Parts | Days | Selection | Search | Downloads | Help
DY: Fachverband Dynamik und Statistische Physik
DY 9: Nonlinear dynamics, synchronization and chaos II
DY 9.5: Talk
Tuesday, March 24, 2009, 11:15–11:30, ZEU 255
Computing by Switching? Heteroclinic Bifurcation in Spiking Neural Networks — •Fabio Schittler-Neves, Christoph Kirst, Andreas Sorge, and Marc Timme — Network Dynamics Group, MPI for Dynamics and Self-Organization, Göttingen, Germany
Networks of spiking neurons may often exhibit saddle periodic orbits that are heteroclinically connected among each other. For fast interaction responses, these systems may show heteroclinically connected unstable attractors [1-3]. Here we show that pulse-coupled (spiking) systems exhibit an analog of heteroclinic bifurcation that has both discrete and continuous parts [4]. On this basis, we demonstrate how input signals may be encoded by spiking neural networks in a novel way.
M. Timme et al., Phys. Rev. Lett. 89:154105 (2002)
[2] P. Ashwin and M. Timme, Nonlinearity 18:2035-2060 (2005)
[3] P. Ashwin and M. Timme, Nature 436:36-37 (2005)
[4] C. Kirst and M. Timme, http://arxiv.org/abs/0709.3432v1,
Phys. Rev. E, Rapid Communications (accepted)