Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Towards Deeper Graph Neural Networks with Differentiable Group Normalization

About

Graph neural networks (GNNs), which learn the representation of a node by aggregating its neighbors, have become an effective computational tool in downstream applications. Over-smoothing is one of the key issues which limit the performance of GNNs as the number of layers increases. It is because the stacked aggregators would make node representations converge to indistinguishable vectors. Several attempts have been made to tackle the issue by bringing linked node pairs close and unlinked pairs distinct. However, they often ignore the intrinsic community structures and would result in sub-optimal performance. The representations of nodes within the same community/class need be similar to facilitate the classification, while different classes are expected to be separated in embedding space. To bridge the gap, we introduce two over-smoothing metrics and a novel technique, i.e., differentiable group normalization (DGN). It normalizes nodes within the same group independently to increase their smoothness, and separates node distributions among different groups to significantly alleviate the over-smoothing issue. Experiments on real-world datasets demonstrate that DGN makes GNN models more robust to over-smoothing and achieves better performance with deeper GNNs.

Kaixiong Zhou, Xiao Huang, Yuening Li, Daochen Zha, Rui Chen, Xia Hu• 2020

Related benchmarks

TaskDatasetResultRank
Node ClassificationCora (test)
Mean Accuracy76.3
951
Node ClassificationCiteseer (test)
Accuracy0.545
945
Node ClassificationPubMed (test)
Accuracy78.43
586
Node ClassificationReddit (test)
Accuracy51.22
201
Graph RegressionZINC 12K (test)
MAE0.2705
173
Graph ClassificationMolHIV
ROC AUC74.37
102
Graph ClassificationMOLBACE
ROC AUC0.7596
42
Graph ClassificationMOLTOX21
ROC-AUC0.7448
42
Molecular property predictionMOLESOL
RMSE1.087
37
Showing 9 of 9 rows

Other info

Follow for update