Structured Neural Marked Point Processes for Interpretable Event Interaction Modeling

2026-05-17

AI Summary

Existing neural point process models struggle to explicitly capture the structured dependencies among event types in multivariate event streams. To address this limitation, this work proposes the Structured Neural Marked Point Process (SNMPP), which, for the first time, decouples symbolic event interaction topologies—encoding excitatory, inhibitory, or neutral relationships—from delay-aware monotonic temporal decay patterns, and integrates them via a multiplicative neural influence kernel. By combining a symbolic interaction graph, a delay-aware temporal network, and a hierarchical Monte Carlo estimator, SNMPP enables efficient and interpretable learning. Experiments on both synthetic and real-world datasets demonstrate that SNMPP not only achieves superior event prediction accuracy but also effectively recovers the underlying structured interaction relationships, offering a compelling balance between predictive performance and interpretability.

Abstract

Multi-class event streams arise in numerous real-world applications, where uncovering structured, interpretable inter-event relationships, together with accurate prediction, remains a central challenge. Existing neural point process models are highly expressive but encode event interactions in a black-box manner, preventing explicit discovery of structured dependencies. In this paper, we propose a structured neural marked point process (SNMPP) that achieves high modeling flexibility while enabling explicit event-wise and class-wise relationship discovery from data. Our model constructs a product-form neural influence kernel composed of a signed interaction network over event types and a delay-aware monotonic temporal network. This design enables explicit characterization of inter-class influence topology -- including excitation, inhibition, and neutrality -- while flexibly capturing diverse temporal decay patterns and potential influence delays. For efficient learning, we develop a stratified Monte Carlo estimator for stochastic training. Extensive experiments on synthetic and real-world benchmark datasets validate the ability of our approach to uncover structured relationships and deliver strong predictive performance.

Problem

• neural point processes
• interpretable event interaction
• structured dependencies
• multi-class event streams
• influence topology

Innovation

• structured neural marked point process
• interpretable event interaction
• signed interaction network
• delay-aware temporal kernel
• stratified Monte Carlo estimator