pith. sign in

arxiv: 1906.10232 · v1 · pith:JO5MIG6Vnew · submitted 2019-06-24 · 🧮 math.NA · cs.NA

Mean-field limit of interacting 2D nonlinear stochastic spiking neurons

Benjamin Aymard , Fabien Campillo , Romain Veltz This is my paper

Pith reviewed 2026-05-25 16:48 UTC · model grok-4.3

classification 🧮 math.NA cs.NA
keywords mean-field limitspiking neuronspropagation of chaosHopf bifurcationstochastic neural networksfinite volume method
0
0 comments X

The pith

Networks of stochastic spiking neurons converge to a mean-field description that develops synchronized activity through a Hopf bifurcation at high connectivity.

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

The paper presents a nonlinear stochastic model for networks of interacting 2D spiking neurons and derives its mean-field limit through a heuristic approach. It introduces a Monte Carlo method to simulate the full microscopic system and a finite volume scheme for the mean-field equations that maintains conservation and positivity. Numerical tests show that as the number of neurons grows, individual neuron behaviors exhibit propagation of chaos and converge to the mean-field solution. The mean-field model further displays a Hopf bifurcation when connectivity strength passes a threshold, which corresponds to the emergence of synchronized firing across the population. These results indicate that mean-field reductions can capture essential collective dynamics in large neural networks.

Core claim

The authors construct a 2D nonlinear stochastic model of spiking neurons with interactions, heuristically obtain its mean-field limit, and demonstrate through simulation that the finite network converges to this limit as size increases, with the mean-field equations exhibiting a Hopf bifurcation that signals synchronized activity once connectivity exceeds a critical value.

What carries the argument

The heuristically derived mean-field limit of the interacting 2D nonlinear stochastic spiking neuron network, discretized by an upwind implicit finite volume method that enforces numerical conservation and positivity.

If this is right

  • Propagation of chaos holds as the network size tends to infinity.
  • Trajectories of individual neurons converge to the mean-field description.
  • The mean-field system undergoes a Hopf bifurcation at sufficiently high connectivity, producing synchronized activity.
  • The finite volume discretization preserves the conservation and positivity properties of the continuous mean-field model, guaranteeing a unique numerical solution.

Where Pith is reading between the lines

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

  • Simplified mean-field equations may suffice to predict collective synchronization without tracking every neuron individually in very large populations.
  • The connectivity threshold for the bifurcation could be used to explore how network parameters control the switch between asynchronous and oscillatory regimes.
  • The same heuristic derivation and numerical approach might extend to variants of the model that incorporate different nonlinear response functions or additional spatial dimensions.

Load-bearing premise

The heuristic steps used to pass from the finite network of interacting neurons to the mean-field equations remain valid in the infinite-size limit.

What would settle it

A sequence of direct Monte Carlo simulations with steadily increasing network size that stops converging to the finite volume solution of the mean-field equations, or that shows no Hopf bifurcation even at high connectivity, would disprove the claimed limit behavior.

read the original abstract

In this work, we propose a nonlinear stochastic model of a network of stochastic spiking neurons. We heuristically derive the mean-field limit of this system. We then design a Monte Carlo method for the simulation of the microscopic system, and a finite volume method (based on an upwind implicit scheme) for the mean-field model. The finite volume method respects numerical versions of the two main properties of the mean-field model, conservation and positivity, leading to existence and uniqueness of a numerical solution. As the size of the network tends to infinity, we numerically observe propagation of chaos and convergence from an individual description to a mean-field description. Numerical evidences for the existence of a Hopf bifurcation (synonym of synchronised activity) for a sufficiently high value of connectivity, are provided.

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

3 major / 1 minor

Summary. The paper proposes a 2D nonlinear stochastic model for networks of spiking neurons. It heuristically derives the mean-field limit, implements a Monte Carlo scheme for the particle system and an upwind implicit finite-volume discretization for the resulting PDE (which preserves positivity and conservation, yielding existence/uniqueness of the discrete solution), and reports numerical observations of propagation of chaos together with convergence to the mean-field description as network size tends to infinity. Numerical evidence is also given for a Hopf bifurcation (interpreted as onset of synchronized activity) at sufficiently large connectivity.

