Max Planck Institute for Dynamics and Self-Organization -- Department for Nonlinear Dynamics and Network Dynamics Group
Personal tools
Log in

MPI Advances - Quantifying causal interactions from time series of complex systems, Jakob Runge (PIK Potsdam)

Date:   Monday, 23.09.2013 14:15
Contact:  Marc Timme

Description

Quantifying causal interactions from time series of complex systems

Jakob Runge (Potsdam Institute for Climate Impact Research (PIK), Potsdam
Research Domain IV: Transdisciplinary Concepts and Methods)

Sept. 23, 14 ct, MPIDS - Bunsenstraße 10, House 8, Lecture hall

This talk deals with the problem of detecting the existence and quantifying the strength of (Granger-) causal interactions from time series of complex systems. I address this topic from an information-theoretic perspective with as few assumptions as possible about the underlying process. In the first part of the talk, the primary goal is to infer and distinguish the existence of direct from indirect interactions and the second part focuses on the problem of defining a meaningful measure of a direct interaction strength. To this end, I will analytically and numerically analyze the interpretation of the well-known measures mutual information and transfer entropy and argue why these measures have some short-comings in quantifying a direct interaction strength. To overcome these problems, I will discuss several alternative measures based on the idea of momentary information transfer and show that these measures can be used to quantify very specific aspects of an interaction. The usefulness and potential of the novel approach will be illustrated by studying the mechanism of the Walker circulation in the tropical Pacific.                                        Authors: Jakob Runge, J. Heitzig, V. Petoukhov, N. Marwan, and J. Kurths

Remarks

MPIDS - Bunsenstraße 10, House 8, Lecture hall

back to overview