Research map / Label-efficient and robust learning / Noisy labels
Generalized DataWeighting via Class-Level Gradient Manipulation
Can Chen, Shuhao Zheng, Xi Chen, Erqun Dong, Xue Liu, Hao Liu, Dejing Dou
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 class-level gradient weighting (GDW) in place of instance-level weighting | MWNet · CIFAR10 uniform noise 60% | test accuracy (%): 81.14 -> 84.11 | Improved |
| Add class-level gradient weighting (GDW) in place of instance-level weighting | MWNet · CIFAR100 uniform noise 60% | test accuracy (%): 50.21 -> 53.33 | Improved |
| Add class-level gradient weighting (GDW) in place of instance-level weighting | MWNet · CIFAR10 uniform noise 0% | test accuracy (%): 92.95 -> 92.94 | No clear effect |
| Use GDW instead of learned label methods Soft-label/Gen-label | Meta-learning label correction · CIFAR10 uniform noise 40% | test accuracy (%): 86.52 -> 88.14 | Improved |
| Use GDW instead of GLC under 40% flip noise | GLC · CIFAR10 flip noise 40% | test accuracy (%): 89.74 -> 87.70 | Hurt |
| Add GDW on long-tailed data | MWNet · Long-tailed CIFAR10 imbalance 0.01 | test accuracy (%): 62.70 -> 71.31 | Improved |
| Fine-tune the base model on the balanced meta set | BaseModel · Long-tailed CIFAR10 imbalance 0.01 | test accuracy (%): 69.77 -> 59.76 | Hurt |
| Fine-tune the base model on the meta set | BaseModel · CIFAR10 uniform noise 60% | test accuracy (%): 77.92 -> 78.41 | No clear effect |
| Remove the zero-mean constraint on class-level weights (ablation) | GDW · CIFAR10 uniform noise 40% | training stability | Hurt |
| Add GDW on a pretrained ResNet-18 on real-world noisy labels | MWNet · Clothing1M | test accuracy (%): 68.46 -> 69.39 | Improved |
| L2RW with pretrained model instead of training from scratch | L2RW · Clothing1M | test accuracy (%): 68.80 | Mixed |
| Add GDW on a dataset with many classes | MWNet · Places-LT | test accuracy (%): 18.08 -> 19.17 | Improved |
Closest papers by meaning
- CMW-Net: Learning a Class-Aware Sample Weighting Mapping for Robust Deep Learning (TPAMI 2023 · similarity 0.87)
- Gradient-Weight Alignment as a Train-Time Proxy for Generalization in Classification Tasks (NeurIPS 2025 · similarity 0.87)
- Foster Adaptivity and Balance in Learning with Noisy Labels (ECCV 2024 · similarity 0.87)
- Improving Generalization via Meta-Learning on Hard Samples (CVPR 2024 · similarity 0.86)
- Improved Balanced Classification with Theoretically Grounded Loss Functions (NeurIPS 2025 · similarity 0.85)
- Learning Imbalanced Data with Beneficial Label Noise (ICML 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.
