Berlin 2015 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

BP: Fachverband Biologische Physik

BP 2: Neurophysics I

BP 2.13: Vortrag

Montag, 16. März 2015, 12:45–13:00, H 1058

Nonlinear population dynamics for finite-size spiking neural networks with adaptation -- Non-Gaussian fluctuations and information filtering — •Tilo Schwalger, Moritz Deger, and Wulfram Gerstner — Brain Mind Institute, École polytechnique fédérale de Lausanne, CH-1015 Lausanne

Bridging the scale from the microscopic dynamics of single neurons to the global population activities of pulse-coupled neurons is crucial for multi-scale modeling of the nervous system. Current theories mostly consider the limit of large networks. However, this approach is limited to the mean activity and neglects fluctuation effects. In realistic neural circuits, the number of neurons of a given type can be rather small (N=50--1000), which requires a theory for the fluctuating population dynamics. Existing finite-size theories are either based on rather simplified neuron models or rely on heuristic assumptions. Using mean field theory and a quasi-renewal approximation [1,2], we present stochastic population equations for the large class of generalized integrate-and-fire neurons with spike-frequency adaptation. Our theory goes beyond the Gaussian approximation and thus applies to rather small populations. We study spontaneous transitions between up and down states in a bistable network induced by finite-size noise. Furthermore, for the asynchronous state, we analytically calculate the power spectrum of the fluctuations. This allows us to investigate information filtering by coupled populations of adapting neurons [2].

References: [1] R. Naud, W. Gerstner, PLoS Comp. Biol. (2012); [2] M. Deger, T. Schwalger, R. Naud, W. Gerstner, Phys. Rev. E (2014)