Causal Discovery from Time Series of Complex Systems: Detection, Quantification, Prediction

Abstract.  Causal discovery methods for time series datasets aim at detecting potentially causal statistical associations that cannot be
explained by other variables in the dataset. The knowledge of potentially causal drivers can then help to quantify causal effects,
improve time series prediction schemes, test interaction hypotheses, and also to assess global features of large-scale systems via causal complex network measures. In this seminar I will give an overview over these approaches, discuss challenges and limitations, and present some applications to the Earth's climate system.

Short bio: Jakob Runge received his Diploma (2010) and Ph.D. (2014) both in physics from Humboldt University Berlin and the Potsdam
Institute for Climate Impact Research. Since 2016 he is a Research Associate at Imperial College London funded by a JSMF Postdoctoral
Fellowship Award in Studying Complex Systems.