Significance. If the numerical observations are free of discretization artifacts, the work supplies concrete evidence for mean-field limits and synchronization transitions in this class of stochastic neuron models. The structure-preserving finite-volume scheme is a methodological strength that could be useful beyond this specific application.

major comments (3)
  1. [Numerical experiments (Monte Carlo and finite-volume sections)] The central claims rest on numerical observation of propagation of chaos and convergence, yet the manuscript supplies no consistency analysis, truncation-error bounds, or convergence studies (grid refinement for the finite-volume scheme or particle-number scaling for Monte Carlo). Without these, it is impossible to separate genuine mean-field behavior from scheme artifacts.
  2. [Mean-field derivation] The mean-field limit is obtained heuristically; the text provides neither a rigorous derivation nor quantitative error estimates between the microscopic system and the PDE. This directly affects the interpretation of the reported convergence as network size grows.
  3. [Bifurcation results] The Hopf bifurcation is asserted on the basis of numerical time series at selected connectivity values. No continuation, eigenvalue tracking, or parameter-sensitivity analysis is described that would confirm the transition is a true bifurcation rather than a numerical transition.
minor comments (1)
  1. Notation for the microscopic and mean-field variables could be unified more clearly across sections to ease comparison between the two descriptions.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the careful reading and constructive comments on our manuscript. We address each major comment below in a point-by-point manner. Where the comments identify gaps in numerical validation, we agree and will revise accordingly; on the heuristic nature of the derivation we maintain the manuscript's stated scope.

read point-by-point responses

  1. Referee: [Numerical experiments (Monte Carlo and finite-volume sections)] The central claims rest on numerical observation of propagation of chaos and convergence, yet the manuscript supplies no consistency analysis, truncation-error bounds, or convergence studies (grid refinement for the finite-volume scheme or particle-number scaling for Monte Carlo). Without these, it is impossible to separate genuine mean-field behavior from scheme artifacts.

    Authors: We acknowledge that the manuscript does not contain formal consistency analysis, truncation-error bounds, or systematic convergence studies such as grid refinement or particle-number scaling. The upwind implicit finite-volume scheme is constructed to preserve positivity and conservation at the discrete level, which provides some control on artifacts, yet this does not replace quantitative convergence checks. We will add grid-refinement studies for the PDE solver and particle-number scaling experiments for the Monte Carlo scheme in the revised manuscript to better separate mean-field behavior from discretization effects. revision: yes

  2. Referee: [Mean-field derivation] The mean-field limit is obtained heuristically; the text provides neither a rigorous derivation nor quantitative error estimates between the microscopic system and the PDE. This directly affects the interpretation of the reported convergence as network size grows.

    Authors: The manuscript states explicitly that the mean-field limit is derived heuristically. A fully rigorous derivation for this nonlinear stochastic spiking model involves substantial technical difficulties (nonlocal interaction, jump processes, and propagation of chaos in two dimensions) that lie outside the present scope. The reported numerical convergence as network size increases is offered as supporting evidence rather than a proof. We will insert an additional remark underscoring the heuristic character and the role of the numerics. revision: no

  3. Referee: [Bifurcation results] The Hopf bifurcation is asserted on the basis of numerical time series at selected connectivity values. No continuation, eigenvalue tracking, or parameter-sensitivity analysis is described that would confirm the transition is a true bifurcation rather than a numerical transition.

    Authors: The claim rests on direct numerical integration showing a qualitative change from steady to oscillatory behavior once connectivity exceeds a threshold. While formal continuation methods or linearization around equilibria are not performed, the time-series data at multiple connectivity values exhibit the expected onset of periodic spiking. We will enlarge the numerical section with a denser sampling of the connectivity parameter and additional diagnostics (e.g., amplitude growth) to strengthen the evidence that the observed transition is not an artifact. revision: partial

Circularity Check

0 steps flagged

No circularity detected in derivation or claims

full rationale

The paper states a heuristic derivation of the mean-field limit, followed by separate Monte Carlo simulation of the microscopic system and finite-volume discretization of the mean-field PDE. Numerical observations of propagation of chaos, convergence as N tends to infinity, and Hopf bifurcation are presented as simulation outcomes, without any reduction of a claimed prediction to a fitted parameter, self-definition, or self-citation chain. No equations or steps in the provided text exhibit the enumerated circularity patterns; the central claims rest on independent numerical comparison rather than tautological re-labeling of inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Abstract-only review provides no explicit free parameters, axioms, or invented entities; the primary unverified premise is the validity of the heuristic mean-field derivation.

