pith. sign in

arxiv: 2606.22964 · v1 · pith:IK6AKPXNnew · submitted 2026-06-22 · 🧮 math.PR

Signature McKean-Vlasov stochastic differential equations

Fred Espen Benth , Salvador Ortiz-Latorre , Leonardo Tarquini This is my paper

Pith reviewed 2026-06-26 07:52 UTC · model grok-4.3

classification 🧮 math.PR
keywords McKean-Vlasov SDEssignaturesrough pathspropagation of chaosparticle systemspath-dependent SDEsexistence and uniqueness
0
0 comments X

The pith

Signature McKean-Vlasov SDEs have unique strong solutions and approximate general path-dependent versions.

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

The paper focuses on McKean-Vlasov SDEs whose coefficients are given by linear functionals of the expected signature of the geometric p-rough path lift of the solution, for p in (2,3). It first proves strong existence and uniqueness for solutions to these equations. It then shows that such signature-based equations can approximate arbitrary path-dependent McKean-Vlasov SDEs. Finally, it establishes propagation of chaos for the associated particle systems. A sympathetic reader would care because the signature transform packages path information into an object that makes the mean-field interaction tractable while still capturing dependence on the full trajectory.

Core claim

After establishing the strong existence and uniqueness of a solution, the authors prove how such an equation can approximate a general class of path-dependent McKean-Vlasov SDEs; they then consider the associated particle system and establish propagation of chaos.

What carries the argument

Linear functional of the expected signature of the geometric p-rough path lift of the solution, which determines the coefficients.

If this is right

  • Strong existence and uniqueness holds for the signature McKean-Vlasov equations under the stated conditions.
  • These signature equations can approximate any path-dependent McKean-Vlasov SDE in a suitable sense.
  • Propagation of chaos holds for the interacting particle systems that approximate the signature McKean-Vlasov limit.

Where Pith is reading between the lines

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

  • Truncations of the signature could yield practical finite-dimensional approximations whose error is controlled by the remainder of the signature series.
  • The same linear-functional dependence on expected signatures might extend the approximation technique to other classes of path-dependent mean-field equations driven by rough paths.
  • Propagation of chaos implies that empirical measures from finite-particle simulations converge to the law of the infinite-particle limit, enabling Monte-Carlo type schemes.

Load-bearing premise

The coefficients depend on the solution only through a linear functional of the expected signature of its geometric p-rough path lift.

What would settle it

An explicit coefficient functional built from the expected signature for which the corresponding McKean-Vlasov equation has either no strong solution or at least two distinct strong solutions would falsify the existence-uniqueness claim.

read the original abstract

McKean-Vlasov-type stochastic differential equations (SDEs) are characterized by coefficients depending on both the state and the law of the solution. In this work, we focus on a class of such equations where the coefficients depend on a linear combination of the expected signature of the geometric $p$-rough path lift of its solution, with $p\in(2,3)$. After establishing the strong existence and uniqueness of a solution, we prove how such an equation can approximate a general class of path-dependent McKean-Vlasov SDEs. Finally, we consider the associated particle system and propagation of chaos is established.

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

1 major / 1 minor

Summary. The manuscript introduces Signature McKean-Vlasov SDEs in which the coefficients depend on the state and on a linear functional of the expected signature of the geometric p-rough path lift of the solution (p ∈ (2,3)). After proving strong existence and uniqueness, the authors establish that this class approximates a general class of path-dependent McKean-Vlasov SDEs and prove propagation of chaos for the associated particle system.

Significance. If the approximation result is valid, the framework supplies a well-posed subclass of path-dependent mean-field equations whose coefficients are linear in averaged iterated integrals, together with a propagation-of-chaos theorem. This could enable analysis and simulation of certain path-dependent interactions via rough-path tools. The existence/uniqueness proof appears to rest on standard rough-path and SDE arguments.

major comments (1)
  1. [Approximation result / Theorem on approximation of general path-dependent MV-SDEs] Approximation theorem (the result claiming density in the class of path-dependent McKean-Vlasov equations): the argument appears to invoke the universal approximation property of signatures for individual paths, yet the functional is applied after taking expectation. Consequently the approximable functionals are linear in the expected signature coordinates. This does not automatically include nonlinear functionals of the law μ (e.g., variance of a signature component or dependence on the support of μ). A precise statement of the topology on the space of measures and an explicit density argument that accounts for the expectation step are required to support the central approximation claim.
minor comments (1)
  1. [Existence and uniqueness section] Clarify the precise regularity assumed on the linear functional l and verify that the rough-path lift exists under the stated coefficient conditions (the abstract mentions p ∈ (2,3) but does not detail the lift construction).

