Node Classification Beyond Homophily: Towards a General Solution

Xu, Zhe; Chen, Yuzhong; Zhou, Qinghai; Wu, Yuhang; Pan, Menghai; Yang, Hao; Tong, Hanghang

doi:10.1145/3580305.3599446

Cited by 5 publications

(2 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Leskovec 2017; Veličković et al 2018). Recent GCN approaches over heterophilic graphs can be grouped into multihop-based ones (Abu-El-Haija et al 2019;Zhu et al 2020;Jin et al 2021b;Wang and Derr 2021;Wang et al 2022b), ranking-based ones (Liu, Wang, and Ji 2021;Wang et al 2022a;Yang et al 2022), and the ones using GCN architecture refinement (Bo et al 2021;Yang et al 2021;Suresh et al 2021a;Yan et al 2021;Luan et al 2022;Xu et al 2023;Li, Kim, and Wang 2023;Zheng et al 2023). These methods have achieved remarkable success in graph node classification.…”

Section: Related Work Graph Convolution Networkmentioning

confidence: 99%

“…UGCN and SimP-GCN employ a similarity preservation scheme for structure learning on heterophilic graphs and BM-GCN employs a selective aggregation on structure via a block-guided strategy. We also compare our model with a recently proposed spectral-based method ALT-GCN (Xu et al 2023). Settings: We implement our method by Pytorch and Pytorch Geometric and use Adam Optimizer on all datasets with the learning rate as 0.001.…”

Section: Experiments Datasets Baselines and Settingsmentioning

confidence: 99%

See 1 more Smart Citation

Refining Latent Homophilic Structures over Heterophilic Graphs for Robust Graph Convolution Networks

Qiu,

Nan,

Xiong

et al. 2024

AAAI

View full text Add to dashboard Cite

Graph convolution networks (GCNs) are extensively utilized in various graph tasks to mine knowledge from spatial data. Our study marks the pioneering attempt to quantitatively investigate the GCN robustness over omnipresent heterophilic graphs for node classification. We uncover that the predominant vulnerability is caused by the structural out-of-distribution (OOD) issue. This finding motivates us to present a novel method that aims to harden GCNs by automatically learning Latent Homophilic Structures over heterophilic graphs. We term such a methodology as LHS. To elaborate, our initial step involves learning a latent structure by employing a novel self-expressive technique based on multi-node interactions. Subsequently, the structure is refined using a pairwisely constrained dual-view contrastive learning approach. We iteratively perform the above procedure, enabling a GCN model to aggregate information in a homophilic way on heterophilic graphs. Armed with such an adaptable structure, we can properly mitigate the structural OOD threats over heterophilic graphs. Experiments on various benchmarks show the effectiveness of the proposed LHS approach for robust GCNs.

show abstract