Research map / Label-efficient and robust learning / Noisy labels
Meta-Learning Dynamic Center Distance: Hard Sample Mining for Learning with Noisy Labels
Chenyu Mu, Yijun Qu, Jiexi Yan, Erkun Yang, Cheng Deng
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 meta-learning pipeline for weights (ablation) | DCD · CIFAR-10 (sym 90%) | test accuracy: 93.82 -> 95.14 | Improved |
| add meta-learning pipeline for weights (ablation) | DCD · CIFAR-100 (sym 90%) | test accuracy: 63.91 -> 65.32 | Improved |
| add Dynamic Center Distance weighting (ablation) | DCD · CIFAR-10 (sym 90%) | test accuracy: 93.03 -> 95.14 | Improved |
| add Dynamic Center Distance weighting (ablation) | DCD · CIFAR-100 (sym 90%) | test accuracy: 63.30 -> 65.32 | Improved |
| add contrastive loss (ablation) | DCD · CIFAR-10 (sym 90%) | test accuracy: 94.23 -> 95.14 | Improved |
| add contrastive loss (ablation) | DCD · CIFAR-100 (sym 90%) | test accuracy: 64.73 -> 65.32 | Improved |
| learn alpha and beta independently instead of alpha+beta=1 (ablation) | DCD · CIFAR-100 (sym 80%) | test accuracy: 72.2 -> 73.3 | Improved |
| include noisy-label weight alpha (vs alpha=0) (ablation) | DCD · CIFAR-100 (sym 80%) | test accuracy: 60.9 -> 73.3 | Improved |
| include pseudo-label weight beta (vs beta=0) (ablation) | DCD · CIFAR-100 (sym 80%) | test accuracy: 64.8 -> 73.3 | Improved |
| use DCD-normalized weight Gamma instead of direct meta weight lambda (ablation) | DCD · CIFAR-10 (sym 20%) | test accuracy: 94.82 -> 96.78 | Improved |
| use DCD-normalized weight Gamma instead of direct meta weight lambda (ablation) | DCD · CIFAR-10 (sym 60%) | test accuracy: 91.94 -> 95.61 | Improved |
| increase meta data size from 2% to 6% (ablation) | DCD · CIFAR-10 (sym 90%) | test accuracy: 95.1% -> 95.7% | No clear effect |
Closest papers by meaning
- Dynamic Loss for Robust Learning (TPAMI 2023 · similarity 0.87)
- Enhanced Sample Selection with Confidence Tracking: Identifying Correctly Labeled Yet Hard-to-Learn Samples in Noisy Data (AAAI 2025 · similarity 0.86)
- Sample Selection via Contrastive Fragmentation for Noisy Label Regression (NeurIPS 2024 · similarity 0.85)
- Improving Generalization via Meta-Learning on Hard Samples (CVPR 2024 · similarity 0.84)
- CGMatch: A Different Perspective of Semi-supervised Learning (CVPR 2025 · similarity 0.84)
- Learning from Noisy Labels via Self-Taught On-the-Fly Meta Loss Rescaling (AAAI 2025 · similarity 0.84)
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.
