讲座题目👩🏼⚖️：State space based nonlinear time series analysis using complex network methods 主讲人：Reik Donner 开始时间🤸🏽♀️：2019-09-06 08:30:00 结束时间👨🏻🦲：2019-09-06 09:30:00 讲座地址：闵行校区物理楼226报告厅 主办单位：物理与电子科学学院
报告人简介： Since 2018 - Professor of Mathematics (Data Science and Stochastic Modeling), Magdeburg-Stendal University of Applied Sciences, Magdeburg, Germany 2014–2019 Research Group Leader, Potsdam Institute for Climate Impact Research 2011 EGU Division Outstanding Young Scientist Award for Nonlinear Processes 2007 JSPS Postdoctoral Fellowship, 2009 Guest Professorship, Osaka Prefecture University, Sakai, Japan 2007-2014 Postdoctoral positions at Dresden University of Technology, MPI for Physics of Complex Systems (Dresden), Potsdam Institute for Climate Impact Research, MPI for Biogeochemistry (Jena) 1997 – 2007 Study of Physics and Mathematics at Potsdam University, Germany
Division Science Officer for Time Series Analysis and Big Data of the EGU Division Nonlinear Processes Co-PI of the Belmont Forum/JPI Climate project GOTHAM, the German-Brazilian Research Training Group “Complex Processes in Networks” and the EU Research Training Group CAFÉ Editorial Board Member in currently four international journals 报告内容🩻： Over the last about four decades, state space based methods have gained considerable importance in the field of nonlinear time series analysis. Beyond the notion of fractal dimensions that directly derives from this framework, the concept of recurrences in state space and their quantitative analysis has become a valuable starting point for the detailed characterization of complex systems based on observational time series. In my talk, I will demonstrate how the proximity or similarity of dynamical states on a sample trajectory (defining their recurrence in the state space) of a system under study can be directly translated into a network representation in terms of a random geometric graph, which is referred to as the associated recurrence network. The potentials of this new perspective for complex system characterization will be discussed along with selected recent achievements in the field based on both paradigmatic model systems and real-world observational time series. Specific emphasis will be put on the interpretation of network transitivity as a generalized fractal dimension concept, together with some theoretical and practical challenges arising from it. Finally, I will discuss some recent achievements regarding the associated threshold selection and significance testing problems. |
