Research map / Label-efficient and robust learning / Noisy labels
Tackling Noisy Labels With Network Parameter Additive Decomposition
Jingyi Wang, Xiaobo Xia, Long Lan, Xinghao Wu, Jun Yu, Wenjing Yang, Bo Han, 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 parameter decomposition alone (no constraints) (ablation) | Standard · CIFAR-10 (Sym. 20%) | test accuracy (%): 90.19 -> 90.09 | No clear effect |
| add constraint on gamma to parameter decomposition (ablation) | PD · CIFAR-10 (Sym. 40%) | test accuracy (%): 86.63 -> 87.68 | Improved |
| add constraint on sigma to PD + gamma constraint (ablation) | PD + gamma · CIFAR-10 (Sym. 40%) | test accuracy (%): 87.68 -> 87.83 | Improved |
| add constraint on gamma to parameter decomposition (ablation) | PD · CIFAR-100 (Sym. 20%) | test accuracy (%): 66.31 -> 68.39 | Improved |
| add constraint on sigma to PD + gamma constraint (ablation) | PD + gamma · CIFAR-100 (Sym. 20%) | test accuracy (%): 68.39 -> 69.46 | Improved |
| add parameter decomposition alone (no constraints) (ablation) | Standard · CIFAR-100 (Sym. 40%) | test accuracy (%): 60.58 -> 60.29 | Hurt |
| use power function for decreasing constraint over constant or other schedules (ablation) | TNLPAD · CIFAR-10 / F-MNIST | test accuracy (%) | Improved |
| use constant constraint function (ablation) | TNLPAD · CIFAR-10 / F-MNIST | test accuracy (%) | No clear effect |
| add TNLPAD to Self | Self · CIFAR-10 (Sym. 40%) | test accuracy (%): 88.67 -> 89.47 | Improved |
| add TNLPAD to DivideMix | DivideMix · CIFAR-10 (Sym. 20%) | test accuracy (%): 96.10 -> 95.99 | Hurt |
| add TNLPAD to DivideMix | DivideMix · CIFAR-10 (Asym. 40%) | test accuracy (%): 92.16 -> 92.54 | Improved |
| use TNLPAD instead of Standard on Clothing1M | Standard · Clothing1M | test accuracy (%): 68.98 -> 71.51 | Improved |
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.
- Self — SELF: Learning to Filter Noisy Labels with Self-Ensembling (ICLR 2020) · better on 1 of 1 shared result
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
- IOFM: Using the Interpolation Technique on the Over-Fitted Models to Identify Clean-Annotated Samples (AAAI 2024 · similarity 0.88)
- To Grok or not to Grok: Disentangling Generalization and Memorization on Corrupted Algorithmic Datasets (ICLR 2024 · similarity 0.87)
- Unveiling Memorization–Generalization Coexistence: A Case Study on Arithmetic Tasks with Label Noise (NeurIPS 2026 · similarity 0.87)
- On the Over-Memorization During Natural, Robust and Catastrophic Overfitting (ICLR 2024 · similarity 0.86)
- Handling Label Noise via Instance-Level Difficulty Modeling and Dynamic Optimization (NeurIPS 2025 · similarity 0.86)
- Memorization in Graph Neural Networks (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.
