Research map / Graph learning / GNN architectures
GraphMix: Improved Training of GNNs for Semi-Supervised Learning
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 GraphMix training to GCN | GCN · Cora, Citeseer, and Pubmed, standard split | test accuracy: 81.30$\pm$0.66 -> 83.94$\pm$0.57; 70.61$\pm$0.22 -> 74.72$\pm$0.59; 79.86$\pm$0.34 -> 80.98$\pm$0.55 | Helped |
| Add GraphMix training to GCN | GCN · Cora, Citeseer, and Pubmed, 10 random splits | test accuracy: 77.84$\pm$1.45 -> 82.07$\pm$1.17; 72.56$\pm$2.46 -> 76.45$\pm$1.57; 78.74$\pm$0.99 -> 80.72$\pm$1.08 | Helped |
| Add GraphMix training to GCN | GCN · Cora-Full, Coauthor-CS, and Coauthor-Physics | test accuracy: 60.13$\pm$0.57 -> 61.80$\pm$0.54; 91.27$\pm$0.56 -> 91.83$\pm$0.51; 92.90$\pm$0.92 -> 94.49$\pm$0.84 | Helped |
| Add GraphMix training to GCN | GCN · Bitcoin OTC and Bitcoin Alpha | F1 score: 65.72$\pm$0.38 -> 66.35$\pm$0.41; 64.00$\pm$0.19 -> 65.34$\pm$0.19 | Helped |
| Add Manifold Mixup and predicted targets to FCN training (ablation) | GraphMix with GCN · Cora, Citeseer, and Pubmed, 10 labels per class | test accuracy: 68.78$\pm$3.54 -> 79.30$\pm$1.36; 61.01$\pm$1.24 -> 70.78$\pm$1.41; 72.56$\pm$1.08 -> 77.13$\pm$3.60 | Helped |
| Add predicted targets to Manifold Mixup training (ablation) | GraphMix with GCN · Cora, Citeseer, and Pubmed, 10 labels per class | test accuracy: 73.55$\pm$3.29 -> 79.30$\pm$1.36; 65.72$\pm$2.10 -> 70.78$\pm$1.41; 75.74$\pm$1.69 -> 77.13$\pm$3.60 | Helped |
| Add Manifold Mixup to predicted-target training (ablation) | GraphMix with GCN · Cora, Citeseer, and Pubmed, 10 labels per class | test accuracy: 69.08$\pm$5.03 -> 79.30$\pm$1.36; 62.66$\pm$1.80 -> 70.78$\pm$1.41; 73.07$\pm$0.94 -> 77.13$\pm$3.60 | Helped |
| Use Manifold Mixup without predicted targets (ablation) | GraphMix with GCN · Cora, Citeseer, and Pubmed, 10 labels per class | test accuracy: 68.78$\pm$3.54 -> 73.55$\pm$3.29; 61.01$\pm$1.24 -> 65.72$\pm$2.10; 72.56$\pm$1.08 -> 75.74$\pm$1.69 | Helped |
Closest papers by meaning
- AGMixup: Adaptive Graph Mixup for Semi-supervised Node Classification (AAAI 2025 · similarity 0.87)
- Generalizing CNNs to graphs with learnable neighborhood quantization (NeurIPS 2024 · similarity 0.87)
- A Manifold Perspective on the Statistical Generalization of Graph Neural Networks (ICML 2025 · similarity 0.87)
- Normalize Then Propagate: Efficient Homophilous Regularization for Few-Shot Semi-Supervised Node Classification (AAAI 2025 · similarity 0.87)
- NodeMixup: Tackling Under-Reaching for Graph Neural Networks (AAAI 2024 · similarity 0.86)
- Diffusing to the Top: Boost Graph Neural Networks with Minimal Hyperparameter Tuning (ICLR 2025 · similarity 0.86)
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.