Abstract: Teaching Graph Neural Networks (GNNs) to accurately classify nodes under severely noisy labels is an important problem in real-world graph learning applications, but is currently underexplored. Although pairwise training methods have demonstrated promise in supervised metric learning and unsupervised contrastive learning, they remain less studied on noisy graphs, where the structural pairwise interactions (PI) between nodes are abundant and thus might benefit label noise learning rather than the pointwise methods. This paper bridges the gap by proposing a pairwise framework for noisy node classification on graphs, which relies on the PI as a primary learning proxy in addition to the pointwise learning from the noisy node class labels. Our proposed framework PI-GNN contributes two novel components: (1) a confidence-aware PI estimation model that adaptively estimates the PI labels, which are defined as whether the two nodes share the same node labels, and (2) a decoupled training approach that leverages the estimated PI labels to regularize a node classification model for robust node classification. Extensive experiments on different datasets and GNN architectures demonstrate the effectiveness of PI-GNN, yielding a promising improvement over the state-of-the-art methods. Code is publicly available at https://github.com/TianBian95/pi-gnn.
Submission Length: Regular submission (no more than 12 pages of main content)
Changes Since Last Submission: We thank all the reviewers for their time and valuable comments. We have addressed the reviewers’ comments and concerns in individual responses to each reviewer. The reviews allowed us to strengthen our manuscript and the major changes made (highlighted in blue) are summarized below:
+ [AoRd] Added sensitivity analysis to model initialization.
+ [AoRd] Added results on a larger graph dataset.
+ [AoRd, kfuH] Added discussion on related works.
+ [8ntc, kfuH] Fixed writing glitches.
+ [8ntc] Clarified the assumption on homophilous graphs.
+ [8ntc] Added experiments on PI estimation using low-rank approximation.
+ [8ntc] Added comparison among different training schemes.
+ [kfuH] Clarified the motivation for using PI estimation.
+ [kfuH] Added results on using subgraph sampling on smaller graphs.
+ [kfuH] Added comparison with baselines on the OGB dataset.
+ [kfuH] Clarified the manually generated noise.
Code: https://github.com/TianBian95/pi-gnn
Assigned Action Editor: ~Shinichi_Nakajima2
License: Creative Commons Attribution 4.0 International (CC BY 4.0)
Submission Number: 819
Loading