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their thorough and constructive review of our manuscript. We address the major comment below.

read point-by-point responses

  1. Referee: [Approximation result / Theorem on approximation of general path-dependent MV-SDEs] Approximation theorem (the result claiming density in the class of path-dependent McKean-Vlasov equations): the argument appears to invoke the universal approximation property of signatures for individual paths, yet the functional is applied after taking expectation. Consequently the approximable functionals are linear in the expected signature coordinates. This does not automatically include nonlinear functionals of the law μ (e.g., variance of a signature component or dependence on the support of μ). A precise statement of the topology on the space of measures and an explicit density argument that accounts for the expectation step are required to support the central approximation claim.

    Authors: We appreciate the referee highlighting this subtlety in the approximation result. The construction in the manuscript indeed relies on linear functionals of the expected signature, so the approximable coefficients are those linear in the expected signature coordinates. We will revise the manuscript to include a precise statement of the topology on the space of measures (the topology of convergence of all finite-dimensional marginals of the expected signature, which metrizes weak convergence under standard moment bounds) and an explicit density argument: the signature provides a dense linear span for continuous functionals on compact subsets of path space in the uniform topology; integrating against any fixed measure μ then yields approximation of the integrated functional by the corresponding linear functional of the expected signature. We agree that this linear structure does not capture nonlinear functionals of μ (such as variances or support-dependent quantities) and will update the description of the 'general class' of path-dependent McKean-Vlasov equations, as well as the statement of the approximation theorem, to reflect this scope accurately. These clarifications will appear in the revised version. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The derivation defines a restricted class of McKean-Vlasov SDEs whose coefficients are linear functionals of the expected signature of the geometric p-rough path lift, then invokes standard rough-path and SDE theory to obtain strong existence/uniqueness. The subsequent approximation result for general path-dependent McKean-Vlasov equations and the propagation-of-chaos statement for the particle system are presented as consequences of that theory rather than as identities or fitted quantities that reduce to the inputs by construction. No self-definitional steps, fitted-input predictions, or load-bearing self-citations appear in the stated claims; the argument chain remains self-contained against external benchmarks from rough-path analysis.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

Abstract-only review; cannot enumerate free parameters, axioms, or invented entities beyond what is stated. The central construction relies on existence of geometric p-rough path lifts and on the linear functional of the expected signature being well-defined.

axioms (2)
  • domain assumption Geometric p-rough path lift of the solution exists for p in (2,3)
    Invoked to define the signature that enters the coefficients.
  • domain assumption Coefficients are Lipschitz or satisfy standard conditions for SDE well-posedness once the signature functional is fixed
    Required for the claimed strong existence and uniqueness.

pith-pipeline@v0.9.1-grok · 5630 in / 1447 out tokens · 19087 ms · 2026-06-26T07:52:44.770827+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

50 extracted references · 4 canonical work pages

  1. [1]

    Numerical approximation of McKean–Vlasov SDEs via stochastic gradient descent

    A. Agarwal, A. Amato, S. Pagliarani, and G. dos Reis. “Numerical approximation of McKean–Vlasov SDEs via stochastic gradient descent”.Advances in Applied Probability (2026), pp. 1–33

    2026

  2. [2]

    Rate of convergence of a particle method to the solution of the McKean–Vlasov equation

    F. Antonelli and A. Kohatsu-Higa. “Rate of convergence of a particle method to the solution of the McKean–Vlasov equation”.The Annals of Applied Probability12.2 (2002), pp. 423–476

    2002

  3. [3]

    Fundamentals of Stochastic Filtering

    A. Bain and D. Crisan. “Fundamentals of Stochastic Filtering”. In:Stochastic Modelling and Applied Probability Series. Vol. 60. 2009

    2009

  4. [4]