axioms (1)
  • domain assumption The mean-field limit exists and can be heuristically derived from the microscopic stochastic model
    Stated directly in the abstract as the basis for subsequent numerical work.

pith-pipeline@v0.9.0 · 5660 in / 1064 out tokens · 43503 ms · 2026-05-25T16:48:12.787389+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

47 extracted references · 47 canonical work pages · 1 internal anchor

  1. [1]

    Augustin, J

    M. Augustin, J. Ladenbauer, F. Baumann, and K. Obermayer , Low-dimensional spike rate models derived from networks of adaptive integrate-and-fire neurons: Comparison and implementation, PLOS Computational Biology, 13 (2017)

    work page 2017

  2. [2]

    Berger, Über das Elektrenkephalogramm des Menschen, Archiv fur Psychiatrie und Ner- venkrankheiten, 87 (1929), pp

    H. Berger, Über das Elektrenkephalogramm des Menschen, Archiv fur Psychiatrie und Ner- venkrankheiten, 87 (1929), pp. 527–70

    work page 1929

  3. [3]

    Boyer, Analysis of the upwind finite volume method for general initial- and boundary-value transport problems, IMA Journal of Numerical Analysis, 32 (2012), pp

    F. Boyer, Analysis of the upwind finite volume method for general initial- and boundary-value transport problems, IMA Journal of Numerical Analysis, 32 (2012), pp. 1404–1439

    work page 2012

  4. [4]

    Brette and W

    R. Brette and W. Gerstner , Adaptive exponential integrate-and-fire model as an effective description of neuronal activity, Journal of Neurophysiology, 94 (2005), pp. 3637–3642

    work page 2005

  5. [5]

    Brunel , Dynamics of sparsely connected networks of excitatory and inhibitory spiking neurons, Journal of Computational Neuroscience, 8 (2000), pp

    N. Brunel , Dynamics of sparsely connected networks of excitatory and inhibitory spiking neurons, Journal of Computational Neuroscience, 8 (2000), pp. 183–208

    work page 2000

  6. [6]

    Carlen, R

    E. Carlen, R. Chatelin, P. Degond, and B. Wennberg , Kinetic hierarchy and propaga- tion of chaos in biological swarm models, Physica D: Nonlinear Phenomena, 260 (2013), pp. 90–111

    work page 2013

  7. [7]

    J. A. Carrillo, M. d. M. González, M. P. Gualdani, and M. E. Schonbek , Classical solutions for a nonlinear Fokker-Planck equation arising in computational neuroscience, Communications in Partial Differential Equations, 38 (2013), pp. 385–409

    work page 2013

  8. [8]

    Chev allier, M

    J. Chev allier, M. J. Cáceres, M. Doumic, and P. Reynaud-Bouret , Microscopic ap- proach of a time elapsed neural model, Mathematical Models and Methods in Applied Sciences, 25 (2015), pp. 2669–2719

    work page 2015

  9. [9]

    M. H. A. Da vis , Markov models and optimization, Monographs on statistics and applied probability, Chapman & Hall, London ; New York, 1st ed ed., 1993

    work page 1993

  10. [10]

    Delarue, J

    F. Delarue, J. Inglis, S. Rubenthaler, and E. Tanré , Global solvability of a networked integrate-and-fire model of McKean–Vlasov type, The Annals of Applied Probability, 25 (2015), pp. 2096–2133

    work page 2015

  11. [11]

    di Volo, A

    M. di Volo, A. Romagnoni, C. Capone, and A. Destexhe , Biologically Realistic Mean- Field Models of Conductance-Based Networks of Spiking Neurons with Adaptation, Neural Computation, 31 (2019), pp. 653–680

    work page 2019

  12. [12]

    Ditlevsen and P

    S. Ditlevsen and P. Lansky , Firing variability is higher than deduced from the empirical coefficient of variation, Neural Computation, 23 (2011), pp. 1944–1966

    work page 2011

  13. [13]

    Drogoul and R

    A. Drogoul and R. Veltz , Hopf bifurcation in a nonlocal nonlinear transport equation stem- ming from stochastic neural dynamics, Chaos: An Interdisciplinary Journal of Nonlinear Science, 27 (2017), p. 021101

    work page 2017

  14. [14]

