Research map / Label-efficient and robust learning / Noisy labels
Learning With Neighbor Consistency for Noisy Labels
Check 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 |
|---|---|---|---|
| Add neighbor consistency regularization | Standard training · mini-ImageNet-Red, 20% noise | accuracy: 66.9\exstd{0.4} -> {\bf 69.0\exstd{0.5}} | Helped |
| Add neighbor consistency regularization | Standard training · mini-ImageNet-Blue, 80% noise | accuracy: 13.4\exstd{0.4} -> 11.5\exstd{0.5} | Hurt |
| Add Mixup to NCR | NCR · mini-ImageNet-Blue, 20% and 40% noise | accuracy: 67.8\exstd{0.3} -> 68.3\exstd{0.1}; 60.6\exstd{0.5} -> 61.4\exstd{0.3} | Mixed |
| Add Mixup to NCR | NCR · mini-ImageNet-Red, 80% noise | accuracy: 51.2\exstd{0.4} -> 48.9\exstd{0.6} | Hurt |
| Add NCR and data augmentation to Mixup | Mixup · mini-WebVision | accuracy: 77.2 -> 80.5 | Helped |
| Add NCR and data augmentation to Mixup | Mixup · Clothing1M | accuracy: 72.2 -> 74.6 | Helped |
| Add NCR and data augmentation to Mixup | Mixup · WebVision | accuracy: 75.5 -> 76.8 | No clear effect |
| Add NCR to ELR | ELR · CIFAR-10 and CIFAR-100 across noise ratios | accuracy | Helped |
| Use a smaller neighborhood, such as k=10 (ablation) | NCR · CIFAR-10 with nonzero synthetic noise | validation accuracy | Helped |
| Set NCR weight alpha high (0.9) (ablation) | NCR · CIFAR-10 with nonzero synthetic noise | validation accuracy | Helped |
| Enable NCR from the start instead of waiting several epochs (ablation) | NCR · CIFAR-10 with 40% and 80% noise | validation accuracy | Helped |
| Increase batch size from 256 to 1024 (ablation) | NCR · WebVision | accuracy: 73.9 -> 75.7 | Helped |
Closest papers by meaning
- Augmentation Strategies for Learning With Noisy Labels (CVPR 2021 · similarity 0.90)
- Tackling Instance-Dependent Label Noise via a Universal Probabilistic Model (AAAI 2021 · similarity 0.90)
- Learning with Structural Labels for Learning with Noisy Labels (CVPR 2024 · similarity 0.89)
- Toward Robustness against Label Noise in Training Deep Discriminative Neural Networks (NeurIPS 2017 · similarity 0.89)
- Directional Label Diffusion Model for Learning from Noisy Labels (CVPR 2025 · similarity 0.89)
- Generative Modeling Reinvents Supervised Learning: Label Repurposing with Predictive Consistency Learning (ICML 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.