报告人：Alexander Fuchs-Kreiss （University of Hildesheim）

报告时间：2026年4月28日（周二）下午14:00 - 15:00

报告地点：腾讯会议388-984-701

报告摘要：We study actors in a network who can cast instantaneousevents, e.g., posts on a social media platform. These events will bemodelled as multivariate Hawkes process allowing for mutual excitation,that is, events cast by one actor are allowed to increase the activityof other actors. Such influences can, in turn, be interpreted as edgesof a network. Our primary target is estimation of this influencenetwork. However, we believe that a second relevant driver of suchevents are globally available covariates, e.g., time of the day, orinformation about the actors. We therefore study a model that alsoincludes such common drivers.We derive rates of convergence for all of the model parameters when boththe number of actors and the time horizon tends to infinity. To preventan exploding network, sparseness is assumed, and enforced by aLASSO-type estimator. In order to allow for inference, we study ade-biasing procedure.We will also discuss numerical aspects, and discuss in this context anextension to multiple Hawkes processes, that is, if several classes ofinteractions are observed that follow the same dynamics. To conductinference on the network, we use a permutation type test.This is joint work with Enno Mammen, Eckehard Olbrich, and WolfgangPolonik.