CMStatistics 2020: Start Registration
View Submission - CMStatistics
Title: Bayesian nonparametrics for sparse dynamic networks Authors:  Cian Naik - University of Oxford (United Kingdom) [presenting]
Francois Caron - University of Oxford (United Kingdom)
Judith Rousseau - University of Oxford (United Kingdom)
Yee Whye Teh - Oxford University (United Kingdom)
Konstantina Palla - University of Cambridge (United Kingdom)
Abstract: A Bayesian nonparametric approach for sparse time-varying networks is proposed. A positive parameter is associated with each node of a network, which models the sociability of that node. Sociabilities are assumed to evolve over time and are modelled via a dynamic point process model. The model is able to (a) capture long term evolution of the sociabilities, and (b) yields sparse graphs, where the number of edges grows subquadratically with the number of nodes. The evolution of the sociabilities is described by a tractable time-varying generalised gamma process. We provide some theoretical insights into the model and apply it to three real-world datasets: a network of hyperlinks between communities on Reddit, email exchanges between members of the Democratic National Congress, and co-occurrences of words in Reuters news articles after the September $11^{th}$ attacks.