Research map / Label-efficient and robust learning / Noisy labels
Prototype-Guided Supervision for Graph Learning with Noisy and Sparse Labels
Qiyu Li, Xianxian Li, De Li, Jinyan Wang
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 |
|---|---|---|---|
| add mixup CE, orthogonality and radius regularization on top of prototype CE (ablation) | prototype-guided GNN · CiteSeer, uniform noise 0.6 | accuracy: 49.34 -> 60.08 | Improved |
| add mixup CE, orthogonality and radius regularization on top of prototype CE (ablation) | prototype-guided GNN · CiteSeer, pair noise 0.4 | accuracy: 64.31 -> 72.21 | Improved |
| add radius regularization (compactness) to the objective (ablation) | prototype-guided GNN · CiteSeer high noise | accuracy | Improved |
| add orthogonality constraint on prototypes (ablation) | prototype-guided GNN · CiteSeer | accuracy | Improved |
| supervise on class prototypes instead of training node labels | GNN classifier · noisy-label node classification (uniform 0.2, 0.4) | accuracy | Improved |
| use the proposed method versus DNDNet on Cora ML at 0.6 uniform noise | prototype-guided GNN · Cora ML, uniform noise 0.6 | accuracy: 63.40 -> 70.04 | Improved |
| clean-label setting on PubMed versus CRGNN | prototype-guided GNN · PubMed, clean | accuracy: 83.23 -> 82.32 | Hurt |
| vary top-k competing classes in inter-class mixup (ablation) | prototype-guided GNN · CiteSeer | accuracy | No clear effect |
| raise noise rate from 0.2 to 0.4 when varying regularization strengths (ablation) | prototype-guided GNN · CiteSeer, gamma_ortho and gamma_rad | accuracy | Mixed |
Closest papers by meaning
- DREAM: Dual-Standard Semantic Homogeneity with Dynamic Optimization for Graph Learning with Label Noise (ICML 2026 · similarity 0.88)
- Decoupled Self-supervised Learning for Graphs (NeurIPS 2022 · similarity 0.88)
- Robust Node Classification on Graph Data with Graph and Label Noise (AAAI 2024 · similarity 0.87)
- Posterior Label Smoothing for Node Classification (AAAI 2026 · similarity 0.87)
- HGDL: Heterogeneous Graph Label Distribution Learning (NeurIPS 2024 · similarity 0.86)
- Two Heads Are Better Than One: Boosting Graph Sparse Training via Semantic and Topological Awareness (ICML 2024 · 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.
