Research map / Label-efficient and robust learning / Noisy labels
How does a Neural Network's Architecture Impact its Robustness to Noisy Labels?
Jingling Li, Mozhi Zhang, Keyulu Xu, John P Dickerson, Jimmy Ba
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 |
|---|---|---|---|
| evaluate predictive power of learned representations instead of raw test accuracy | DivideMix PreAct ResNet18 · CIFAR-10 flipped noise 80% | test accuracy (%): 56.2 -> 93.6 | Improved |
| evaluate predictive power of learned representations instead of raw test accuracy | DivideMix 4-layer MLP · CIFAR-Easy uniform noise 90% | test accuracy (%): 16.22 -> 99.94 | Improved |
| use CNN-based network instead of MLP (better alignment with target) | DivideMix predictive power · CIFAR-10 uniform noise 80% | test accuracy (%): 38.8 -> 93.5 | Improved |
| use MLP instead of CNN (better alignment with target) | DivideMix predictive power · CIFAR-Easy uniform noise 90% | test accuracy (%): 10.09 -> 99.94 | Improved |
| add DivideMix on top of vanilla training | 4-layer MLP · CIFAR-10 uniform noise 80% | test accuracy (%): 32.5 -> 34.4 | No clear effect |
| add DivideMix on top of vanilla training | 9-layer CNN · CIFAR-Easy uniform noise 90% | test accuracy (%): 10.36 -> 10.00 | No clear effect |
| add DivideMix on top of vanilla training | 9-layer CNN · CIFAR-Easy uniform noise 80% | test accuracy (%): 10.15 -> 100.00 | Improved |
| add DivideMix on top of vanilla training | PreAct ResNet18 · CIFAR-10 uniform noise 80% | test accuracy (%): 27.3 -> 92.9 | Improved |
| add DivideMix on top of vanilla training | 4-layer MLP · CIFAR-10 flipped noise 80% | test accuracy (%): 43.0 -> 44.0 | No clear effect |
| use predictive power instead of test accuracy for weakly aligned MLP | DivideMix 4-layer MLP · CIFAR-10 flipped noise 80% | test accuracy (%): 44.0 -> 38.8 | Hurt |
| use predictive power of DivideMix on well-aligned CNN under extreme uniform noise | 9-layer CNN · CIFAR-Easy uniform noise 90% | test accuracy (%): 10.00 -> 10.09 | No clear effect |
| increase mutual information between noisy and clean labels | well-aligned networks · predictive power vs noise type | test accuracy (%) | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DivideMix — DivideMix: Learning with Noisy Labels as Semi-supervised Learning (ICLR 2020) · better on 13 of 13 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Neural Architecture Design and Robustness: A Dataset (ICLR 2023 · 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)
- On the Role of Label Noise in the Feature Learning Process (ICML 2025 · similarity 0.87)
- Learning with Structural Labels for Learning with Noisy Labels (CVPR 2024 · similarity 0.86)
- Soften to Defend: Towards Adversarial Robustness via Self-Guided Label Refinement (CVPR 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.
