We here focus on the task of learning Granger causality matrices for multivariate point processes. In order to accomplish this task, our work is the first to explorethe use of Wold processes. By doing so, we are able to develop asymptotically fast MCMC learning algorithms. With Nbeing the total number of events and K the number of processes, our learning algorithm has aO(N( log(N) + log(K)))costper iteration. This is much faster than theO(N3K2)orO(K3) for the state of theart. Our approach, called GRANGER-BUSCA, is validated on nine datasets. This is an advance in relation to most prior efforts which focus mostly on subsets of the Memetracker data. Regarding accuracy, GRANGER-BUSCA is three times more accurate (in Precision@10) than the state of the art for the commonly explored subsets Memetracker. Due to GRANGER-BUSCA’s much lower training complexity,our approach is the only one able to train models for larger, full, sets of data.
Where: NIPS - Thirty-second Conference on Neural Information Processing Systems, December 2019 (Montreal, Brazil)