Research map / Graph learning / GNN architectures
Measuring and Relieving the Over-Smoothing Problem for Graph Neural Networks from the Topological View
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add AdaEdge prediction-guided edge adjustment | GCN · CORA, CiteSeer, PubMed, Amazon Photo, Coauthor Physics | accuracy (%): CORA: 81.2 ±0.8 -> 82.3**_{±0.8}; CiteSeer: 69.3 ±0.7 -> 69.7**_{±0.9}; PubMed: 76.3 ±0.5 -> 77.4**_{±0.5}; Amazon Photo | Helped |
| Add AdaEdge prediction-guided edge adjustment | ChebGCN · CORA, CiteSeer, PubMed, Amazon Computers, Coauthor CS | accuracy (%): CORA: 78.6 ±0.6 -> 80.1**_{±0.5}; CiteSeer: 67.4 ±1.0 -> 67.8*_{±1.2}; PubMed: 76.7 ±0.1 -> 77.5**_{±0.6}; Amazon Comput | Helped |
| Add AdaEdge prediction-guided edge adjustment | GAT · CORA, CiteSeer, PubMed, Amazon Photo, Coauthor CS, Coauthor Physics | accuracy (%): CORA: 76.3 ±3.1 -> 77.9**_{±2.0}; CiteSeer: 68.9 ±0.6 -> 69.1*_{±0.8}; PubMed: 75.9 ±0.5 -> 76.6**_{±0.2}; Amazon Photo: | Helped |
| Add AdaEdge prediction-guided edge adjustment | GAT · Amazon Computers | accuracy (%): 81.4 ±1.5 -> 81.1^{*}_{±1.6} | Hurt |
| Add AdaEdge prediction-guided edge adjustment | HighOrder · CORA, CiteSeer, Amazon Computers | accuracy (%): CORA: 76.6 ±1.2 -> 72.5**_{±4.1}; CiteSeer: 64.2 ±1.0 -> 63.3**_{±1.0}; Amazon Computers: 26.3 ±12.7 -> 23.9*_{±13.4} | Hurt |
| Add AdaEdge prediction-guided edge adjustment | GGNN · CORA, CiteSeer | accuracy (%): CORA: 47.3 ±6.1 -> 44.7**_{±3.5}; CiteSeer: 55.5 ±2.8 -> 47.9**_{±3.4} | Hurt |
| Add AdaEdge prediction-guided edge adjustment | GNN models · 7 node-classification datasets | accuracy (%) | Mixed |
| Add MADGap regularizer (ablation) | 1-layer GCN · CORA, CiteSeer, PubMed | accuracy (%): CORA: 54.6 -> 54.9; CiteSeer: 39.7 -> 41.6; PubMed: 52.8 -> 57.1 | Helped |
| Add MADGap regularizer (ablation) | 2-layer GCN · CORA, CiteSeer, PubMed | accuracy (%): CORA: 80.8 -> 80.6; CiteSeer: 68.7 -> 67.1; PubMed: 76.6 -> 75.2 | Mixed |
| Add MADGap regularizer (ablation) | 3-layer GCN · CORA, CiteSeer, PubMed | accuracy (%): CORA: 79.0 -> 79.9; CiteSeer: 65.9 -> 65.4; PubMed: 76.2 -> 76.5 | Mixed |
| Add MADGap regularizer (ablation) | 4-layer GCN · CORA, CiteSeer, PubMed | accuracy (%): CORA: 70.8 -> 72.1; CiteSeer: 55.6 -> 62.2; PubMed: 67.3 -> 69.4 | Helped |
| Add MADGap regularizer (ablation) | 5-layer GCN · CORA, CiteSeer, PubMed | accuracy (%): CORA: 46.9 -> 48.0; CiteSeer: 35.8 -> 46.5; PubMed: 55.1 -> 56.1 | Helped |
Closest papers by meaning
- Deep Graph Neural Networks via Posteriori-Sampling-based Node-Adaptative Residual Module (NeurIPS 2024 · similarity 0.90)
- Making Classic GNNs Strong Baselines Across Varying Homophily: A Smoothness–Generalization Perspective (NeurIPS 2025 · similarity 0.89)
- Two Heads Are Better Than One: Boosting Graph Sparse Training via Semantic and Topological Awareness (ICML 2024 · similarity 0.89)
- Improving Graph Neural Networks by Learning Continuous Edge Directions (ICLR 2025 · similarity 0.89)
- A Manifold Perspective on the Statistical Generalization of Graph Neural Networks (ICML 2025 · similarity 0.89)
- Graph Neural Networks Are More Than Filters: Revisiting and Benchmarking from A Spectral Perspective (ICLR 2025 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.