Research map / Graph learning / GNN architectures
Dirichlet Energy Constrained Learning for Deep Graph Neural Networks
Kaixiong Zhou, Xiao Huang, Daochen Zha, Rui Chen, Li Li, Soo-Hyun Choi, Xia Hu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 EGNN instead of the strongest reported baseline at 64 layers | Cora node classification · Cora, 64 layers | accuracy (%): 85.4 -> 85.7 | Improved |
| Use EGNN instead of the strongest reported baseline at 64 layers | Pubmed node classification · Pubmed, 64 layers | accuracy (%): 79.9 -> 80.1 | Improved |
| Use EGNN instead of the strongest reported baseline at 32 layers | Coauthor-Physics node classification · Coauthor-Physics, 32 layers | accuracy (%): 92.9 -> 93.3 | Improved |
| Use EGNN instead of the strongest reported baseline at 32 layers | Ogbn-arxiv node classification · Ogbn-arxiv, 32 layers | accuracy (%): 71.4 -> 72.7 | Improved |
| Replace orthogonal weight initialization and regularization with Glorot initialization and Frobenius norm regularization (ablation) | EGNN · Cora, 64 layers | accuracy (%): 85.7 -> 23.6 | Hurt |
| Replace orthogonal weight initialization and regularization with Glorot initialization and Frobenius norm regularization (ablation) | EGNN · Ogbn-arxiv, 32 layers | accuracy (%): 72.7 -> 72.7 | No clear effect |
| Remove the lower energy limit by setting c_min to 0 (ablation) | EGNN · Cora, 64 layers | accuracy (%): 85.7 -> 12.9 | Hurt |
| Increase the lower limit c_min to 0.95 (ablation) | EGNN · Cora, 64 layers | accuracy (%): 85.7 -> 71.5 | Hurt |
| Replace SReLU with linear identity activation (ablation) | EGNN · Cora, 64 layers | accuracy (%): 85.7 -> 85.5 | No clear effect |
| Replace SReLU with ReLU (ablation) | EGNN · Ogbn-arxiv, 32 layers | accuracy (%): 72.7 -> 72.4 | Hurt |
Closest papers by meaning
- MuseGNN: Forming Scalable, Convergent GNN Layers that Minimize a Sampling-Based Energy (ICLR 2025 · similarity 0.89)
- Graph Neural Networks Exponentially Lose Expressive Power for Node Classification (ICLR 2020 · similarity 0.89)
- TopoTune: A Framework for Generalized Combinatorial Complex Neural Networks (ICML 2025 · similarity 0.89)
- TINED: GNNs-to-MLPs by Teacher Injection and Dirichlet Energy Distillation (ICML 2025 · similarity 0.88)
- How Graph Neural Networks Learn: Lessons from Training Dynamics (ICML 2024 · similarity 0.87)
- Can Classic GNNs Be Strong Baselines for Graph-level Tasks? Simple Architectures Meet Excellence (ICML 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.
