pith. sign in

arxiv: 2604.11385 · v1 · submitted 2026-04-13 · 🧮 math.PR

Quantitative Large Population Limit for Non Exchangeable Diffusions in Fisher Information

Jules Grass (ICJ , UCBL) This is my paper

Pith reviewed 2026-05-10 15:35 UTC · model grok-4.3

classification 🧮 math.PR
keywords non-exchangeable particle systemsgraphon mean-field limitsrelative Fisher informationrelative entropylarge population limitsdiffusive interactionsstability estimates
0
0 comments X

The pith

Non-exchangeable diffusive particle systems with converging interaction matrices are approximated quantitatively in Fisher information by independent projections that reduce to graphon mean-field limits.

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

The paper shows that when an interaction matrix for N diffusive particles converges to a graphon, the marginal laws of the full system stay close in relative Fisher information to those of a simpler independent projection system. This equivalence then lets the authors derive explicit stability bounds for the corresponding graphon mean-field system, both in relative entropy and in Fisher information. A reader cares because the result supplies rates that control how heterogeneous, non-symmetric interactions produce a tractable large-population limit without forcing all particles to be exchangeable.

Core claim

When the interaction matrix converges suitably to a graphon, quantitative bounds hold on the relative Fisher information between the marginal laws of the original non-exchangeable particle system and those of the independent projection system. The independent projection system is equivalent to a graphon mean-field system whose relative entropy and Fisher information satisfy quantitative stability estimates as the number of particles grows.

What carries the argument

The equivalence between the independent projection system and the graphon mean-field system, which converts relative Fisher information approximation into quantitative stability estimates for the large-population limit.

If this is right

  • Relative Fisher information between the original particle marginals and the independent projection vanishes at an explicit rate when the matrix converges to the graphon.
  • The graphon mean-field system inherits quantitative stability bounds in both relative entropy and Fisher information.
  • The large-population limit for non-exchangeable diffusions therefore holds with explicit control in Fisher information.
  • The same reduction yields stability estimates that apply uniformly to families of graphons satisfying the convergence condition.

Where Pith is reading between the lines

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

  • The same quantitative bounds may be adaptable to other information functionals such as relative entropy directly on the particle system.
  • Numerical checks on finite-N systems with a fixed graphon limit would confirm whether the predicted rates match observed convergence in Fisher information.
  • The framework could apply to heterogeneous network models in biology or social dynamics where exchangeability is unrealistic.

Load-bearing premise

The interaction matrix of the particle system converges in a suitable sense to a graphon.

What would settle it

A concrete counterexample in which the interaction matrix converges to a graphon yet the relative Fisher information between the particle marginals and the independent projection fails to vanish at the claimed quantitative rate.

read the original abstract

This paper builds upon the methods developed in [22] and [15] to investigate the large population behavior of non exchangeable systems of N diffusive particles when the interaction matrix converges (in some sense) to a graphon. We first prove that the particle system is well approximated in Fisher information by the so-called independent projection system by proving quantitative bounds on the relative Fisher information between the marginal laws of both systems. We then use a convenient equivalence between the independent projection system and a graphon mean field system to investigate its large population behavior by proving quantitative stability estimates for graphon mean field systems in both relative entropy and Fisher information.

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 manuscript proves quantitative bounds on the relative Fisher information between the marginal laws of an N-particle non-exchangeable diffusion system and an independent projection system. It then invokes an equivalence between the independent projection system and a graphon mean-field system to derive quantitative stability estimates for the large-population limit in both relative entropy and Fisher information, under the assumption that the interaction matrix converges in a suitable sense to a graphon.

