Research map / Label-efficient and robust learning / Domain adaptation and generalization
Do We Really Need to Access the Source Data? Source Hypothesis Transfer for Unsupervised 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 |
|---|---|---|---|
| adapt target features while keeping the source classifier fixed, using information maximization and self-supervised pseudo-labels | SHOT · Office-Home (closed-set, average) | accuracy: 67.6 to 71.8 | Helped |
| adapt target features while keeping the source classifier fixed, using information maximization and self-supervised pseudo-labels | SHOT · VisDA-C (per-class) | accuracy: 76.4 to 82.9 | Helped |
| add self-supervised pseudo-labeling to SHOT-IM (ablation) | SHOT-IM · Office-Home (closed-set, average) | accuracy: 70.5 to 71.8 | Helped |
| add self-supervised pseudo-labeling to SHOT-IM (ablation) | SHOT-IM · Office (average) | accuracy: 87.3 to 88.6 | Helped |
| add self-supervised pseudo-labeling to SHOT-IM (ablation) | SHOT-IM · VisDA-C (per-class) | accuracy: 80.4 to 82.9 | Helped |
| replace self-supervised pseudo-labels with naive pseudo-labels (ablation) | L_ent + L_div + self-supervised PL · Office-Home (closed-set, average) | accuracy: 71.8 to 70.3 | Hurt |
| replace self-supervised pseudo-labels with naive pseudo-labels (ablation) | L_ent + L_div + self-supervised PL · Office (average) | accuracy: 88.6 to 87.5 | Hurt |
| replace self-supervised pseudo-labels with naive pseudo-labels (ablation) | L_ent + L_div + self-supervised PL · VisDA-C (per-class) | accuracy: 82.9 to 82.9 | No clear effect |
| add the diversity-promoting objective to entropy minimization (ablation) | L_ent · Office-Home (closed-set, average) | accuracy: 55.5 to 70.5 | Helped |
| add the diversity-promoting objective to entropy minimization (ablation) | L_ent · VisDA-C (per-class) | accuracy: 63.3 to 80.4 | Helped |
| add the diversity-promoting objective to entropy minimization (ablation) | L_ent · Office (average) | accuracy: 83.5 to 87.3 | Helped |
| use entropy minimization with self-supervised pseudo-labeling but without the diversity objective (ablation) | L_ent + L_div + self-supervised PL · Office-Home (closed-set, average) | accuracy: 71.8 to 55.5 | Hurt |
Closest papers by meaning
- Revisiting Source-Free Domain Adaptation: Insights into Representativeness, Generalization, and Variety (CVPR 2025 · similarity 0.90)
- Can We Evaluate Domain Adaptation Models Without Target-Domain Labels? (ICLR 2024 · similarity 0.89)
- Understanding and Improving Source-free Domain Adaptation from a Theoretical Perspective (CVPR 2024 · similarity 0.88)
- LEAD: Learning Decomposition for Source-free Universal Domain Adaptation (CVPR 2024 · similarity 0.88)
- A General Representation-Based Approach to Multi-Source Domain Adaptation (ICML 2025 · similarity 0.88)
- Revisiting the Domain Shift and Sample Uncertainty in Multi-source Active Domain Transfer (CVPR 2024 · similarity 0.88)
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.