Reconstruction of a scalar voltage-based neural field network from observed time series

Institute for Physics and Astronomy, University of Potsdam - Karl-Liebknecht-Str. 24/25, 14476 Potsdam-Golm, Germany and Research Institute for Supercomputing, Nizhni Novgorod State University - Gagarin Av. 23, 606950 Nizhni Novgorod, Russia

Received: 22 August 2017
Accepted: 24 September 2017

Abstract

We present a general method for the reconstruction of a network of nonlinearly coupled neural fields from observations. A prominent example of such a system is a dynamical random neural network model studied by Sompolinsky et al. (Phys. Rev. Lett., 61 (1988) 259). We develop a technique for inferring the properties of the system from the observations of the chaotic voltages. Only the structure of the model is assumed to be known, while the nonlinear gain functions of the interactions, the matrix of the coupling constants, and the time constants of the local dynamics are reconstructed from the time series.