Regensburg 2019 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
BP: Fachverband Biologische Physik
BP 15: Focus session: Collective Dynamics in Neural Networks
BP 15.6: Talk
Wednesday, April 3, 2019, 11:15–11:30, H11
Resonant chaos in random networks of adapting neurons — Samuel Muscinelli1, Wulfram Gerstner1, and •Tilo Schwalger2,3 — 1Brain Mind Institute, École polytechnique fédérale de Lausanne (EPFL), 1015 Lausanne, Switzerland — 2Institut für Mathematik, Technische Universität Berlin, 10623 Berlin — 3Bernstein Center for Computational Neuroscience, 10115 Berlin
The dynamical response of cortical neurons to inputs is governed by several history-dependent mechanisms. One prominent example are slow negative feedback mechanisms that lead to spike frequency adaptation -- a widely observed feature of neurons. Despite the importance of adaptation, it is theoretically poorly understood how such neuronal properties shape the collective activity of recurrent networks. Here, we study the dynamics of a random recurrent network of multi-dimensional rate neurons admitting adaptation. Using dynamical mean-field theory and an iteratative map for the self-consistent second-order statistics of neural activity, we show how local adaptation and recurrent feedback from the network give rise to two distinct types of chaotic behavior, resonant and non-resonant chaos. The type of chaos as well as the resonance frequency can be predicted by the single neuron susceptibility. Interestingly, the emerging correlation time of the network activity cannot be increased by slow adaptation. We also find that suppression of chaos is maximized by input frequencies close to the resonant one. More generally, our work sheds light on the complex interplay between local neuron properties and recurrent network connectivity beyond adaptation.