A Survey of Graph Neural Networks and Their Industrial Applications DOI
Haoran Lu, Lei Wang, Xiaoliang Ma

et al.

Published: Jan. 1, 2024

Graph Neural Networks (GNNs) have emerged as a powerful tool for analyzing and modeling graph-structured data. In recent years, GNNs gained significant attention in various domains. This review paper aims to provide an overview of the state-of-the-art graph neural network techniques their industrial applications.First, we introduce fundamental concepts architectures GNNs, highlighting ability capture complex relationships dependencies We then delve into variants evolution graphs, including directed heterogeneous dynamic hypergraphs. Next, discuss interpretability GNN, GNN theory augmentation, expressivity, over-smoothing.Finally, specific use cases settings, finance, biology, knowledge recommendation systems, Internet Things (IoT), distillation. highlights immense potential solving real-world problems, while also addressing challenges opportunities further advancement this field.

Language: Английский

Current and future directions in network biology DOI Creative Commons
Marinka Žitnik, Michelle M. Li, A. V. Wells

et al.

Bioinformatics Advances, Journal Year: 2024, Volume and Issue: 4(1)

Published: Jan. 1, 2024

Abstract Summary Network biology is an interdisciplinary field bridging computational and biological sciences that has proved pivotal in advancing the understanding of cellular functions diseases across systems scales. Although been around for two decades, it remains nascent. It witnessed rapid evolution, accompanied by emerging challenges. These stem from various factors, notably growing complexity volume data together with increased diversity types describing different tiers organization. We discuss prevailing research directions network biology, focusing on molecular/cellular networks but also other such as biomedical knowledge graphs, patient similarity networks, brain social/contact relevant to disease spread. In more detail, we highlight areas inference comparison multimodal integration heterogeneous higher-order analysis, machine learning network-based personalized medicine. Following overview recent breakthroughs these five areas, offer a perspective future biology. Additionally, scientific communities, educational initiatives, importance fostering within field. This article establishes roadmap immediate long-term vision Availability implementation Not applicable.

Language: Английский

Citations

19
Multi-HGNN: Multi-modal hypergraph neural networks for predicting missing reactions in metabolic networks DOI

Yamei Huang,

Xudong Liang, Tao Lin

et al.

Information Sciences, Journal Year: 2025, Volume and Issue: unknown, P. 121960 - 121960

Published: Feb. 1, 2025

Language: Английский

Citations

1
DHHNN: A Dynamic Hypergraph Hyperbolic Neural Network based on variational autoencoder for multimodal data integration and node classification DOI

Zhangyu Mei,

Xiao Bi,

Dianguo Li

et al.

Information Fusion, Journal Year: 2025, Volume and Issue: unknown, P. 103016 - 103016

Published: Feb. 1, 2025

Language: Английский

Citations

1
A Survey on Hypergraph Mining: Patterns, Tools, and Generators DOI Creative Commons
Geon Lee, Fanchen Bu, Tina Eliassi‐Rad

et al.

ACM Computing Surveys, Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 20, 2025

Hypergraphs, which belong to the family of higher-order networks, are a natural and powerful choice for modeling group interactions in real world. For example, when collaboration may involve not just two but three or more people, use hypergraphs allows us explore beyond pairwise (dyadic) patterns capture groupwise (polyadic) patterns. The mathematical complexity offers both opportunities challenges hypergraph mining. goal mining is find structural properties recurring real-world across different domains, we call To patterns, need tools. We divide tools into categories: (1) null models (which help test significance observed patterns), (2) elements (i.e., substructures such as open closed triangles), (3) quantities numerical computing transitivity). There also generators, whose objective produce synthetic that faithful representation hypergraphs. In this survey, provide comprehensive overview current landscape mining, covering tools, generators. taxonomies each offer in-depth discussions future research on

Language: Английский

Citations

1
EduGraph: Learning Path-Based Hypergraph Neural Networks for MOOC Course Recommendation DOI
Ming Li, Zhao Li, Changqin Huang

et al.

