Research map / Graph learning / Knowledge and temporal graphs
Distilling Knowledge From Graph Convolutional Networks
Check 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 local-structure distillation with an RBF kernel | GAT student · PPI test | F1 Score: 95.7 -> 96.1 | Helped |
| Add local-structure distillation with an RBF kernel | DGCNN student · ModelNet40 | Acc: 91.2 -> 91.9 | Helped |
| Add local-structure distillation with an RBF kernel | DGCNN student · ModelNet40 | mAcc: 87.5 -> 88.6 | Helped |
| Use local-structure distillation instead of KD | DGCNN student · ModelNet40 | Acc: 91.6 -> 91.9 | Helped |
| Use local-structure distillation instead of attention transfer | DGCNN student · ModelNet40 | Acc: 91.6 -> 91.9 | Helped |
| Use local-structure distillation instead of FitNet | DGCNN student · ModelNet40 | Acc: 91.1 -> 91.9 | Helped |
| Use L2 Norm instead of RBF kernel (ablation) | LSP with DGCNN student · ModelNet40 | Acc: 91.9 -> 91.4 | Hurt |
| Use polynomial kernel instead of RBF kernel (ablation) | LSP with DGCNN student · ModelNet40 | Acc: 91.9 -> 91.7 | Hurt |
| Use linear kernel instead of RBF kernel (ablation) | LSP with DGCNN student · ModelNet40 | Acc: 91.9 -> 91.5 | Hurt |
| Add more channels to each layer (ablation) | LSP DGCNN student · ModelNet40 | Acc: 91.9 -> 92.3 | Helped |
| Add one graph convolutional layer (ablation) | LSP DGCNN student · ModelNet40 | Acc: 91.9 -> 92.1 | Helped |
| Add one fully connected layer (ablation) | LSP DGCNN student · ModelNet40 | Acc: 91.9 -> 91.8 | No clear effect |
Closest papers by meaning
- Multi-Scale Distillation from Multiple Graph Neural Networks (AAAI 2022 · similarity 0.92)
- End-to-End Structure-Aware Convolutional Networks for Knowledge Base Completion (AAAI 2019 · similarity 0.91)
- Demystifying GNN-to-MLP Knowledge Transfer: Theoretical Grounding and Dual-Stream Distillation Method (AAAI 2026 · similarity 0.90)
- DistillHGNN: A Knowledge Distillation Approach for High-Speed Hypergraph Neural Networks (ICLR 2025 · similarity 0.90)
- VQGraph: Rethinking Graph Representation Space for Bridging GNNs and MLPs (ICLR 2024 · similarity 0.90)
- Large Language Model Meets Graph Neural Network in Knowledge Distillation (AAAI 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.