Propagation of chaos in neural fields

DOI10.1214/13-AAP950MaRDI QIDQ2454411zbMATH OpenWikidataFDO

Publication date 13 June 2014

Published in The Annals of Applied Probability (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1108.2414, https://projecteuclid.org/euclid.aoap/1398258102

delayed stochastic differential equations infinite-dimensional stochastic processes mean-field limits neural fields propagation of chaos

Mathematics Subject Classification ID

Convergence of probability measures (60B10) Limit theorems in probability theory (60F99) Stochastic ordinary differential equations (aspects of stochastic analysis) (60H10) Interacting random processes; statistical mechanics type models; percolation theory (60K35) Neural biology (92C20)

Abstract: We consider the problem of the limit of bio-inspired spatially extended neuronal networks including an infinite number of neuronal types (space locations), with space-dependent propagation delays modeling neural fields. The propagation of chaos property is proved in this setting under mild assumptions on the neuronal dynamics, valid for most models used in neuroscience, in a mesoscopic limit, the neural-field limit, in which we can resolve the quite fine structure of the neuron's activity in space and where averaging effects occur. The mean-field equations obtained are of a new type: they take the form of well-posed infinite-dimensional delayed integro-differential equations with a nonlocal mean-field term and a singular spatio-temporal Brownian motion. We also show how these intricate equations can be used in practice to uncover mathematically the precise mesoscopic dynamics of the neural field in a particular model where the mean-field equations exactly reduce to deterministic nonlinear delayed integro-differential equations. These results have several theoretical implications in neuroscience we review in the discussion.

Review text

The author considers a so-called neural field, made of \(N\) neurons grouped into \(P(N)\) populations having location \(r_1,\dots, r_{P(N)}\) and sizes \(N_1(N),\dots, N_{P(N)}(N)\), respectively. The membrane potential \(X^{i,N}_t\) of the neuron \(i\) belonging to the population \(p(i)\equiv\alpha\), at time \(t\), has its evolution modellized by a stochastic differential equation \[ dX^{i,N}_t= \sigma(r_\alpha) dW^i_t+ f(r_\alpha, t, X^{i,N}_t)+ {1\over P(N)} \sum^{P(N)}_{\gamma=1}\, \sum_{j\in p^{-1}(\gamma)} {dt\over N_\gamma} b(r_\alpha, r_\gamma, X^{i,N}_t, X^{j,N}_{t-\tau(r_\alpha, r_\gamma)}), \] where \((W^i_t)\) is an \(N\)-dimensional Brownian motion, and the smooth function \(Z(r_\alpha,r_\gamma)\) represents a time delay into the interaction function \(b\). This article addresses the limiting behaviour, as \(N\to\infty\), of such neural fields, under mild regularity assumptions and the regime condition \[ e(N):= {1\over P(N)} \sum^{P(N)}_{\gamma=1} {1\over N_\gamma(N)}\to 0. \] The main result is the convergence in law of \((X^{i,N}_t)\), to a mean field process \((\overline X_t(r_\alpha))\) solving the following McKean-Vlasov equation with delay: \[ d\overline X_t= \sigma\,dW_t+ f(r,t,\overline X_t)\,dt+ \int\mathbb{E}_{\overline Z}[(b(r,r',\overline X_t, \overline Z_{t-\tau(r,r')}(r')]\,\lambda(dr')\, dt, \] where \((W_t)\) is a specially chaotic Brownian motion (\(W_t\) and \(W_t(r')\) are independent for any \(r\neq r'\)), \((\overline Z)\) is an independent copy of \((\overline X)\), and \(\lambda\) is a special law, under which the different locations \(r'\) are i.i.d. The author furthermore specifies the speed of this convergence: the expected value of the distance \(\sup_{-\tau\leq s\leq t} |X^{i,N}_s-\overline X^i_s(r_\alpha)|^2\) is shown to be \(O(e(N)+ 1/P(N))\). Moreover, the above systems of evolution equations and of mean field equations are both shown to admit a unique, square integrable, string solution, with infinite lifetime. Finally, theoretical implications to neuroscience are discussed.

Recommendations

Cites work

Cited in

(56)

This page was built for publication: Propagation of chaos in neural fields

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2454411)