IEEE Transactions on Big Data, Journal Year: 2024, Volume and Issue: 10(6), P. 706 - 719

Published: Sept. 3, 2024

Language: Английский

Citations

8
Molecular hypergraph neural networks DOI Creative Commons
Junwu Chen, Philippe Schwaller

The Journal of Chemical Physics, Journal Year: 2024, Volume and Issue: 160(14)

Published: April 10, 2024

Graph neural networks (GNNs) have demonstrated promising performance across various chemistry-related tasks. However, conventional graphs only model the pairwise connectivity in molecules, failing to adequately represent higher order connections, such as multi-center bonds and conjugated structures. To tackle this challenge, we introduce molecular hypergraphs propose Molecular Hypergraph Neural Networks (MHNNs) predict optoelectronic properties of organic semiconductors, where hyperedges A general algorithm is designed for irregular high-order which can efficiently operate on with orders. The results show that MHNN outperforms all baseline models most tasks photovoltaic, OCELOT chromophore v1, PCQM4Mv2 datasets. Notably, achieves without any 3D geometric information, surpassing utilizes atom positions. Moreover, better than pretrained GNNs under limited training data, underscoring its excellent data efficiency. This work provides a new strategy more representations property prediction related connections.

Language: Английский

Citations

7
Multi-modal trajectory forecasting with Multi-scale Interactions and Multi-pseudo-target Supervision DOI
Cong Zhao,

Andi Song,

Zimu Zeng

et al.

Knowledge-Based Systems, Journal Year: 2024, Volume and Issue: 296, P. 111903 - 111903

Published: May 14, 2024

Language: Английский

Citations

6
Multi-aspect Knowledge-enhanced Hypergraph Attention Network for Conversational Recommendation Systems DOI
Xiaokang Li, Yihao Zhang, Yonghao Huang

et al.

Knowledge-Based Systems, Journal Year: 2024, Volume and Issue: 299, P. 112119 - 112119

Published: June 13, 2024

Language: Английский

Citations

5
Hypergraph Transformer for Semi-Supervised Classification DOI Open Access
Zexi Liu, Bohan Tang,

Ziyuan Ye

et al.

ICASSP 2022 - 2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Journal Year: 2024, Volume and Issue: unknown, P. 7515 - 7519

Published: March 18, 2024

Hypergraphs play a pivotal role in the modelling of data featuring higher-order relations involving more than two entities. Hypergraph neural networks emerge as powerful tool for processing hypergraph-structured data, delivering remarkable performance across various tasks, e.g., hypergraph node classification. However, these models struggle to capture global structural information due their reliance on local message passing. To address this challenge, we propose novel learning framework, HyperGraph Transformer (HyperGT). HyperGT uses Transformer-based network architecture effectively consider correlations among all nodes and hyperedges. incorporate information, has distinct designs: i) positional encoding based incidence matrix, offering valuable insights into node-node hyperedge-hyperedge interactions; ii) structure regularization loss function, capturing connectivities between Through designs, achieves comprehensive representation by incorporating interactions while preserving connectivity patterns. Extensive experiments conducted real-world classification tasks showcase that consistently outperforms existing methods, establishing new state-of-the-art benchmarks. Ablation studies affirm effectiveness individual designs our model.

Language: Английский

Citations

4
Developing a novel approach in estimating urban commute traffic by integrating community detection and hypergraph representation learning DOI Creative Commons
Yuhuan Li, Shaowu Cheng, Yuxiang Feng

et al.

Expert Systems with Applications, Journal Year: 2024, Volume and Issue: 249, P. 123790 - 123790

Published: March 21, 2024

The efficiency of urban traffic management and congestion alleviation relies heavily on accurate forecasting Origin-Destination (O-D) demand matrices. Existing models primarily focus estimating O-D for various travel purposes throughout the day, which is characterised by its pulsating nature. However, these often compromise precision peak-hour forecasts, leading to unreliable dynamic control challenges in effectively reducing congestion. To tackle this challenge, paper proposes a novel method predicting commuting Our employs community detection algorithms road networks precisely partition commute regions, incorporating Points Interest (POIs). We also present spatio-temporal weighted hypergraph model that leverages partitioned time characteristics from observed trips, meteorological data improve forecasting. Comparative analyses with contemporary ablation studies indicate our significantly enhances prediction accuracy, approximately 5%. These findings imply proposed more encompasses varied during peak hours, thereby providing matrices management.

Language: Английский

Citations

4