Research map / Label-efficient and robust learning / Domain adaptation and generalization
Active Universal Domain Adaptation
Xinhong Ma, Junyu Gao, Changsheng Xu
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 adversarial curriculum loss Ladv (instead of naive adversarial loss) (ablation) | AUAN · Office-Home (6 challenging tasks) | avg class accuracy: 56.57 -> 58.91 | Improved |
| add diverse curriculum loss Ldiv (ablation) | AUAN · Office-Home (6 challenging tasks) | avg class accuracy: 49.61 -> 58.91 | Improved |
| add prototype classification loss Lp (ablation) | AUAN · Office-Home (6 challenging tasks) | avg class accuracy: 55.71 -> 58.91 | Improved |
| add self-supervised cluster loss Lnc (ablation) | AUAN · Office-Home (6 challenging tasks) | avg class accuracy: 56.95 -> 58.91 | Improved |
| use labeled target data DLT during adaptation (ablation) | AUAN · Office-Home (6 challenging tasks) | avg class accuracy: 54.70 -> 58.91 | Improved |
| use pseudo-labeled DPLT when learning Gp (ablation) | AUAN · Office-Home (6 challenging tasks) | avg class accuracy: 57.67 -> 58.91 | Improved |
| use ADCL instead of UAN as domain adaptation method | active learning pipeline · Office-Home (avg over 4 AL strategies) | avg class accuracy: 32.95 -> 47.33 | Improved |
| use ADCL instead of UAN | active learning pipeline · Office-31 (avg over 4 AL strategies) | avg class accuracy: 63.13 -> 79.75 | Improved |
| use CNTGE instead of BADGE active learning | ADCL · VisDA S->R (15th round) | avg class accuracy: 64.00 -> 73.91 | Improved |
| use CNTGE instead of BADGE active learning | ADCL · Office-31 (avg, 15th round) | avg class accuracy: 80.51 -> 81.18 | Improved |
| use Coreset instead of CNTGE active learning | ADCL · Office-31 (avg, 15th round) | avg class accuracy: 81.18 vs 80.79 | No clear effect |
| use Entropy instead of Random active learning | ADCL · VisDA S->R (15th round) | avg class accuracy: 63.15 -> 60.06 | Hurt |
Closest papers by meaning
- Category-Aware Active Domain Adaptation (ICML 2024 · similarity 0.92)
- Gradient-Based Sample Selection for Black-Box Universal Domain Adaptation (AAAI 2025 · similarity 0.92)
- Universal Domain Adaptation (CVPR 2019 · similarity 0.91)
- Universal Semi-Supervised Domain Adaptation by Mitigating Common-Class Bias (CVPR 2024 · similarity 0.91)
- ADU: Adaptive Detection of Unknown Categories in Black-Box Domain Adaptation (CVPR 2025 · similarity 0.89)
- Enhancing Semi-supervised Domain Adaptation via Effective Target Labeling (AAAI 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.
