Regensburg 2022 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
BP: Fachverband Biologische Physik
BP 2: Computational Biophysics and Neuroscience
BP 2.7: Talk
Monday, September 5, 2022, 11:30–11:45, H13
Mesoscopic description of metastability and hippocampal replay in neural networks with short-term plasticity — Bastian Pietras1, Valentin Schmutz2, and •Tilo Schwalger3 — 1Universitat Pompeu Fabra, Barcelona, Spain — 2Brain Mind Institute, School of CÉcole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland — 3Technische Universität Berlin
Sequences of metastable states in neuronal population activities have been linked to various sensory and cognitive functions. Two prominent mechanisms of metastable dynamics are noise-induced transitions among attractors and deterministic transitions induced by slow fatigue processes. The dependence of these mechansisms on neural circuit parameters at the microscopic scale are largely unclear. Starting with a network of linear-nonlinear Poisson spiking neurons with synaptic short-term plasticity, we use a bottom-up approach and derive a stochastic neural-mass model at the mesocopic scale that links to the microscopic circuit parameters. We apply the mesoscopic model to investigate hippocampal "replay" events, i.e. spontaneous sequences of metastable activations of place cells. We study a spiking-neural-network for depression-induced metastability of place-cell activity. The corresponding mesoscopic model precisely reproduces the statistics of metastastable events in the microscopic network model. This enables us to efficiently explore the full range of neuron numbers including the thermodynamic limit. We find a novel dynamical regime in finite-size networks where metastable replay events are fluctuation-driven and exhibit biologically plausible irregularity.