Research map / Label-efficient and robust learning / Noisy labels
Understanding and Utilizing Deep Neural Networks Trained 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 |
|---|---|---|---|
| Use INCV-selected cleaner samples and then Co-teaching instead of standard Co-teaching | Co-teaching · CIFAR-10 symmetric noise 0.2, 0.5, and 0.8; asymmetric noise 0.4 | clean-test accuracy (%): 89.05 -> 89.71; 82.12 -> 84.78; 16.21 -> 52.27; 84.55 -> 86.04 | Helped |
| Use INCV-selected cleaner samples and then Co-teaching instead of standard Co-teaching | Co-teaching · WebVision validation, first 50 classes | Top-1 accuracy (%): 63.58 -> 65.24 | Helped |
| Use INCV-selected cleaner samples and then Co-teaching instead of standard Co-teaching | Co-teaching · WebVision validation, first 50 classes | Top-5 accuracy (%): 85.20 -> 85.34 | Helped |
| Use INCV-selected cleaner samples and then Co-teaching instead of standard Co-teaching | Co-teaching · ImageNet ILSVRC12 validation, trained on WebVision first 50 classes | Top-1 accuracy (%): 61.48 -> 61.60 | Helped |
| Use INCV-selected cleaner samples and then Co-teaching instead of standard Co-teaching | Co-teaching · ImageNet ILSVRC12 validation, trained on WebVision first 50 classes | Top-5 accuracy (%): 84.70 -> 84.98 | Helped |
| Iterate cross-validation to add further identified samples | INCV · CIFAR-10 with synthetic noisy labels | label recall: about 90% after four iterations at symmetric noise 0.5 | Helped |
| Increase INCV iterations | INCV · CIFAR-10 with synthetic noisy labels | label precision | Hurt |
| Vary the INCV remove ratio slightly | INCV · CIFAR-10 with synthetic noisy labels | test accuracy | No clear effect |
Closest papers by meaning
- Coresets for Robust Training of Deep Neural Networks against Noisy Labels (NeurIPS 2020 · similarity 0.91)
- Distilling Effective Supervision From Severe Label Noise (CVPR 2020 · similarity 0.91)
- Learning with Structural Labels for Learning with Noisy Labels (CVPR 2024 · similarity 0.90)
- Confidence-based Reliable Learning under Dual Noises (NeurIPS 2022 · similarity 0.90)
- On the Role of Label Noise in the Feature Learning Process (ICML 2025 · similarity 0.89)
- How Does Label Noise Gradient Descent Improve Generalization in the Low SNR Regime? (NeurIPS 2025 · 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.