On Friday, 04.09.2020., from 16.15-17.15,

Robert Rosenbaum, University of Notre Dame,

will give a talk in Dynamical Systems Seminar under the title

Spatiotemporal dynamics and reliable computations in recurrent spiking neural networks

Abstract: Randomly connected networks of spiking neuron models provide a parsimonious model of neural variability, but are notoriously unreliable for performing computations. We show that this difficulty can be overcome by incorporating the well-documented dependence of synaptic connection probability on the distance between neurons. Using a Fokker-Planck formalism, we show that spatially extended spiking networks exhibit symmetry-breaking, Turing-Hopf bifurcations to generate intricate, high-dimensional spatiotemporal patterns. The resulting dynamics can be trained to perform dynamical computations using a reservoir computing approach.

The talk will be held via Zoom platform,
https://us02web.zoom.us/j/89011911019

Meeting ID: 890 1191 1019

Everybody is invited.