On Addressing the Limitations of Graph Neural Networks

This report gives a summary of two problems about graph convolutional networks (GCNs): over-smoothing and heterophily challenges, and outlines future directions to explore.

cs.LG Advanced Graph Neural Networks Computer Science Mathematics

View graph Source

Paper

References (72)

01When Do We Need GNN for Node Classification?Sitao Luan, Chenqing Hua, Qincheng Lu et al.2022 · arXiv.org · 19 citations

02Revisiting Heterophily For Graph Neural NetworksSitao Luan, Chenqing Hua, Qincheng Lu et al.2022 · Neural Information Processing Systems · 326 citations In Library

03Graph Neural Networks Intersect Probabilistic Graphical Models: A SurveyChenqing Hua2022 · arXiv.org · 7 citations In Library

04Is Heterophily A Real Nightmare For Graph Neural Networks To Do Node Classification?Sitao Luan, Chenqing Hua, Qincheng Lu et al.2021 · arXiv.org · 131 citations In Library

05A Consciousness-Inspired Planning Agent for Model-Based Reinforcement LearningMingde Zhao, Zhen Liu, Sitao Luan et al.2021 · Neural Information Processing Systems · 41 citations In Library

06New Benchmarks for Learning on Non-Homophilous GraphsDerek Lim, Xiuyu Li, Felix Hohne et al.2021 · arXiv.org · 121 citations In Library

07Two Sides of the Same Coin: Heterophily and Oversmoothing in Graph Convolutional Neural NetworksYujun Yan, Milad Hashemi, Kevin Swersky et al.2021 · Industrial Conference on Data Mining · 310 citations In Library

08Beyond Low-frequency Information in Graph Convolutional NetworksDeyu Bo, Xiao Wang, C. Shi et al.2021 · AAAI Conference on Artificial Intelligence · 817 citations In Library

09Reward Propagation Using Graph Convolutional NetworksMartin Klissarov, Doina Precup2020 · Neural Information Processing Systems · 26 citations In Library

10Graph Neural Networks with HeterophilyJiong Zhu, Ryan A. Rossi, Anup B. Rao et al.2020 · AAAI Conference on Artificial Intelligence · 409 citations In Library

11Graph Representation LearningWilliam L. Hamilton2020 · Synthesis Lectures on Artificial Intelligence and Machine Learning · 1.3k citations

12Training Matters: Unlocking Potentials of Deeper Graph Convolutional Neural NetworksSitao Luan, Mingde Zhao, X. Chang et al.2020 · International Workshop on Complex Networks & Their Applications · 11 citations In Library

Scroll for more · 38 remaining

Similar papers

On Addressing the Limitations of Graph Neural NetworksSitao Luan2023 · Preprints.org In Library

On Addressing the Limitations of Graph Neural NetworksSitao Luan2023 · Zenodo (CERN European Organization for Nuclear Research) In Library

Two Sides of the Same Coin: Heterophily and Oversmoothing in Graph Convolutional Neural NetworksYujun Yan, Milad Hashemi, Kevin Swersky et al.2022 · 2022 IEEE International Conference on Data Mining (ICDM) In Library

Two Sides of the Same Coin: Heterophily and Oversmoothing in Graph Convolutional Neural NetworksYujun Yan, Milad Hashemi, Kevin Swersky et al.2021 · arXiv (Cornell University) · 309 citations In Library

Addressing the Limitations of Graph Neural Networks on Node-level TasksSitao Luan2024 In Library

Graph Neural Networks Do Not Always OversmoothBastian Epping, Alexandre René, Moritz Helias et al.2024 · arXiv (Cornell University) · 16 citations In Library

Deeper Insights into Graph Convolutional Networks for Semi-Supervised LearningQimai Li, Zhichao Han, Xiao-Ming Wu2018 · AAAI Conference on Artificial Intelligence · 3.5k citations In Library

Review of Image Classification Algorithms Based on Graph Convolutional NetworksWenhao Tang2023 · EAI Endorsed Transactions on AI and Robotics In Library

Tackling Over-Smoothing: Graph Hollow Convolution Network with Topological Layer FusionJunhong Wan, Yao Fu, Weihao Jiang et al.2022 · 2022 International Joint Conference on Neural Networks (IJCNN) In Library

Towards a Spectrum of Graph Convolutional NetworksMathias Niepert, Alberto García-Durán2018 · Data Science Workshop · 2 citations In Library

Patents near this paper

GRAPH CONVOLUTIONAL NETWORKS WITH MOTIF-BASED ATTENTIONUS 11,544,535 · ADOBE INC. · 2023 SGCNN: Structural Graph Convolutional Neural NetworkUS 11,853,903 · SIEMENS AKTIENGESELLSCHAFT · 2023 SPATIO-TEMPORAL ANOMALY DETECTION IN COMPUTER NETWORKS USING GRAPH CONVOLUTIONAL RECURRENT NEURAL NETWORKS (GCRNNS)US 10,771,488 · CISCO TECHNOLOGY, INC. · 2020 Deep Learning Testability Analysis with Graph Convolutional NetworksUS 10,657,306 · NVIDIA CORP. · 2020 GRAPH DATA PROCESSING METHOD, DEVICE, AND COMPUTER PROGRAM PRODUCTUS 11,609,936 · EMC IP HOLDING COMPANY LLC · 2023

Source of record

arXiv