    Mean-field description and propa- gation of chaos in networks of Hodgkin-Huxley and FitzHugh-Nagumo neurons

    J. Baladron, D. Fasoli, O. Faugeras, and J. Touboul. “Mean-field description and propa- gation of chaos in networks of Hodgkin-Huxley and FitzHugh-Nagumo neurons”.Journal of Mathematical Neuroscience2.1 (2012)

    2012

  5. [5]

    Baldi.Stochastic Calculus

    P. Baldi.Stochastic Calculus. An Introduction Through Theory and Exercises. Springer Cham, 2017

    2017

  6. [6]

    Projected particle methods for solving McKean– Vlasov stochastic differential equations

    D. Belomestny and J. Schoenmakers. “Projected particle methods for solving McKean– Vlasov stochastic differential equations”.SIAM Journal on Numerical Analysis56.6 (2018), pp. 3169–3195

    2018

  7. [7]

    The signature of a rough path: unique- ness

    H. Boedihardjo, X. Geng, T. Lyons, and D. Yang. “The signature of a rough path: unique- ness”.Advances in Mathematics293 (2016), pp. 720–737

    2016

  8. [8]

    A stochastic particle method for the McKean–Vlasov and the Burgers equation

    M. Bossy and D. Talay. “A stochastic particle method for the McKean–Vlasov and the Burgers equation”.Mathematics of Computation66.217 (1997), pp. 157–192

    1997

  9. [9]

    Carmona and F

    R. Carmona and F. Delarue.Probabilistic Theory of Mean Field Games with Applications I. Mean Field FBSDEs, Control, and Games. 1st ed. Springer, 2018. 42

    2018

  10. [10]

    Carmona and F

    R. Carmona and F. Delarue.Probabilistic Theory of Mean Field Games with Applications II. Mean Field Games with Common Noise and Master Equations. 1st ed. Springer, 2018

    2018

  11. [11]

    Mean field games and systemic risk

    R. Carmona, J. Fouque, and L. Sun. “Mean field games and systemic risk”.Communica- tions in Mathematical Sciences13.4 (2015), pp. 911–933

    2015

  12. [12]

    Convergence analysis of machine learning algorithms for the numerical solution of mean field control and games I: the ergodic case

    R. Carmona and M. Lauri´ ere. “Convergence analysis of machine learning algorithms for the numerical solution of mean field control and games I: the ergodic case”.SIAM Journal on Numerical Analysis59.3 (2021), pp. 1455–1485

    2021

  13. [13]

    Convergence analysis of machine learning algorithms for the numerical solution of mean field control and games II: the finite horizon case

    R. Carmona and M. Lauri` ere. “Convergence analysis of machine learning algorithms for the numerical solution of mean field control and games II: the finite horizon case”.The Annals of Applied Probability32.6 (2022), pp. 4065–4105

    2022

  14. [14]

    Probabilistic approach for granular media equa- tions in the non-uniformly convex case

    P. Cattiaux, A. Guillin, and F. Malrieu. “Probabilistic approach for granular media equa- tions in the non-uniformly convex case”.Probability Theory and Related Fields140.1-2 (2008), pp. 19–40

    2008

  15. [15]

    Propagation of chaos: a review of models, methods and ap- plications. I. Models and methods

    L. Chaintron and A. Diez. “Propagation of chaos: a review of models, methods and ap- plications. I. Models and methods”.Kinetic and Related Models15.6 (2022), pp. 895– 1015

    2022

  16. [16]

    A cubature based algorithm to solve decoupled McKean–Vlasov forward–backward stochastic differential equations

    P.E. Chaudru de Raynal and C.A. Garcia Trillos. “A cubature based algorithm to solve decoupled McKean–Vlasov forward–backward stochastic differential equations”.Stochastic Processes and Their Applications125.6 (2015), pp. 2206–2255

    2015

  17. [17]

    Iterated path integrals

    K. T. Chen. “Iterated path integrals”.Bulletin of the American Mathematical Society83.5 (1977), pp. 831–879

    1977

  18. [18]

    Integration of paths, geometric invariants and a generalized Baker-Hausdorff formula

    K.T. Chen. “Integration of paths, geometric invariants and a generalized Baker-Hausdorff formula”.Annals of Mathematics65.1 (1957), pp. 163–178

