Predicting Higher-Order Dynamics With Unknown Hypergraph Topology

Journal Article (2024)
Author(s)

Zili Zhou (Fudan University)

C. Li (TU Delft - Network Architectures and Services, Fudan University)

Piet Mieghem (TU Delft - Network Architectures and Services)

Xiang Li (Tongji University)

Research Group
Network Architectures and Services
DOI related publication
https://doi.org/10.1109/TCSI.2024.3513406
More Info
expand_more
Publication Year
2024
Language
English
Research Group
Network Architectures and Services
Bibliographical Note
Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.@en
Issue number
4
Volume number
72
Reuse Rights

Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.

Abstract

Predicting future dynamics on networks is challenging, especially when the complete and accurate network topology is difficult to obtain in real-world scenarios. Moreover, the higher-order interactions among nodes, which have been found in a wide range of systems in recent years, such as the nets connecting multiple modules in circuits, further complicate accurate prediction of dynamics on hypergraphs. In this work, we proposed a two-step method called the topology-agnostic higher-order dynamics prediction (TaHiP) algorithm. The observations of nodal states of the target hypergraph are used to train a surrogate matrix, which is then employed in the dynamical equation to predict future nodal states in the same hypergraph, given the initial nodal states. TaHiP outperforms three latest Transformer-based prediction models in different real-world hypergraphs. Furthermore, experiments in synthetic and real-world hypergraphs show that the prediction error of the TaHiP algorithm increases with mean hyperedge size of the hypergraph, and could be reduced if the hyperedge size distribution of the hypergraph is known.

Files

Predicting_Higher-Order_Dynami... (pdf)
(pdf | 1.73 Mb)
- Embargo expired in 16-06-2025
License info not available