B1741
Title: Bayesian structure learning in high-dimensional graphical models
Authors: Reza Mohammadi - University of Amsterdam (Netherlands) [presenting]
Abstract: Graphical models have been used in many application areas for learning conditional independence structure among a (possibly large) collection of variables. For these models, Bayesian structure learning, while providing a natural and principled way for model uncertainty, often lag behind frequentist approaches in terms of computational efficiency and scalability. Hence, scalable approaches with theoretical and computational safeguards are critical to leveraging the advantages of posterior inference. We discuss the computational problems related to Bayesian structure learning, and we offer several solutions to cope with the computational issues. To show its empirical usefulness, we present the application of our approach to high-dimensional fMRI data for brain connectivity studies. In addition, we have implemented our method in the R packages BDgraph and ssgraph, which are available at CRAN.
