The Hawkes Edge Partition Model for Continuous-time Event-based Temporal Networks

Sikun Yang, Heinz Koeppl

Abstract: We propose a novel probabilistic framework to model continuous-time interaction events data. Our goal is to infertheimplicitcommunity structure underlying the temporal interactions among entities, and also to exploit how thecommunity structure influences the interaction dynamics among these nodes. To this end, we model the reciprocatinginteractions between individuals using mutually-exciting Hawkes processes. The base rate of the Hawkes processfor each pair of individuals is built upon the latent representations inferred using the hierarchical gamma processedge partition model (HGaP-EPM). In particular, our model allows the interaction dynamics between each pair ofindividuals to be modulated by their respective affiliated communities. Moreover, our model can flexibly incorporatethe auxiliary individuals’ attributes, or covariates associated with interaction events. Efficient Gibbs sampling andExpectation-Maximization algorithms are developed to perform inference via Pólya-Gamma data augmentation strategy.Experimental results on real-world datasets demonstrate that our model not only achieves competitive performancefor temporal link prediction compared with state-of-the-art methods, but also discovers interpretable latent structurebehind the observed temporal interactions.