Intensity-Free Learning of Temporal Point Processes

TL;DR

This paper introduces intensity-free models for temporal point processes that directly model inter-event times, leveraging normalizing flows and mixture models to improve flexibility, efficiency, and applicability over traditional intensity-based methods.

Contribution

It proposes a novel approach to learn temporal point processes without intensity functions, using flow-based and mixture models for better flexibility and computational efficiency.

Findings

Achieves state-of-the-art performance in prediction tasks

Enables learning sequence embeddings and missing data imputation

Demonstrates flexibility and efficiency over traditional methods

Abstract

Temporal point processes are the dominant paradigm for modeling sequences of events happening at irregular intervals. The standard way of learning in such models is by estimating the conditional intensity function. However, parameterizing the intensity function usually incurs several trade-offs. We show how to overcome the limitations of intensity-based approaches by directly modeling the conditional distribution of inter-event times. We draw on the literature on normalizing flows to design models that are flexible and efficient. We additionally propose a simple mixture model that matches the flexibility of flow-based models, but also permits sampling and computing moments in closed form. The proposed models achieve state-of-the-art performance in standard prediction tasks and are suitable for novel applications, such as learning sequence embeddings and imputing missing data.