Name:
Pekka Marttinen

Affiliation:
University of Helsinki, visiting PhD. student at KERMIT, Faculty of 
Bioscience Engineering, Ghent University

Presentation preference: poster

Title:
Bayesian learning of causal graphs for multivariate time-series

Abstract:
Graphical modelling strategies have been recently discovered as a versatile 
tool for analyzing multivariate stochastic processes. However, majority of 
the existing statistical methods for this purpose are restricted to 
undirected graphs, which leave the most intriguing questions about 
causality open and may also yield spurious associations between variables 
in the dynamic context. We introduce a Bayesian method for unsupervised 
learning of causal graphical models for multivariate stochastic processes, 
which allows for non-decomposable graphs and structural breaks in the 
processes. Contrary to static graphical models, the number of possible 
dynamic causal graphical models is extremely large even for small systems, 
and consequently, standard Bayesian computation based on Markov chain Monte 
Carlo is not in practice a feasible alternative for model learning. To 
obtain a numerically efficient approach we utilize a recent Bayesian 
information theoretic criterion for model learning, which has attractive 
properties when the potential model complexity is large relative to the 
size of the observed data set. The performance of our method is illustrated 
by analyzing both simulated and real data sets.