Evolving-Graph Gaussian Processes

TL;DR

This paper introduces evolving-Graph Gaussian Processes (e-GGPs), a novel method for modeling dynamic graph structures over time, improving upon static GGPs for time-series regression tasks involving evolving graphs.

Contribution

The paper presents e-GGPs, a new approach that learns transition functions on dynamic graphs using a neighborhood kernel, extending GGPs to evolving graph data.

Findings

e-GGPs outperform static GGPs in time-series regression on evolving graphs

The method effectively models connectivity and interaction changes over time

Demonstrated benefits in real-world dynamic graph scenarios

Abstract

Graph Gaussian Processes (GGPs) provide a data-efficient solution on graph structured domains. Existing approaches have focused on static structures, whereas many real graph data represent a dynamic structure, limiting the applications of GGPs. To overcome this we propose evolving-Graph Gaussian Processes (e-GGPs). The proposed method is capable of learning the transition function of graph vertices over time with a neighbourhood kernel to model the connectivity and interaction changes between vertices. We assess the performance of our method on time-series regression problems where graphs evolve over time. We demonstrate the benefits of e-GGPs over static graph Gaussian Process approaches.