Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
SOE: Fachverband Physik sozio-ökonomischer Systeme
SOE 24: Extreme Events (joint session DY / SOE)
SOE 24.6: Vortrag
Donnerstag, 19. März 2015, 16:30–16:45, BH-N 243
Forecasting extreme events in high-dimensional excitable systems — •Stephan Bialonski1, Gerrit Ansmann2,3,4, and Holger Kantz1 — 1Max-Planck-Institute for the Physics of Complex Systems, Dresden, Germany — 2Department of Epileptology, University of Bonn, Bonn, Germany — 3Helmholtz Institute for Radiation and Nuclear Physics, University of Bonn, Bonn, Germany — 4Interdisciplinary Center for Complex Systems, University of Bonn, Bonn, Germany
The dynamics of many high-dimensional systems, ranging from nature to technology and society, can exhibit extreme events, i.e. large deviations from the average behaviour. Since extreme events can pose severe threats and can have implications for economy, politics, or health, a successful and reliable prediction of such events is highly desirable. We investigate extreme events in a high-dimensional deterministic system: a network of FitzHugh-Nagumo units. Mimicking field studies, we assume that the temporal evolution of only some degrees of freedom of the system is observed and that the exact equations of motion are unknown. Addressing these challenges, we present a data-driven approach to predict extreme events which is only based on the time series of some observables and on the coupling topology of the network. By iterative predictions, we are able to forecast the onset of an extreme event as well as the propagation and extinction of excitation, i.e. the full life-cycle of an extreme event.