Meta Temporal Point Processes

TL;DR

This paper introduces a novel meta learning approach for temporal point processes by framing them as neural processes, utilizing context sets and local history matching to improve modeling of event sequences.

Contribution

It proposes a new meta learning framework for TPPs using neural processes, incorporating context sets and local history matching for enhanced sequence modeling.

Findings

Outperforms existing TPP methods on benchmark datasets

Demonstrates improved modeling of event sequences

Shows the effectiveness of neural process framing for TPPs

Abstract

A temporal point process (TPP) is a stochastic process where its realization is a sequence of discrete events in time. Recent work in TPPs model the process using a neural network in a supervised learning framework, where a training set is a collection of all the sequences. In this work, we propose to train TPPs in a meta learning framework, where each sequence is treated as a different task, via a novel framing of TPPs as neural processes (NPs). We introduce context sets to model TPPs as an instantiation of NPs. Motivated by attentive NP, we also introduce local history matching to help learn more informative features. We demonstrate the potential of the proposed method on popular public benchmark datasets and tasks, and compare with state-of-the-art TPP methods.