Title: Labelled dynamic Bayesian network: Framework and structure learning for application to ecological network
Authors: Etienne Auclair - UTeam (France) [presenting]
Abstract: A ``labelled dynamic Bayesian network'' is a particular case of dynamic Bayesian network where the probabilities of the random variables are not described by conditional probabilities tables, but by a small set of parameters shared by every variable of the network. The interest of this model is the fact that the number of parameters of the model is fixed, and does not depend on the size and the complexity of the network. This model will be described with an algorithm developed to learn the structure of a labelled dynamic Bayesian network, illustrated on an ecological problem: learning the structure of an ecological network using dynamic data of presence/absence of the species.