Wasserstein Learning of Deep Generative Point Process Models
This addresses the problem of modeling asynchronous sequential data more effectively for applications in real-world phenomena, though it appears incremental as it builds on existing point process frameworks.
The paper tackles the limitations of intensity-based point process models by proposing an intensity-free approach that transforms nuisance processes to a target one, and trains the model using a likelihood-free method leveraging Wasserstein distance. Experiments on synthetic and real-world data show the proposed model outperforms conventional ones.
Point processes are becoming very popular in modeling asynchronous sequential data due to their sound mathematical foundation and strength in modeling a variety of real-world phenomena. Currently, they are often characterized via intensity function which limits model's expressiveness due to unrealistic assumptions on its parametric form used in practice. Furthermore, they are learned via maximum likelihood approach which is prone to failure in multi-modal distributions of sequences. In this paper, we propose an intensity-free approach for point processes modeling that transforms nuisance processes to a target one. Furthermore, we train the model using a likelihood-free leveraging Wasserstein distance between point processes. Experiments on various synthetic and real-world data substantiate the superiority of the proposed point process model over conventional ones.