Research map / Label-efficient and robust learning / Domain adaptation and generalization
Feature-Space Planes Searcher: A Universal Domain Adaptation Framework for Interpretability and Computational Efficiency
Zhitong Cheng, Yiran Jiang, Yulong Ge, Yufeng Li, Zhongheng Qin, Rongzhi Lin, Jianwei Ma
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 entropy core (SE+CE) to SCE baseline (ablation) | FPS · Office-Home Clipart->Product (ViT-B/16) | Macro-F1: 85.1 -> 88.4 | Improved |
| add entropy core (SE+CE) to SCE baseline (ablation) | FPS · VisDA-2017 (ResNet-101) | Macro-F1: 65.5 -> 82.8 | Improved |
| add sample entropy loss alone to SCE baseline (ablation) | FPS · Office-Home Clipart->Product (ViT-B/16) | accuracy: 86.2 -> 78.3 | Hurt |
| add category entropy loss alone to SCE baseline (ablation) | FPS · Office-Home Clipart->Product (ViT-B/16) | accuracy: 86.2 -> 87.6 | Improved |
| add consistency regularizer L_CR to entropy core (ablation) | FPS · Office-Home Clipart->Product (ViT-B/16) | accuracy: 90.1 -> 90.2 | No clear effect |
| add consistency regularizer L_CR to entropy core (ablation) | FPS · VisDA-2017 (ResNet-101) | Macro-F1: 82.8 -> 84.3 | Improved |
| add margin-aware confidence weighting to core + CR (ablation) | FPS · Office-Home Clipart->Product (ViT-B/16) | accuracy: 90.2 -> 91.8 | Improved |
| add entropy core under degraded features (alpha,beta=1.5) (ablation) | FPS · VisDA-2017 degraded (ResNet-101) | Macro-F1: 27.0 -> 51.1 | Improved |
| fine-tune full network instead of head only (ablation) | Frozen head-only · Office-Home Art->Product (ResNet-50) | target accuracy (%): 79.97 -> 77.37 | Hurt |
| fine-tune last block instead of head only (ablation) | Frozen head-only · Office-Home Art->Product (ResNet-50) | target accuracy (%): 79.97 -> 79.32 | Hurt |
| single-step full-dataset optimization with global statistics removed (FPS*) vs batch-wise head-only (ablation) | FPS* · Office-Home Art->Product (ResNet-50) | target accuracy (%): 79.97 -> 80.15 | No clear effect |
| apply FPS to ViT features on EuroSAT RGB | FPS · EuroSAT_MS RGB | accuracy (%): 87.23 -> 90.78 | Improved |
Closest papers by meaning
- Structure-Aware Feature Fusion for Unsupervised Domain Adaptation (AAAI 2020 · similarity 0.88)
- Cluster Alignment With a Teacher for Unsupervised Domain Adaptation (ICCV 2019 · similarity 0.88)
- Maximum Structural Generation Discrepancy for Unsupervised Domain Adaptation (TPAMI 2022 · similarity 0.88)
- Progressive Feature Alignment for Unsupervised Domain Adaptation (CVPR 2019 · similarity 0.87)
- Source-Free Domain Adaptation via Target Prediction Distribution Searching (IJCV 2023 · similarity 0.87)
- Joint Domain Alignment and Discriminative Feature Learning for Unsupervised Deep Domain Adaptation (AAAI 2019 · similarity 0.87)
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.
