Die DPG-Frühjahrstagung in Dresden musste abgesagt werden! Lesen Sie mehr ...
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
CPP: Fachverband Chemische Physik und Polymerphysik
CPP 42: Data analytics for dynamical systems I (joint session SOE/BP/CPP/DY)
CPP 42.9: Vortrag
Dienstag, 17. März 2020, 12:30–12:45, GÖR 226
Network inference from event sequences: Disentangling synchrony from serial dependency — •Reik Donner1,2, Forough Hassanibesheli2,3, Frederik Wolf2,3, and Adrian Odenweller4,5 — 1Magdeburg-Stendal University of Applied Sciences, Magdeburg — 2Potsdam Institute for Climate Impact Research — 3Department of Physics, Humboldt University, Berlin — 4Center for Earth System Research and Sustainability, University of Hamburg — 5Max Planck Institute for Meteorology, Hamburg
Inferring coupling among interacting units or quantifying their synchronization based on the timing of discrete events has vast applications in neuroscience, climate, or economics. Here, we focus on two prominent concepts that have been widely used in the past - event synchronization (ES) and event coincidence analysis (ECA). Numerical performance studies for two different types of spreading processes on paradigmatic network architectures reveal that both methods are generally suitable for correctly identifying the unknown links. By further applying both concepts to spatiotemporal climate datasets, we demonstrate that unlike ECA, ES systematically underestimates linkages in the presence of temporal event clustering, which needs to be accounted for in network reconstruction from data. In turn, for spike train data from multi-channel EEG recordings (with relatively narrow inter-event time distributions), the obtained results are practically indistinguishable. Our findings allow deriving practical recommendations for suitable data preprocessing in the context of network inference and synchronization assessment from event data.