Past Events

AbstractAs networks proliferate in diverse fields of modern endeavor, so grows the need to control them on a large scale (e.g. in wireless or sensor networks), and interpret the high-dimensional data they generate (e.g. in social or biological networks). In this talk I present a broad-based approach to both challenges, based on a popular, but currently unrelated, formalism for multivariate statistical inference: Markov random fields (MRFs).The popularity of MRFs derives from the empirical success of graph-based heuristics, like Belief Propagation (BP), in performing basic statistical tasks.