    1957

  19. [19]

    Chevyrev and A

    I. Chevyrev and A. Kormilitzin.A primer on the signature method in machine learning

  20. [20]

    arXiv:1603.03788 [stat.ML]

    work page arXiv

  21. [21]

    Signature moments to characterize laws of stochastic processes

    I. Chevyrev and H. Oberhauser. “Signature moments to characterize laws of stochastic processes”.Journal of Machine Learning Research23.1 (2022)

    2022

  22. [22]

    Cuchiero and J

    C. Cuchiero and J. M¨ oller.Polynomial McKean–Vlasov SDEs. 2025. arXiv:2502.19203 [math.PR]

    work page arXiv 2025

  23. [23]

    Universal approximation theorems for continuous functions of c` adl` ag paths and L´ evy-type signature models

    C. Cuchiero, F. Primavera, and S. Svaluto-Ferro. “Universal approximation theorems for continuous functions of c` adl` ag paths and L´ evy-type signature models”.Finance and Stochastics29.2 (2025), pp. 289–342

    2025

  24. [24]

    Global universal approximation of func- tional input maps on weighted spaces

    C. Cuchiero, P. Schmocker, and J. Teichmann. “Global universal approximation of func- tional input maps on weighted spaces”.Constructive Approximation(2026)

    2026

  25. [25]

    Cuchiero, S

    C. Cuchiero, S. Svaluto-Ferro, and J. Teichmann.Signature SDEs from an affine and polynomial perspective. 2025. arXiv:2302.01362 [math.PR]

    work page arXiv 2025

  26. [26]

    Global solvability of a networked integrate-and-fire model of McKean–Vlasov type

    F. Delarue, J. Inglis, S. Rubenthaler, and E. Tanr´ e. “Global solvability of a networked integrate-and-fire model of McKean–Vlasov type”.The Annals of Applied Probability25.4 (2015), pp. 2096–2133

    2015

  27. [27]

    Fristedt and L

    B. Fristedt and L. F. Gray.A Modern Approach to Probability Theory. 1st ed. Birkh¨ auser, 1997

    1997

  28. [28]

    P. K. Friz and M. Hairer.A Course on Rough Paths: With an Introduction to Regularity Structures. 2nd Edition. Springer Nature, 2020

    2020

  29. [29]

    P. K. Friz and N. B. Victoir.Multidimensional Stochastic Processes as Rough Paths: Theory and Applications. Cambridge University Press, 2010

    2010

  30. [30]

    Analytical approximations of non-linear SDEs of McK- ean–Vlasov type

    E. Gobet and S. Pagliarani. “Analytical approximations of non-linear SDEs of McK- ean–Vlasov type”.Journal of Mathematical Analysis and Applications466.1 (2018), pp. 71– 106

    2018

  31. [31]

    Noisy bounded confidence models for opinion dynamics: the effect of boundary conditions on phase transitions

    B.D. Goddard, B. Gooding, H. Short, and G.A. Pavliotis. “Noisy bounded confidence models for opinion dynamics: the effect of boundary conditions on phase transitions”. IMA Journal of Applied Mathematics87.1 (2022), pp. 80–110. 43

    2022

  32. [32]

    On the mean field and classical limits of quantum mechanics

    F. Golse, C. Mouhot, and T. Paul. “On the mean field and classical limits of quantum mechanics”.Communications in Mathematical Physics343.1 (2016), pp. 165–205

    2016

  33. [33]

    Hyperbolic groups

    M. Gromov. “Hyperbolic groups”. In:Essays in Group Theory. Ed. by S. M. Gersten. Springer New York, 1987, pp. 75–263

    1987

  34. [34]

    Control of McKean–Vlasov SDEs with contagion through killing at a state-dependent intensity

    B. Hambly and P. Jettkant. “Control of McKean–Vlasov SDEs with contagion through killing at a state-dependent intensity”.The Annals of Applied Probability35.6 (2025)

    2025

  35. [35]

    Solving high-dimensional partial differential equations using deep learning

    J. Han, A. Jentzen, and W. E. “Solving high-dimensional partial differential equations using deep learning”.Proceedings of the National Academy of Sciences - PNAS115.34 (2018), pp. 8505–8510

    2018

  36. [36]

