modelling reciprocating relationships with hawkes strongprocessesstrong.pdfVIP

Modelling Reciprocating Relationships with Hawkes Processes Charles Blundell Katherine A. Heller Gatsby Computational Neuroscience Unit Duke University University College London Durham, NC, USA London, United Kingdom kheller@ c.blundell@gatsby.ucl.ac.uk Jeffrey M. Beck University of Rochester Rochester, NY, USA jbeck@ Abstract We present a Bayesian nonparametric model that discovers implicit social struc- ture from interaction time-series data. Social groups are often formed implicitly, through actions among members of groups. Yet many models of social networks use explicitly declared relationships to infer social structure. We consider a par- ticular class of Hawkes processes, a doubly stochastic point process, that is able to model reciprocity between groups of individuals. We then extend the Infinite Relational Model by using these reciprocating Hawkes processes to parameterise its edges, making events associated with edges co-dependent through time. Our model outperforms general, unstructured Hawkes processes as well as structured Poisson process-based models at predicting verbal and email turn-taking, and mil- itary conflicts among nations. 1 Introduction As social animals, people constantly organise themselves into social groups. These social groups can revolve around particular activities, such as sports teams, particular roles, such as store managers, or general social alliances,