Research map / Label-efficient and robust learning / Domain adaptation and generalization
Unsupervised Multi-Source Domain Adaptation Without Access to Source Data
Check 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 |
|---|---|---|---|
| combine source models using learned weights and jointly adapt their feature encoders without source data | DECISION · Digits | accuracy: 85.8 -> 94.0; average increase of 8.2% | Helped |
| combine source models using learned weights and jointly adapt their feature encoders without source data | DECISION · Office | accuracy: 89.3 -> 91.1; average increase of 5.2% | Helped |
| combine source models using learned weights and jointly adapt their feature encoders without source data | DECISION · Office-Home | accuracy: 74.3 -> 75.5; mean increase of 2% | Helped |
| combine source models using learned weights and jointly adapt their feature encoders without source data | DECISION · Office-Caltech | accuracy: 97.0 -> 98.0 | Helped |
| combine source models using learned weights and jointly adapt their feature encoders without source data | DECISION · DomainNet | accuracy: 46.2 -> 45.9 | Mixed |
| add the diversity loss to entropy minimization (ablation) | DECISION · Office, D,W → A | accuracy: 68.5 -> 71.6 | Helped |
| add the diversity loss to entropy minimization (ablation) | DECISION · Office, A,D → W | accuracy: 96.6 -> 95.9 | Hurt |
| add pseudo-label loss to weighted information maximization (ablation) | DECISION · Office, A,D → W | accuracy: 95.9 -> 98.4 | Helped |
| add pseudo-label loss to weighted information maximization (ablation) | DECISION · Office, A,W → D | accuracy: 99.0 -> 99.6 | Helped |
| add pseudo-label loss to weighted information maximization (ablation) | DECISION · Office, D,W → A | accuracy: 71.6 -> 75.4 | Helped |
| replace the full objective with pseudo-label loss alone (ablation) | DECISION · Office (Avg.) | accuracy: 91.1 -> 90.5 | Hurt |
| freeze feature extractors and learn source weights instead of using equal weights (ablation) | DECISION · Office-Home (Avg.) | accuracy: 69.2 -> 70.2 | Helped |
Closest papers by meaning
- Do We Really Need to Access the Source Data? Source Hypothesis Transfer for Unsupervised Domain Adaptation (ICML 2020 · similarity 0.93)
- Distributionally Robust Classification for Multi-source Unsupervised Domain Adaptation (ICLR 2026 · similarity 0.92)
- Can We Evaluate Domain Adaptation Models Without Target-Domain Labels? (ICLR 2024 · similarity 0.92)
- Towards Reliable Model Selection for Unsupervised Domain Adaptation: An Empirical Study and A Certified Baseline (NeurIPS 2024 · similarity 0.89)
- Understanding and Improving Source-free Domain Adaptation from a Theoretical Perspective (CVPR 2024 · similarity 0.89)
- Revisiting Source-Free Domain Adaptation: Insights into Representativeness, Generalization, and Variety (CVPR 2025 · 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.