Research map / Label-efficient and robust learning / Noisy labels
Understanding and Improving Early Stopping for Learning with Noisy Labels
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 traditional early stopping with progressive early stopping (ablation) | ResNet-18 CE base model · CIFAR-10 Sym-50% | test accuracy (%): 70.76 -> 75.87 | Helped |
| replace traditional early stopping with PES (ablation) | ResNet-18 · CIFAR-10 Pair-45% | test accuracy (%): 60.62 -> 62.40 | Helped |
| replace traditional early stopping with PES (ablation) | ResNet-18 · CIFAR-10 Inst-40% | label precision: 90.77 -> 90.63 | No clear effect |
| use PES instead of early stopping for confident example selection (ablation) | ResNet-18 · CIFAR-10 Sym-20% | label recall: 88.51 -> 92.67 | Helped |
| use PES base model with MixMatch instead of DivideMix | PreAct ResNet-18 · CIFAR-100 Pair-45% | test accuracy (%): 48.2 -> 73.6 | Helped |
| combine PES with semi-supervised learning vs ELR+ | PreAct ResNet-18 · CIFAR-100 Sym-80% | test accuracy (%): 50.4 -> 61.6 | Helped |
| PES single network vs best ensemble baseline | ResNet-50 · Clothing-1M | test accuracy (%): 74.81 -> 74.64 | No clear effect |
| vary T2 from 0 to 10 (ablation) | PES · CIFAR-10 | test accuracy | Mixed |
| add progressive training time over plain training | PES · CIFAR-10 Sym-50% | training time (h): 0.9 -> 1.0 | Hurt |
Closest papers by meaning
- Early Stopping Against Label Noise Without Validation Data (ICLR 2024 · similarity 0.91)
- On the Role of Label Noise in the Feature Learning Process (ICML 2025 · similarity 0.88)
- Directional Label Diffusion Model for Learning from Noisy Labels (CVPR 2025 · similarity 0.87)
- Learning with Structural Labels for Learning with Noisy Labels (CVPR 2024 · similarity 0.87)
- How Does Label Noise Gradient Descent Improve Generalization in the Low SNR Regime? (NeurIPS 2025 · similarity 0.86)
- IOFM: Using the Interpolation Technique on the Over-Fitted Models to Identify Clean-Annotated Samples (AAAI 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.