Significance. If the quantitative estimates hold with explicit rates that incorporate the graphon convergence modulus, the work would strengthen the toolkit for non-exchangeable mean-field limits by providing Fisher-information control that complements existing entropy-based approaches. The reduction via the independent projection system is a potentially useful technical device for handling heterogeneity.

major comments (2)
  1. [Abstract and §1] Abstract and §1: The quantitative large-population claims rest on the interaction matrix converging to a graphon, yet the error terms in the relative Fisher information bounds between the particle marginals and the independent projection system appear to treat this convergence as a fixed background assumption rather than inserting an explicit rate or modulus of continuity. If the convergence holds only qualitatively or in a topology weaker than the one controlling the Fisher information, the claimed N-dependent quantitative limit does not follow uniformly.
  2. [Main results on equivalence and stability] The equivalence between the independent projection system and the graphon mean-field system is invoked to transfer stability estimates in relative entropy and Fisher information. It is not clear whether this equivalence is exact or introduces additional error terms whose size must be controlled quantitatively in terms of N and the graphon approximation; without an explicit accounting, the stability estimates for the original particle system may lose their quantitative character.
minor comments (2)
  1. Clarify the precise mode of graphon convergence (e.g., cut norm, L^2, or pointwise) already in the statement of the main theorems so that readers can immediately assess compatibility with the Fisher-information topology.
  2. Ensure that all constants appearing in the quantitative bounds are tracked explicitly with respect to the graphon and the dimension, even if they are not optimal.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and insightful comments on our manuscript. We address the two major comments point by point below, providing clarifications and indicating revisions where appropriate to strengthen the quantitative aspects of the results.

read point-by-point responses

  1. Referee: [Abstract and §1] Abstract and §1: The quantitative large-population claims rest on the interaction matrix converging to a graphon, yet the error terms in the relative Fisher information bounds between the particle marginals and the independent projection system appear to treat this convergence as a fixed background assumption rather than inserting an explicit rate or modulus of continuity. If the convergence holds only qualitatively or in a topology weaker than the one controlling the Fisher information, the claimed N-dependent quantitative limit does not follow uniformly.

    Authors: We appreciate the referee highlighting this point. The relative Fisher information bounds between the N-particle marginals and the independent projection system (Theorem 2.3) are quantitative in N for any fixed interaction matrix and do not depend on the graphon limit. The graphon convergence assumption enters only when combining these bounds with the stability estimates for the graphon mean-field system to obtain the overall large-population limit. To ensure the N-dependent quantitative character is fully explicit and uniform, we will revise the abstract, Section 1, and the statements of Theorems 2.3, 4.1, and 4.2 to insert the modulus of continuity for the graphon convergence (in the appropriate topology) directly into the final error terms. This makes the dependence on both N and the graphon approximation rate transparent. revision: yes

  2. Referee: [Main results on equivalence and stability] The equivalence between the independent projection system and the graphon mean-field system is invoked to transfer stability estimates in relative entropy and Fisher information. It is not clear whether this equivalence is exact or introduces additional error terms whose size must be controlled quantitatively in terms of N and the graphon approximation; without an explicit accounting, the stability estimates for the original particle system may lose their quantitative character.

    Authors: We thank the referee for this observation. By construction, the equivalence between the independent projection system and the graphon mean-field system is exact: the independent projection is defined so that its marginal laws coincide precisely with the product measure of the graphon mean-field marginals (see Definition 3.1 and Proposition 3.2). Consequently, no additional error terms arise in the transfer, and the quantitative stability estimates in relative entropy and Fisher information for the graphon system (Theorems 4.1 and 4.2) apply directly without loss of quantitative character. We will add a short clarifying remark after Proposition 3.2 in the revised manuscript to explicitly state this exactness and its role in preserving the rates. revision: yes

Circularity Check

0 steps flagged

No significant circularity; new quantitative bounds derived independently

full rationale

The paper's derivation proceeds by establishing quantitative relative Fisher information bounds between the N-particle marginals and the independent projection system, followed by stability estimates for the equivalent graphon mean-field limit in relative entropy and Fisher information. These steps rely on the external assumption that the interaction matrix converges to a graphon, which is invoked as a hypothesis rather than derived or fitted within the paper. Prior works [22] and [15] supply background methods, but the central quantitative estimates and equivalence arguments are presented as fresh results without reducing by construction to self-citations, self-definitions, or renamed inputs. The chain remains self-contained against the stated assumptions.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The abstract provides limited detail on background assumptions; the main ones are standard existence results for SDEs and the graphon convergence hypothesis needed for the limit.

