Research map / Label-efficient and robust learning / Domain adaptation and generalization
Sample Selection for Universal Domain Adaptation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use the proposed score s(x) instead of UAN's wt(x) (ablation) | the proposed UniDA method · Office31 | average class accuracy: 87.84 -> 92.44 | Helped |
| Use the proposed score s(x) instead of entropy score sh(x) (ablation) | the proposed UniDA method · Office31 | average class accuracy: 87.65 -> 92.44 | Helped |
| Add the domain-classifier factor d(x) to the sample score (ablation) | the proposed UniDA method · Office31 | average class accuracy: 81.67 -> 92.44 | Helped |
| Add the classification-confidence factor max ȳ(x) to the sample score (ablation) | the proposed UniDA method · Office31 | average class accuracy: 84.15 -> 92.44 | Helped |
| Use confidence regularization instead of entropy maximization (ablation) | the proposed UniDA method · Office31 | accuracy: 91.70 -> 92.44 | Helped |
| Use selective pseudo-labeling instead of Zou et al.'s pseudo-labeling scheme (ablation) | the proposed UniDA method · Office31 | accuracy: 88.26 -> 92.44 | Helped |
| Add pseudo-labels (ablation) | the proposed UniDA method · Office-Home | average results: 76.689 -> 79.62 | Helped |
| Add pseudo-labels (ablation) | the proposed UniDA method · Office31 | average results: 89.89 -> 92.44 | Helped |
| Use a dynamic pseudo-label threshold instead of no threshold (sα = 0) (ablation) | the proposed UniDA method · Office31 | average results: 88.19 -> 92.44 | Helped |
| Use a dynamic pseudo-label threshold instead of a static sα = 1.2 (ablation) | the proposed UniDA method · Office-Home | average results: 78.46 -> 79.62 | Helped |
| Use a dynamic pseudo-label threshold instead of a static sα = 1.2 (ablation) | the proposed UniDA method · Office31 | average results: 92.48 -> 92.44 | No clear effect |
| Add confidence regularization (ablation) | the proposed UniDA method · Office-Home | average results: 78.04 -> 79.62 | Helped |
Closest papers by meaning
- Gradient-Based Sample Selection for Black-Box Universal Domain Adaptation (AAAI 2025 · similarity 0.91)
- Tackling Dimensional Collapse toward Comprehensive Universal Domain Adaptation (ICML 2025 · similarity 0.90)
- Can We Evaluate Domain Adaptation Models Without Target-Domain Labels? (ICLR 2024 · similarity 0.89)
- Towards Reliable Model Selection for Unsupervised Domain Adaptation: An Empirical Study and A Certified Baseline (NeurIPS 2024 · similarity 0.89)
- Universal Semi-Supervised Domain Adaptation by Mitigating Common-Class Bias (CVPR 2024 · similarity 0.89)
- Category-Aware Active Domain Adaptation (ICML 2024 · similarity 0.89)
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.