Berlin 2015 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
BP: Fachverband Biologische Physik
BP 8: Neurophysics II
BP 8.8: Vortrag
Montag, 16. März 2015, 16:30–16:45, H 1058
Estimating branching parameters from subsampled systems — •Jens Wilting1, Theo Geisel1,2, and Viola Priesemann1,2 — 1Max-Planck-Institute for Dynamics and Self-Organization, Göttingen, Germany — 2BCCN, Göttingen, Germany
Branching processes are frequently used to model time-varying data in economics, finance, epidemiology, population dynamics, physics, and neuroscience. Depending on the expected number of offspring є , the process shows either stationary fluctuations (sub-critical) or transient growth (super-critical), and has a deep connections to self-organized criticality. Methods to infer є from fully sampled systems are well established. In real-world systems, however, often only a subset of all agents can be sampled. Since under subsampling classical approaches to infer є fail, we have developed a novel approach based on the auto-correlation function that for the first time allows to robustly infer є even under subsampling. Importantly, our method generalizes to auto-regressive processes with both additive and multiplicative noise, making it widely applicable. We demonstrate that our method correctly estimates є under sub-sampling in simulated branching processes, and also in the purely deterministic Bak-Tang-Wiesenfeld model. Moreover, applying our method to necessarily subsampled neuronal population dynamics from macaque monkeys, we show that spiking dynamics reflects a sub-critical regime (є = 0.95)