posedness of McKean-Vlasov stochastic diﬀerential e quations with H¨ older drift.Stochastic Process. Appl. 130 , 1 (2020), 79–107

2020

[5] [3]

On a class of degenerate parabolic equations of Kolmogorov type

Di Francesco, M., and Pascucci, A. On a class of degenerate parabolic equations of Kolmogorov type. AMRX Appl. Math. Res. Express , 3 (2005), 77–116

2005

[6] [5]

Kolokoltsov, V. N. Nonlinear diﬀusions and stable-like processes with coeﬃcients depen ding on the median or VaR. Appl. Math. Optim. 68 , 1 (2013), 85–98

work page 2013

[7] [6]

Sur la th´ eorie du mouvement brownien

Langevin, P. Sur la th´ eorie du mouvement brownien. Compt. Rendus , 146 (1908), 530–533

work page 1908

[8] [7]

Optimal regularity for degenerate Kolmogorov equations in non-divergence form with rough-in-time coeﬃcients

Lucertini, G., Pagliarani, S., and Pascucci, A. Optimal regularity for degenerate Kolmogorov equations in non-divergence form with rough-in-time coeﬃcients. J. Evol. Equ. 23 , 4 (2023), Paper No. 69, 37. 6

work page 2023