Note: This content is accessible to all versions of every browser. However, this browser does not seem to support current Web standards, preventing the display of our site's design details.


Learning the Structure of Continuous Time Bayesian Networks


L. Studer

Master Thesis, HS14 (10359)

Continuous Time Bayesian Networks (CTBNs) model Markov processes of countable states with continuous time dynamics. Their strength lies in decoupling the multivariate dependence structures from the dynamics by introducing a dependence graph and a set of local rate parameters. In this thesis a novel approach to learning a CTBN's dependence graph structure from ensembles of time series data is proposed. It is shown that the graph of a CTBN can be learned eciently by marginalizing out its rate parameters, and hence by-passing explicit estimation of those parameters. To this end the theory and design of a scalable CTBN graph learning algorithm is discussed. In order to cope with the computational complexity arising at relevant problem dimensions, a massively parallelized implementation for use on e.g. the IBM Blue Gene R /Q system is conceived.


Type of Publication:

(12)Diploma/Master Thesis

C. Zechner

File Download:

Request a copy of this publication.
(Uses JavaScript)
% Autogenerated BibTeX entry
@PhdThesis { Xxx:2014:IFA_5030
Permanent link