Neural Spectral Marked Point Processes

Self-and mutually-exciting point processes are popular models in machine learning and statistics for dependent discrete event data. To date, most existing models assume stationary kernels (including the classical Hawkes processes) and simple parametric models. Modern applications with complex event data require more general point process models that can incorporate contextual information of the events, called marks, besides the temporal and location information. Moreover, such applications often require non-stationary models to capture more complex spatio-temporal dependence. To tackle these challenges, a key question is to devise a versatile influence kernel in the point process model. In this paper, we introduce a novel and general neural network-based non-stationary influence kernel with high expressiveness for handling complex discrete events data while providing theoretical performance guarantees. We demonstrate the superior performance of our proposed method compared with the state-of-the-art on synthetic and real data. INTRODUCTION Event sequence data are ubiquitous in our daily life, ranging from traffic incidents, 911 calls, social media posts, earthquake catalog data, and COVID-19 data (see, e.g., Bertozzi et al. (2020) ). Such data consist of a sequence of events indicating when and where each event occurred, with additional descriptive information (called marks) about the event (such as category, volume, or free-text). The distribution of events is of scientific and practical interest, both for prediction purposes and for inferring events' underlying generative mechanism.