Research map / Graph learning / GNN architectures
Boosting Graph Neural Networks via Adaptive Knowledge Distillation
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 |
|---|---|---|---|
| Use BGNN instead of the strongest reported single-teacher baseline | GCN, GraphSage, or GAT student · Collab, IMDB, Enzymes, Cora, Citeseer, Pubmed, A-Computers | accuracy: 82.73±0.34 vs 81.33±0.31 (Collab); 84.63±0.28 vs 83.50±0.27 (Cora) | Helped |
| Add adaptive temperature based on teacher confidence (ablation) | BGNN with GCN student and GAT teacher · Collab, IMDB, Enzymes, Cora, Citeseer, Pubmed, A-Computers | accuracy: 81.27±0.25 -> 82.53±0.47 (Collab); 79.00±0.00 -> 79.33±0.47 (IMDB); 66.67±1.36 -> 68.89±0.79 (Enzymes); 83.30±0.22 -> 83 | Mixed |
| Add weight boosting for training samples misclassified by the teacher (ablation) | BGNN with GCN student and GAT teacher · Collab, IMDB, Enzymes, Cora, Citeseer, Pubmed, A-Computers | accuracy: 82.53±0.09 -> 82.53±0.47 (Collab); 77.67±0.47 -> 79.33±0.47 (IMDB); 68.33±1.36 -> 68.89±0.79 (Enzymes); 83.17±0.42 -> 83 | Mixed |
| Train with multiple teachers sequentially rather than one teacher | BGNN student · Seven classification datasets | accuracy | Mixed |
| Distill sequentially rather than combining teacher outputs in parallel | Multi-teacher GNN distillation · Seven classification datasets | accuracy: 20 out of 21 cases | Helped |
Closest papers by meaning
- Large Language Model Meets Graph Neural Network in Knowledge Distillation (AAAI 2025 · similarity 0.89)
- VQGraph: Rethinking Graph Representation Space for Bridging GNNs and MLPs (ICLR 2024 · similarity 0.89)
- TINED: GNNs-to-MLPs by Teacher Injection and Dirichlet Energy Distillation (ICML 2025 · similarity 0.88)
- Diffusing to the Top: Boost Graph Neural Networks with Minimal Hyperparameter Tuning (ICLR 2025 · similarity 0.88)
- How Graph Neural Networks Learn: Lessons from Training Dynamics (ICML 2024 · similarity 0.88)
- Preference-driven Knowledge Distillation for Few-shot Node Classification (NeurIPS 2025 · similarity 0.87)
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.