Research map / Graph learning / GNN architectures
Graph Mixture of Experts: Learning on Large-Scale Graphs with Explicit Diversity Modeling
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 |
|---|---|---|---|
| replace GCN layers with GMoE layers | GCN · ogbg-molhiv | ROC-AUC: 76.06 -> 77.87 | Helped |
| replace GCN layers with GMoE layers | GCN · ogbn-arxiv | accuracy: 71.74 -> 71.88 | No clear effect |
| replace GIN layers with GMoE layers (no pretrain) | GIN · ogbg-moltoxcast | ROC-AUC: 63.3 -> 62.86 | Hurt |
| add GMoE to pretrained GIN | GIN+Pretrain · ogbg-molbbbp | ROC-AUC: 69.94 -> 68.62 | Hurt |
| add GMoE to pretrained GIN | GIN+Pretrain · average of 4 datasets | ROC-AUC: 70.71 -> 71.31 | Helped |
| use all hop-2 experts instead of all hop-1 (ablation) | GMoE-GCN · ogbg-molhiv | ROC-AUC: 77.53 -> 77.87 | Helped |
| use all hop-2 experts instead of all hop-1 (ablation) | GMoE-GCN · ogbg-molfreesolv | RMSE: 2.532 -> 2.873 | Hurt |
| use sparse top-k (k=1) instead of dense (k=4) (ablation) | GMoE-GCN · ogbg-molfreesolv | RMSE: 2.856 -> 2.500 | Helped |
| add load-balancing loss (lambda 0.1) (ablation) | GMoE-GCN · ogbg-moltox21 | ROC-AUC: 72.87 -> 75.45 | Helped |
| integrate GMoE into Neural FingerPrints | Neural FingerPrints · ogbg-molhiv | ROC-AUC: +0.4 | Helped |
Closest papers by meaning
- Discriminative Mixture-of-Experts on Graphs with Reliable Expert Fusion (ICML 2026 · similarity 0.93)
- Self-Adaptive Graph Mixture of Models (AAAI 2026 · similarity 0.92)
- Mixture of Weak and Strong Experts on Graphs (ICLR 2024 · similarity 0.90)
- Graph Sparsification via Mixture of Graphs (ICLR 2025 · similarity 0.89)
- Graph Neural Network Generalization With Gaussian Mixture Model Based Augmentation (ICML 2025 · similarity 0.89)
- Mixture of Link Predictors on Graphs (NeurIPS 2024 · similarity 0.88)
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.