Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
DY: Fachverband Dynamik und Statistische Physik
DY 26: Data Analytics for Complex Dynamical Systems (joint SOE/DY Focus Session) (joint session SOE/DY)
DY 26.4: Vortrag
Dienstag, 23. März 2021, 12:00–12:20, SOEa
Tipping and transition paths in high-dimensional agent-based models — •Luzie Helfmann1,2,3, Peter Koltai1, Jobst Heitzig3, and Christof Schütte2,1 — 1Freie Universität Berlin — 2Zuse Institute Berlin — 3Potsdam Institute for Climate Impact Research
Agent-based models are a popular choice for modeling complex social systems. Here, we are concerned with studying noise-induced tipping between relevant subsets of the agent state space, e.g., in order to understand drastic opinion changes in a population of agents. Due to the large number of interacting individuals, agent-based models are usually very high-dimensional. We therefore apply Diffusion Maps, a non-linear dimension reduction, to reveal the intrinsic low-dimensional structure of the model dynamics. We will characterize the tipping behavior by means of Transition Path Theory, a theory for gaining statistical understanding of the tipping paths (e.g., their density, flux, rate). We will illustrate our approach on two examples, both exhibiting a multitude of tipping pathways.