    Multilevel Picard approximations for McKean–Vlasov stochastic differential equations

    M. Hutzenthaler, T. Kruse, and T.A. Nguyen. “Multilevel Picard approximations for McKean–Vlasov stochastic differential equations”.Journal of Mathematical Analysis and Applications507.1 (2022)

    2022

  37. [37]

    Karatzas and S

    I. Karatzas and S. Shreve.Brownian Motion and Stochastic Calculus. 2nd ed. Springer New York, 1991

    1991

  38. [38]

    La variation d’ordre p des semi-martingales

    D. Lepingle. “La variation d’ordre p des semi-martingales”.Zeitschrift f¨ ur Wahrschein- lichkeitstheorie und Verwandte Gebiete36.4 (1976), pp. 295–316

    1976

  39. [39]

    Differential equations driven by rough signals

    T. Lyons. “Differential equations driven by rough signals”.Revista Matem´ atica Iberoamer- icana14.2 (1998), pp. 215–310

    1998

  40. [40]

    Lyons and Z

    T. Lyons and Z. Qian.System Control and Rough Paths. Oxford University Press, 2002

    2002

  41. [41]

    Logarithmic Sobolev inequalities for some nonlinear PDE’s

    F. Malrieu. “Logarithmic Sobolev inequalities for some nonlinear PDE’s”.Stochastic Pro- cesses and Their Applications95.1 (2001), pp. 109–132

    2001

  42. [42]

    Propagation of chaos for a class of non-linear parabolic equations

    H. P. McKean. “Propagation of chaos for a class of non-linear parabolic equations”. In: Stochastic Differential Equations. Vol. Session 7. Lecture Series in Differential Equations. Washington, DC: Catholic University, 1967, pp. 41–57

    1967

  43. [43]

    A mean field view of the landscape of two-layer neural networks

    S. Mei, A. Montanari, and P. Nguyen. “A mean field view of the landscape of two-layer neural networks”.Proceedings of the National Academy of Sciences115.33 (2018), E7665– E7671

    2018

  44. [44]

    Asymptotic behaviour of some interacting particle systems; McKean–Vlasov and Boltzmann models

    S. M´ el´ eard, L. Tubaro, and D. Talay. “Asymptotic behaviour of some interacting particle systems; McKean–Vlasov and Boltzmann models”. In:Probabilistic Models for Nonlinear Partial Differential Equations. Lecture Notes in Mathematics. Springer Berlin Heidelberg, 2006, pp. 42–95

    2006

  45. [45]

    Morale, L

    D. Morale, L. Tarquini, and S. Ugolini.A probabilistic interpretation of a non-conservative and path-dependent nonlinear reaction-advection-diffusion system. 2025. arXiv:2407 . 19301 [math.PR]

    2025

  46. [46]

    Morale, L

    D. Morale, L. Tarquini, and S. Ugolini.Killed path-dependent McKean–Vlasov SDEs for a probabilistic representation of non-conservative McKean PDEs. 2025. arXiv:2507.23353

    work page arXiv 2025

  47. [47]

    Algorithm 1004: the iisignature library: efficient calcu- lation of iterated-integral signatures and log signatures

    J.F. Reizenstein and B. Graham. “Algorithm 1004: the iisignature library: efficient calcu- lation of iterated-integral signatures and log signatures”.ACM Transactions on Mathe- matical Software46.1 (2020), pp. 1–21

    2020

  48. [48]

    Topics in propagation of chaos

    A. Sznitman. “Topics in propagation of chaos”. In:Ecole d’Et´ e de Probabilit´ es de Saint- Flour XIX-1989. Lecture Notes in Mathematics. Springer Berlin Heidelberg, 2006, pp. 165– 251

    1989

  49. [49]

    A new McKean–Vlasov stochastic interpretation of the parabolic-parabolic Keller-Segel model: the two-dimensional case

    M. Tomasevic. “A new McKean–Vlasov stochastic interpretation of the parabolic-parabolic Keller-Segel model: the two-dimensional case.”The Annals of Applied Probability31.1 (2021), pp. 432–459

    2021

  50. [50]

    Villani.Optimal Transport: Old and New

    C. Villani.Optimal Transport: Old and New. Springer-Verlag, 2009. 44

    2009