axioms (2)
  • standard math Solutions to the underlying stochastic differential equations for the particle system and mean-field limits exist and are unique under the stated conditions.
    Invoked implicitly to define the marginal laws whose Fisher information is compared.
  • domain assumption The interaction matrix converges to a graphon in a topology that permits passage to the independent projection and graphon mean-field systems.
    Explicitly stated as the setting in which the large-population behavior is investigated.

pith-pipeline@v0.9.0 · 5394 in / 1570 out tokens · 63926 ms · 2026-05-10T15:35:04.801207+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

31 extracted references · 31 canonical work pages

  1. [1]

    Bayraktar, S

    E. Bayraktar, S. Chakraborty, and R. Wu. Graphon mean fiel d systems. Ann. Appl. Probab. , 33(5):3587–3619, 2023

    work page 2023

  2. [2]

    Bayraktar and D

    E. Bayraktar and D. Kim. Concentration of measure for gra phon particle system. Advances in Applied Probability, 56(4):1279–1306, Jan. 2024

    work page 2024

  3. [3]

    Bayraktar and R

    E. Bayraktar and R. Wu. Stationarity and uniform in time c onvergence for the graphon particle system. Stochastic Processes Appl. , 150:532–568, 2022

    work page 2022

  4. [4]

    G. Bet, F. Coppini, and F. R. Nardi. Weakly interacting os cillators on dense random graphs. Journal of Applied Probability , 61(1):255–278, June 2023

    work page 2023

  5. [5]

    Bhamidi, A

    S. Bhamidi, A. Budhiraja, and R. Wu. Weakly interacting p article systems on inhomogeneous random graphs. Stochastic Processes and their Applications , 129(6):2174–2206, 2019

    work page 2019

  6. [6]

    Budhiraja and R

    A. Budhiraja and R. Wu. Some fluctuation results for weakl y interacting multi-type particle systems. Stochastic Processes Appl. , 126(8):2253–2296, 2016

    work page 2016

  7. [7]

    Chaintron, G

    L.-P. Chaintron, G. Conforti, and K. Eichinger. Propaga tion of weak log-concavity along generalised heat flows via hamilton-jacobi equations, 2025

    work page 2025

  8. [8]

    Chaintron and A

    L.-P. Chaintron and A. Diez. Propagation of chaos: A revi ew of models, methods and applications. i. models and methods. Kinetic and Related Models , 15(6):895, 2022. QUANTITATIVE LARGE POPULATION LIMIT FOR NON EXCHANGEABLE D IFFUSIONS IN FISHER INFORMATION 21

    work page 2022

  9. [9]

    Chaintron and A

    L.-P. Chaintron and A. Diez. Propagation of chaos: A revi ew of models, methods and applications. ii. applications. Kinetic and Related Models , 15(6):1017, 2022

    work page 2022

  10. [10]

    F. Collet. Macroscopic limit of a bipartite curie–weis s model: A dynamical approach. Journal of Statistical Physics , 157(6):1301–1319, Sept. 2014

    work page 2014

  11. [11]

    Contucci, I

    P. Contucci, I. Gallo, and G. Menconi. Phase transition s in social science: two-population mean field theory. International Journal of Modern Physics B , 22(14):2199–2212, June 2008

    work page 2008

  12. [12]

    Coppini, A

    F. Coppini, A. De Crescenzo, and H. Pham. Nonlinear grap hon mean-field systems. Stochastic Processes Appl., 190:19, 2025. Id/No 104728

    work page 2025

  13. [13]

    Delattre, G

    S. Delattre, G. Giacomin, and E. Luçon. A note on dynamic al models on random graphs and fokker–planck equations. Journal of Statistical Physics , 165(4):785–798, Nov. 2016

    work page 2016

  14. [14]

    Friedman

    A. Friedman. Partial Differential Equations of Parabolic Type . R.E. Krieger Publishing Company, 1983

    work page 1983

  15. [15]