    Duarte, G

    A. Duarte, G. Ost, and A. A. Rodríguez , Hydrodynamic Limit for Spatially Structured Interacting Neurons, Journal of Statistical Physics, 161 (2015), pp. 1163–1202

    work page 2015

  15. [15]

    F aragó, Á

    I. F aragó, Á. Ha v asi, and Z. Zlatev , Richardson-extrapolated sequential splitting and its application, Journal of Computational and Applied Mathematics, 226 (2009), pp. 218–227

    work page 2009

  16. [16]

    F ardet, M

    T. F ardet, M. Ballandras, S. Bottani, S. Métens, and P. Monceau , Understanding the generation of network bursts by adaptive oscillatory neurons, Frontiers in Neuroscience, 12 (2018)

    work page 2018

  17. [17]

    Fournier and E

    N. Fournier and E. Löcherbach , On a toy model of interacting neurons, Annales de l’Institut Henri Poincaré, Probabilités et Statistiques, 52 (2016), pp. 1844–1876

    work page 2016

  18. [18]

    Fournier, E

    N. Fournier, E. Tanré, and R. Veltz , On a toy network of neurons interacting through their dendrites, arXiv e-prints, (2018)

    work page 2018

  19. [19]

    Gerstner, W

    W. Gerstner, W. Kistler, R. Naud, and L. Paninski , Neuronal Dynamics: From Single Neurons to Networks and Models of Cognition, Cambridge University Press, 2014

    work page 2014

  20. [20]

    Gerstner, W

    W. Gerstner, W. M. Kistler, R. Naud, and L. Paninski ,Neuronal dynamics: from single neurons to networks and models of cognition, Cambridge University Press, Cambridge, United Kingdom, 2014

    work page 2014

  21. [21]

    Glowinski, S

    R. Glowinski, S. J. Osher, and W. Yin (eds.) , Splitting Methods in Communication, Imaging, Science, and Engineering, Springer International Publishing, 2016

    work page 2016

  22. [22]

    A. L. Hodgkin and A. F. Huxley , A quantitative description of membrane current and its application to conduction and excitation in nerve, The Journal of Physiology, 117 (1952), pp. 500–544

    work page 1952

  23. [23]

    Inglis and D

    J. Inglis and D. Talay , Mean-field limit of a stochastic particle system smoothly interacting through threshold hitting-times and applications to neural networks with dendritic compo- nent, SIAM Journal on Mathematical Analysis, 47 (2015)

    work page 2015

  24. [24]

    E. M. Izhikevich , Which model to use for cortical spiking neurons?, IEEE Transactions on Neural Networks, 15 (2004), pp. 1063–1070

    work page 2004

  25. [25]

    M. Kac , Foundations of kinetic theory, in Proceedings of the Third Berkeley Symposium 18 on Mathematical Statistics and Probability, Volume 3: Contributions to Astronomy and Physics, Berkeley, Calif., 1956, University of California Press, pp. 171–197

    work page 1956

  26. [26]

    M. d. Kamps, M. Lepperød, and Y. M. Lai , Computational geometry for modeling neural populations: From visualization to simulation, PLOS Computational Biology, 15 (2019), p. e1006729

    work page 2019

  27. [27]

    Y. A. Kuznetsov , Elements of Applied Bifurcation Theory, Springer New York, New York, NY, 2004

    work page 2004

  28. [28]

    Lapicque, Recherches quantitatives sur l’excitation électrique des nerfs traitée comme po- larisation, J

    L. Lapicque, Recherches quantitatives sur l’excitation électrique des nerfs traitée comme po- larisation, J. Physiol. Pathol. Gen., 9 (1907), pp. 620–635

    work page 1907

  29. [29]

    Lemaire, M

    V. Lemaire, M. Thieullen, and N. Thomas , Exact simulation of the jump times of a class of piecewise deterministic markov processes, Journal of Scientific Computing, 75 (2018), pp. 1776–1807

    work page 2018

  30. [30]

    Marpeau, A

    F. Marpeau, A. Barua, and K. Josić , A finite volume method for stochastic integrate-and- fire models, Journal of Computational Neuroscience, 26 (2009), pp. 445–457

    work page 2009

  31. [31]

    A. D. Masi, A. Gal ves, E. Löcherbach, and E. Presutti , Hydrodynamic limit for inter- acting neurons, Journal of Statistical Physics, 158 (2015), pp. 866–902

    work page 2015

  32. [32]

