Research map / Label-efficient and robust learning / Noisy labels
Co-teaching: Robust training of deep neural networks with extremely 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 peer-selected small-loss examples to update two networks | CNN training with noisy labels · MNIST Pair-45% | test accuracy: 80.88% -> 87.63% | Helped |
| Use peer-selected small-loss examples to update two networks | CNN training with noisy labels · MNIST Symmetry-50% | test accuracy: 90.05% -> 91.32% | Helped |
| Use peer-selected small-loss examples to update two networks | CNN training with noisy labels · MNIST Symmetry-20% | test accuracy: 98.80% -> 97.25% | Hurt |
| Use peer-selected small-loss examples to update two networks | CNN training with noisy labels · CIFAR-10 Pair-45% | test accuracy: 58.14% -> 72.62% | Helped |
| Use peer-selected small-loss examples to update two networks | CNN training with noisy labels · CIFAR-10 Symmetry-50% | test accuracy: 71.10% -> 74.02% | Helped |
| Use peer-selected small-loss examples to update two networks | CNN training with noisy labels · CIFAR-10 Symmetry-20% | test accuracy: 84.55% -> 82.32% | Hurt |
| Use peer-selected small-loss examples to update two networks | CNN training with noisy labels · CIFAR-100 Pair-45% | test accuracy: 31.60% -> 34.81% | Helped |
| Use peer-selected small-loss examples to update two networks | CNN training with noisy labels · CIFAR-100 Symmetry-50% | test accuracy: 41.04% -> 41.37% | Helped |
| Use peer-selected small-loss examples to update two networks | CNN training with noisy labels · CIFAR-100 Symmetry-20% | test accuracy: 61.87% -> 54.23% | Hurt |
| Set tau to 1.25epsilon rather than epsilon (ablation) | Co-teaching · MNIST Pair-45% | test accuracy: 87.63% -> 97.89% | Helped |
| Set tau to 1.5epsilon rather than epsilon (ablation) | Co-teaching · MNIST Pair-45% | test accuracy: 87.63% -> 69.47% | Hurt |
| Set tau to 1.25epsilon rather than epsilon (ablation) | Co-teaching · MNIST Symmetry-50% | test accuracy: 91.32% -> 98.62% | Helped |
Closest papers by meaning
- How does Disagreement Help Generalization against Label Corruption? (ICML 2019 · similarity 0.92)
- Combating Noisy Labels by Agreement: A Joint Training Method with Co-Regularization (CVPR 2020 · similarity 0.90)
- Understanding and Utilizing Deep Neural Networks Trained with Noisy Labels (ICML 2019 · similarity 0.89)
- Coresets for Robust Training of Deep Neural Networks against Noisy Labels (NeurIPS 2020 · similarity 0.88)
- Learning with Structural Labels for Learning with Noisy Labels (CVPR 2024 · similarity 0.88)
- On the Role of Label Noise in the Feature Learning Process (ICML 2025 · 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.