Research map / Label-efficient and robust learning / Domain adaptation and generalization
Semi-Supervised Noise Adaptation: Transferring Knowledge from Noise Domain
Yuan Yao, Jin Song, Huixia Li, Tongtong Yuan, Jiaqi Wu, Yu Zhang
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 NAF (noise domain alignment) to ERM | ResNet-18 · CIFAR-10 | top-1 accuracy: 55.55 -> 67.90 | Improved |
| add NAF to ERM | ResNet-50 · CIFAR-10 | top-1 accuracy: 58.83 -> 73.98 | Improved |
| add NAF to ERM | ResNet-18 · ImageNet-1K | accuracy: 36.11 -> 37.10 | Improved |
| add NAF as plug-in to UDA | UDA · CIFAR-10 epoch 20 | accuracy: 56.11 -> 76.94 | Improved |
| add NAF as plug-in to FixMatch | FixMatch · CIFAR-10 epoch 20 | accuracy: 69.40 -> 79.31 | Improved |
| add NAF as plug-in to DST | DST · CIFAR-10 epoch 20 | accuracy: 85.47 -> 85.53 | No clear effect |
| add L_n loss (ablation) | NAF · CIFAR-100 (4 labels/class) | accuracy: 47.33 -> 49.98 | Improved |
| add L_n,t loss (ablation) | NAF · CIFAR-100 (4 labels/class) | accuracy: 40.64 -> 49.98 | Improved |
| use NDS alignment instead of Euclidean distance (EDD) (ablation) | NAF · CIFAR-100 | accuracy: 20.03 -> 49.98 | Improved |
| give noise class-discriminative structure instead of single point (ablation) | NAF · CIFAR-10 | accuracy: 33.34 -> 58.15 | Improved |
| use learned class means instead of fixed class means (ablation) | NAF · CIFAR-100 | accuracy: 46.68 -> 47.72 | Improved |
| raise noise dimensionality from 1024 to 2048 (ablation) | NAF · CIFAR-100 | accuracy: 49.98 -> 51.04 | No clear effect |
Closest papers by meaning
- De-Confusing Pseudo-Labels in Source-Free Domain Adaptation (ECCV 2024 · similarity 0.89)
- Label-Noise Robust Domain Adaptation (ICML 2020 · similarity 0.88)
- Transferable Curriculum for Weakly-Supervised Domain Adaptation (AAAI 2019 · similarity 0.88)
- Semi-Supervised Domain Adaptation With Source Label Adaptation (CVPR 2023 · similarity 0.88)
- Label Noise: Ignorance Is Bliss (NeurIPS 2024 · similarity 0.86)
- Proxy Denoising for Source-Free Domain Adaptation (ICLR 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.
