Research map / Label-efficient and robust learning / Domain adaptation and generalization
Improving Unsupervised Domain Adaptation: A Pseudo-Candidate Set Approach
Aveen Dayal, Rishabh Lalla, Linga Reddy Cenkeramaddi, C. Krishna Mohan, Abhinav Kumar, Vineeth N Balasubramanian
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 |
|---|---|---|---|
| apply UDPCS refinement to existing UDA methods on VisDA | MCC, NWD, MDD, LeCo, SDAT · VisDA | mean accuracy: 83.8 -> 86.2 (MCC) | Improved |
| apply UDPCS refinement on DomainNet | CDAN, MDD, MCC, SDAT · DomainNet | average accuracy: 52.2 -> 55.7 (MDD) | Improved |
| apply UDPCS refinement to SDAT on DomainNet | SDAT · DomainNet | average accuracy: 51.3 -> 51.4 | No clear effect |
| apply UDPCS refinement on OfficeHome | MDD, MCC, LeCo, SDAT, NWD · OfficeHome | average accuracy: 69.5 -> 71.3 (MDD) | Improved |
| add candidate loss (ablation) | MDD+Ours · OfficeHome | average accuracy: 69.5 -> 70.7 | Improved |
| add non-candidate loss (ablation) | MDD+Ours · OfficeHome | average accuracy: 69.5 -> 70.6 | Improved |
| add confidence threshold gamma to both losses (ablation) | MDD+Ours · OfficeHome | average accuracy: 70.9 -> 71.3 | Improved |
| use strong augmentation instead of weak augmentation in candidate loss (ablation) | MDD+Ours · OfficeHome | average accuracy: 69.6 -> 71.3 | Improved |
| use strong augmentation instead of weak augmentation in candidate loss (ablation) | MCC+Ours · OfficeHome | average accuracy: 72.8 -> 73.6 | Improved |
| vary lambda between 0.5, 1 and 2 (ablation) | MDD+Ours · OfficeHome | average accuracy: 71.1 -> 71.2 | No clear effect |
| vary threshold gamma between 0.8, 0.85 and 0.9 (ablation) | MDD+Ours · OfficeHome | average accuracy: 71.1 / 71.3 / 71.1 | No clear effect |
| replace KL divergence with cross entropy in candidate loss (ablation) | MDD+Ours · OfficeHome | average accuracy: 71.3 -> 71.3 | No clear effect |
Closest papers by meaning
- Guiding Pseudo-Labels With Uncertainty Estimation for Source-Free Unsupervised Domain Adaptation (CVPR 2023 · similarity 0.94)
- Prior Knowledge Guided Unsupervised Domain Adaptation (ECCV 2022 · similarity 0.93)
- Can We Evaluate Domain Adaptation Models Without Target-Domain Labels? (ICLR 2024 · similarity 0.92)
- Make the U in UDA Matter: Invariant Consistency Learning for Unsupervised Domain Adaptation (NeurIPS 2023 · similarity 0.91)
- Universal Semi-Supervised Domain Adaptation by Mitigating Common-Class Bias (CVPR 2024 · similarity 0.90)
- Pseudo-Labeling for Unsupervised Domain Adaptation with Kernel GLMs (NeurIPS 2026 · 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.
