Dr Sebastiano Stramaglia
University of Bari, Italy
Causal approaches to the inference of dynamical networks
Thu Sep 30, 2010. 11.00am.
Granger causality is a major approach to measure the information flow between variables. In its original formulation, Granger causality is based on linear modelling of the data. In the last years, the problem of causality attracted interest also in the machine learning community, and a nonlinear version of Granger causality has been introduced in the frame of kernel methods. I will discuss the application of nonlinear causality analysis to the inference of couplings in dynamical networks, an issue with relevant applications in neuroscience, genomics and many other fields.