    Grass, A

    J. Grass, A. Guillin, and C. Poquet. Propagation of chao s in fisher information, 2025

    work page 2025

  16. [16]

    Grass, A

    J. Grass, A. Guillin, and C. Poquet. Sharp propagation o f chaos for mckean-vlasov equation with non constant diffusion coefficient. Electronic Communications in Probability , 30:1–12, 2025

    work page 2025

  17. [17]

    Jabin, D

    P. Jabin, D. Poyato, and J. Soler. Mean-field limit of non -exchangeable systems. Communications on Pure and Applied Mathematics , 78(4):651–741, Nov. 2024

    work page 2024

  18. [18]

    Jabin and Z

    P.-E. Jabin and Z. Wang. Quantitative estimates of prop agation of chaos for stochastic systems with W − 1,∞ kernels. Inventiones mathematicae, 214(1):523–591, July 2018

    work page 2018

  19. [19]

    H. Kunita. Stochastic differential equations and stoch astic flows of diffeomorphisms. In P. L. Hennequin, editor, École d’Été de Probabilités de Saint-Flour XII - 1982 , pages 143–303, Berlin, Heidelberg, 1984. Springer Berlin Heidelberg

    work page 1982

  20. [20]

    D. Lacker. Hierarchies, entropy, and quantitative pro pagation of chaos for mean field diffusions. Prob- ability and Mathematical Physics , 4(2):377–432, May 2023

    work page 2023

  21. [21]

    Lacker and L

    D. Lacker and L. Le Flem. Sharp uniform-in-time propaga tion of chaos. Probability Theory and Related Fields, 187(1-2):443–480, 2023

    work page 2023

  22. [22]

    Lacker, L

    D. Lacker, L. C. Yeung, and F. Zhou. Quantitative propag ation of chaos for non-exchangeable diffusions via first-passage percolation, 2024

    work page 2024

  23. [23]

    L. Lovász. Large Networks and Graph Limits . American Mathematical Society colloquium publications. American Mathematical Society, 2012

    work page 2012

  24. [24]

    Lovász and B

    L. Lovász and B. Szegedy. Limits of dense graph sequence s. J. Comb. Theory, Ser. B , 96(6):933–957, 2006

    work page 2006

  25. [25]

    Muntean and F

    A. Muntean and F. Toschi. Collective Dynamics from Bacteria to Crowds: An Excursion T hrough Modeling, Analysis and Simulation , volume 553. 01 2014

    work page 2014

  26. [26]

    Nadtochiy and M

    S. Nadtochiy and M. Shkolnikov. Mean field systems on net works, with singular interaction through hitting times. Ann. Probab., 48(3):1520–1556, 2020

    work page 2020

  27. [27]

    Naldi, L

    G. Naldi, L. Pareschi, and G. Toscani, editors. Mathematical modeling of collective behavior in socio- economic and life sciences . Model. Simul. Sci. Eng. Technol. Boston, MA: Birkhäuser, 2 010

    work page

  28. [28]

    R. I. Oliveira and G. H. Reis. Interacting diffusions on r andom graphs with diverging average degrees: Hydrodynamics and large deviations. Journal of Statistical Physics , 176(5):1057–1087, July 2019

    work page 2019

  29. [29]

    Sirignano and K

    J. Sirignano and K. Spiliopoulos. Mean field analysis of neural networks: a law of large numbers. SIAM J. Appl. Math. , 80(2):725–752, 2020

    work page 2020

  30. [30]

    C. Villani. Fisher information in kinetic theory, 2025 . https://arxiv.org/abs/2501.00925

    work page arXiv 2025

  31. [31]

    S. Wang. Sharp local propagation of chaos for mean field p articles with w− 1,∞ kernels. Journal of Functional Analysis, page 111240, Oct. 2025. Université Claude Bernard Lyon 1, CNRS UMR 5208, Institut Ca mille Jordan, 69622 Villeurbanne, France. grass@math.univ-lyon1.fr

    work page 2025