Research map / Label-efficient and robust learning / Noisy labels
NoiseRank: Unsupervised Label Noise Reduction with Dependence Models
Karishma Sharma, Pinar Donmez, Enming Luo, Yan Liu, I. Zeki Yalniz
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 |
|---|---|---|---|
| remove top 1.8% ranked noisy instances from machine-labeled pre-training data | ResNet-50 pre-trained on YFCC100m labels · ImageNet-1k | top-1 accuracy (%): 79.06 -> 79.34 | Improved |
| remove NoiseRank-ranked instances instead of the same number of random instances (ablation) | ResNet-50 pre-trained on YFCC100m labels · ImageNet-1k | top-1 accuracy (%): 78.96 -> 79.34 | Improved |
| remove random 1.8% of instances instead of none (ablation) | ResNet-50 pre-trained on YFCC100m labels · ImageNet-1k | top-1 accuracy (%): 79.06 -> 78.96 | Hurt |
| remove more top-ranked instances (0.6% to 1.2%) (ablation) | NoiseRank pre-training filter · ImageNet-1k | top-1 accuracy (%): 79.13 -> 79.12 | No clear effect |
| fine-tune on NoiseRank-cleaned train set instead of noisy train set | ResNet-50 ImageNet pretrain · Clothing-1M classification | top-1 accuracy (%): 68.94 -> 73.77 | Improved |
| initialize from the noisy-train model before fine-tuning on cleaned set (ablation) | NoiseRank · Clothing-1M classification | top-1 accuracy (%): 73.77 -> 73.82 | No clear effect |
| fine-tune on NoiseRank-cleaned train set instead of noisy train set | ResNet-50 ImageNet pretrain · Food-101N classification | top-1 accuracy (%): 81.44 -> 85.20 | Improved |
| initialize from the noisy-train model before fine-tuning on cleaned set (ablation) | NoiseRank · Food-101N classification | top-1 accuracy (%): 85.20 -> 85.78 | Improved |
| add an additional clean 50k set for fine-tuning | ResNet-50 on noisy train · Clothing-1M classification | top-1 accuracy (%): 68.94 -> 79.43 | Improved |
| add the clean 50k set on top of NoiseRank cleaned training | NoiseRank · Clothing-1M classification | top-1 accuracy (%): 73.82 -> 79.57 | Improved |
| iterate NoiseRank (fine-tune on denoised data and re-run detection) (ablation) | NoiseRank · Food-101N label noise detection | average error rate (%): 24.02 -> 18.43 | Improved |
| iterate NoiseRank (fine-tune on denoised data and re-run detection) (ablation) | NoiseRank · Clothing-1M label noise detection | average error rate (%): 23.54 -> 22.81 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- CleanNet — CleanNet: Transfer Learning for Scalable Image Classifier Training With Label Noise (CVPR 2018) · better on 2 of 4 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
- Deep Noise Label Learning via Effective Rank Reduction (ECCV 2026 · similarity 0.90)
- Learning with Feature-Dependent Label Noise: A Progressive Approach (ICLR 2021 · similarity 0.90)
- Label Noise: Ignorance Is Bliss (NeurIPS 2024 · similarity 0.87)
- NoiseGPT: Label Noise Detection and Rectification through Probability Curvature (NeurIPS 2024 · similarity 0.86)
- Adaptive Integration of Partial Label Learning and Negative Learning for Enhanced Noisy Label Learning (AAAI 2024 · similarity 0.86)
- Handling Label Noise via Instance-Level Difficulty Modeling and Dynamic Optimization (NeurIPS 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.
