Challenges of data analysis in a multiscale context: causality inference and unresolved scales
Staff - Faculty of Informatics
Start date: 18 February 2015
End date: 19 February 2015
Seminar given by Prof. Illia Horenko
DATE: Wednesday, February 18th, 2015
PLACE: USI Lugano campus, room 402, Main building (Via G. Buffi 13)
TIME: 14:00
ABSTRACT:
Data analysis for multiscale applications is an important issue in context of such application areas as biophysics, geophysics and neuroscience.
One of the challenges is to learn about the causality relations in the considered systems, based on the available experimental and/or simulation data. Proper inference of such causality relations, giving an additional insight into such processes, can allow improving the respective mathematical and computational models as well as enable comparing the data sets coming from measurements and simulations. Implications of missing/unresolved scales for this problem will be discussed and an overview of methods for data-driven causality inference will be given. Presented mathematical and statistical concepts will be illustrated on two applications: (i) analysis of torsion angles time series obtained in the molecular dynamics simulation; (ii) analysis of historical climate teleconnection series and inference of their mutual influences.
The talk will be based on the recently published paper
S. Gerber and I. Horenko
"On inference of causality for discrete state models in a multiscale context"
Proceedings of the National Academy of Sciences of USA (PNAS), 111 (41), 14651-14656, 2014
HOST: Prof. Michele Parrinello