    Nicola and S

    W. Nicola and S. A. Campbell , Bifurcations of large networks of two-dimensional integrate and fire neurons, Journal of Computational Neuroscience, 35 (2013), pp. 87–108

    work page 2013

  33. [33]

    Nicola and S

    W. Nicola and S. A. Campbell , Nonsmooth bifurcations of mean field systems of two- dimensional integrate and fire neurons, SIAM Journal on Applied Dynamical Systems, 15 (2016), pp. 391–439

    work page 2016

  34. [34]

    Oga w a, T.-S

    S. Oga w a, T.-S. Lee, A. Nayak, and P. Glynn, Oxygenation-sensitive contrast in magnetic resonance image of rodent brain at high magnetic fields, Magnetic Resonance in Medicine, 14 (1990), pp. 68–78

    work page 1990

  35. [35]

    Ostojic, N

    S. Ostojic, N. Brunel, and V. Hakim , Synchronization properties of networks of electrically coupled neurons in the presence of noise and heterogeneities, Journal of Computational Neuroscience, 26 (2009), pp. 369–392

    work page 2009

  36. [36]

    Ott and T

    E. Ott and T. M. Antonsen , Low dimensional behavior of large systems of globally coupled oscillators, Chaos: An Interdisciplinary Journal of Nonlinear Science, 18 (2008), p. 037113

    work page 2008

  37. [37]

    Pazó and E

    D. Pazó and E. Montbrió , Low-Dimensional Dynamics of Populations of Pulse-Coupled Oscillators, Physical Review X, 4 (2014)

    work page 2014

  38. [38]

    Pazó and E

    D. Pazó and E. Montbrió , From Quasiperiodic Partial Synchronization to Collective Chaos in Populations of Inhibitory Neurons with Delay, Physical Review Letters, 116 (2016)

    work page 2016

  39. [39]

    Renart, N

    A. Renart, N. Brunel, and X.-J. W ang , Mean-field theory of irregularly spiking neuronal populations and working memory in recurrent cortical networks, Boca Raton, CRC Press, 2004

    work page 2004

  40. [40]

    Schw alger, M

    T. Schw alger, M. Deger, and W. Gerstner , Towards a theory of cortical columns: From spiking neurons to interacting neural populations of finite size, PLOS Computational Bi- ology, 13 (2017)

    work page 2017

  41. [41]

    Strang , On the construction and comparison of difference schemes, SIAM Journal on Numerical Analysis, 5 (1968), pp

    G. Strang , On the construction and comparison of difference schemes, SIAM Journal on Numerical Analysis, 5 (1968), pp. 506–517

    work page 1968

  42. [42]

    Sznitman, Topics in propagation of chaos, in École d’Été de Probabilités de Saint-Flour, 1989, Springer, 1991, pp

    A.-S. Sznitman, Topics in propagation of chaos, in École d’Été de Probabilités de Saint-Flour, 1989, Springer, 1991, pp. 164–251

    work page 1989

  43. [43]

    Touboul , Bifurcation analysis of a general class of nonlinear integrate-and-fire neurons, SIAM Journal on Applied Mathematics, 68 (2008), pp

    J. Touboul , Bifurcation analysis of a general class of nonlinear integrate-and-fire neurons, SIAM Journal on Applied Mathematics, 68 (2008), pp. 1045–1079

    work page 2008

  44. [44]

    Touboul and R

    J. Touboul and R. Brette , Dynamics and bifurcations of the adaptive exponential integrate- and-fire model, Biological Cybernetics, 99 (2008), pp. 319–334

    work page 2008

  45. [46]

    A new twist for the simulation of hybrid systems using the true jump method

    R. Veltz , A new twist for the simulation of hybrid systems using the true jump method, arXiv:1504.06873 [math], (2015). arXiv: 1504.06873

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  46. [47]

    Vinokur, An analysis of finite-difference and finite-volume formulations of conservation laws, Journal of Computational Physics, 81 (1989), pp

    M. Vinokur, An analysis of finite-difference and finite-volume formulations of conservation laws, Journal of Computational Physics, 81 (1989), pp. 1–52

    work page 1989

  47. [48]

    Zlatev, I

    Z. Zlatev, I. Dimov, I. F aragó, and Á. Ha v asi , Richardson Extrapolation: Practical As- pects and Applications, De Gruyter, 2017. 19

    work page 2017