Research map / Label-efficient and robust learning / Noisy labels
Maximum Spanning Tree Guided Confidence and Sparse Graph for Robust Noisy Label Learning
Gengfeng Chen, Xu Liu, Boyi Peng, Liangqiu Xiao, Weicheng Xie, Siyang Song, Zitong Yu, Laizhong Cui, Linlin Shen
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 MST-GCSN versus ELR/DULC-style baselines | PreAct ResNet-18 · CIFAR overall average | test accuracy: 82.2 (DULC) -> 82.6 | Improved |
| add MST-GCSN at 90% symmetric noise on CIFAR-100 | PreAct ResNet-18 · CIFAR-100 (90% sym noise) | test accuracy: 54.5 (TCL) -> 56.3 | Improved |
| add MST-GCSN at 80-90% sym noise on CIFAR-10 | PreAct ResNet-18 · CIFAR-10 (80-90% sym noise) | test accuracy: DULC 95.0/93.5 vs 94.7/93.2 | No clear effect |
| add Confidence-Gated Propagation on MST (ablation) | Mixup + CE baseline · CIFAR-10 (90% sym noise) | test accuracy: 18.3 -> 63.1 | Improved |
| add local label estimate in PLP (ablation) | CGP model · CIFAR-10 (90% sym noise) | test accuracy: 63.1 -> 83.6 | Improved |
| add anchor-guided global label estimate in PLP (ablation) | CGP model · CIFAR-10 (90% sym noise) | test accuracy: 63.1 -> 81.9 | Improved |
| combine local and global estimates in PLP (ablation) | CGP model · CIFAR-10 (90% sym noise) | test accuracy: 83.6 -> 87.3 | Improved |
| add entropy-weighted CE loss (ablation) | CGP + PLP · CIFAR-10 (90% sym noise) | test accuracy: 87.3 -> 90.9 | Improved |
| add entropy-weighted CE loss (ablation) | CGP + PLP · CIFAR-100 (50% sym noise) | test accuracy: 70.1 -> 73.5 | Improved |
| add supervised contrastive loss (ablation) | CGP + PLP + WCE · CIFAR-10 (90% sym noise) | test accuracy: 90.9 -> 93.2 | Improved |
| use MST instead of kNN graph (ablation) | graph construction · unreliable edge ratio | unreliable edge ratio (%): 10.05 (5-NN) -> 7.10 | Improved |
Closest papers by meaning
- Identifying and Correcting Label Noise for Robust GNNs via Influence Contradiction (ICML 2026 · similarity 0.89)
- Training Robust Graph Neural Networks by Modeling Noise Dependencies (NeurIPS 2025 · similarity 0.87)
- NoisyGL: A Comprehensive Benchmark for Graph Neural Networks under Label Noise (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)
- Learning with Structural Labels for Learning with Noisy Labels (CVPR 2024 · similarity 0.85)
- NLGT: Neighborhood-based and Label-enhanced Graph Transformer Framework for Node Classification (AAAI 2025 · similarity 0.85)
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.
