Research map / Label-efficient and robust learning / Noisy labels
PADDLES: Phase-Amplitude Spectrum Disentangled Early Stopping for Learning with Noisy Labels
Huaxi Huang, Hui Kang, Sheng Liu, Olivier Salvado, Thierry Rakotoarivelo, Dadong Wang, Tongliang Liu
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 PES training of later layers on top of AS/PS-disentangled early stopping (ablation) | PADDLES_Base · CIFAR-100 symmetric 50% | test accuracy (%): 47.72 -> 59.78 | Improved |
| add PES training of later layers (ablation) | PADDLES_Base · CIFAR-10 pairflip 45% | test accuracy (%): 82.80 -> 89.32 | Improved |
| add phase-amplitude disentangled early stopping (without PES) (ablation) | CE baseline · CIFAR-10 symmetric 50% | test accuracy (%): 75.51 -> 83.40 | Improved |
| add PADDLES in place of PES | supervised confident-sample training · CIFAR-10 symmetric 20% | test accuracy (%): 92.38 -> 92.43 | No clear effect |
| use PADDLES instead of PES in semi-supervised setting | semi-supervised noisy label learning · CIFAR-100 symmetric 80% | test accuracy (%): 61.6 -> 62.9 | Improved |
| use PADDLES instead of SOP+ | CIFAR-10N Aggregate · CIFAR-10N Aggregate | test accuracy (%): 95.61 -> 95.46 | Hurt |
| train longer on phase spectrum (large T_P) before stopping (ablation) | PADDLES · CIFAR-N | test accuracy | Hurt |
| increase amplitude-spectrum stopping epoch T_A from 18 to 30 (ablation) | PADDLES · CIFAR-100N Fine and CIFAR-10N Worst | test accuracy | Improved |
| choose disentangle position after 3rd or 4th ResNet block (ablation) | PADDLES · CIFAR-10 and CIFAR-100 | test accuracy | Improved |
| train on phase spectrum rather than amplitude spectrum under label noise (ablation) | ResNet-18 · CIFAR-10 symmetric 50% | robustness to label noise | Improved |
Later papers that compared with this method
1 later accepted paper reports PADDLES as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Variational Rectification Inference for Learning with Noisy Labels (IJCV 2024 · better on 2 of 2 shared results)
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 2 of 2 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
- Amplitude-Phase Recombination: Rethinking Robustness of Convolutional Neural Networks in Frequency Domain (ICCV 2021 · similarity 0.88)
- Early Stopping Against Label Noise Without Validation Data (ICLR 2024 · similarity 0.87)
- Understanding and Mitigating the Label Noise in Pre-training on Downstream Tasks (ICLR 2024 · similarity 0.85)
- On the Role of Label Noise in the Feature Learning Process (ICML 2025 · similarity 0.85)
- Non-Stationary Predictions May Be More Informative: Exploring Pseudo-Labels with a Two-Phase Pattern of Training Dynamics (ICML 2025 · similarity 0.84)
- Self-cognitive Denoising in the Presence of Multiple Noisy Label Sources (ICML 2024 · 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.
