B0412
Title: Learning structures of Bayesian networks with cyclic structures
Authors: Witold Wiecek - Certara UK Ltd (United Kingdom) [presenting]
Frederic Bois - Certara UK Ltd (United Kingdom)
Ghislaine Gayraud - University of Technology of Compiegne (France)
Abstract: Bayesian networks are a popular approach to modelling networks. Networks in BNs must be acyclic while in many applications they include cycles. Dynamic BNs can be used but they require time series data. We present an alternative model that embeds cyclic structures within acyclic BNs, allowing us to still use the factorization property of BNs and informative priors on network structure. We present an implementation in the linear Gaussian case, where cyclic structures are treated as multivariate nodes. We use a Markov Chain Monte Carlo algorithm for inference, allowing us to work with the whole posterior distribution on the space of graphs. The algorithm implemented as a part of graph$\_$sampler, open-source software for modelling networks.
