Research map / Label-efficient and robust learning / Domain adaptation and generalization
Graphical Modeling for Multi-Source Domain Adaptation
Minghao Xu, Hang Wang, Bingbing Ni
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 global alignment loss L_global (ablation) | CRF-MSDA · Digits-five avg | accuracy: 86.90 -> 90.57 | Improved |
| add local alignment loss L_local (ablation) | CRF-MSDA · Digits-five avg | accuracy: 86.90 -> 89.02 | Improved |
| add both L_global and L_local (ablation) | CRF-MSDA · Digits-five avg | accuracy: 86.90 -> 91.83 | Improved |
| add MLE-based loss L_MLE to classification loss (ablation) | MRF-MSDA · Digits-five avg | accuracy: 91.93 -> 93.74 | Improved |
| add classification loss L_cls to MLE loss (ablation) | MRF-MSDA · Digits-five avg | accuracy: 91.11 -> 93.74 | Improved |
| use MRF-MSDA instead of CRF-MSDA | MSDA graphical model · Digits-five avg | accuracy: 91.8 -> 93.7 | Improved |
| use MRF-MSDA instead of CRF-MSDA | MSDA graphical model · PACS avg | accuracy: 89.9 -> 92.6 | Improved |
| use MRF-MSDA instead of CRF-MSDA | MSDA graphical model · Office-31 avg | accuracy: 84.6 -> 84.7 | No clear effect |
| use MRF-MSDA instead of CRF-MSDA | MSDA graphical model · DomainNet avg | accuracy: 47.4 -> 47.9 | No clear effect |
| use MRF-MSDA instead of CRF-MSDA for training speed | MSDA graphical model · Digits-five training time (100 iterations) | training time (s): 17.41 -> 16.54 | Improved |
| raise temperature tau to 0.4 from 0.1 (ablation) | MRF-MSDA · Digits-five ->mm | accuracy: nearly 15% decrease at tau = 0.4 | Hurt |
| make adjacency too dense or sparse (sigma > 0.05 or < 0.0005 vs 0.005) (ablation) | CRF-MSDA · Digits-five | accuracy: best at sigma around 0.005 | Hurt |
Closest papers by meaning
- Partial Feature Selection and Alignment for Multi-Source Domain Adaptation (CVPR 2021 · similarity 0.90)
- STEM: An Approach to Multi-Source Domain Adaptation With Guarantees (ICCV 2021 · similarity 0.90)
- On Balancing Bias and Variance in Unsupervised Multi-Source-Free Domain Adaptation (ICML 2023 · similarity 0.90)
- Learning to Combine: Knowledge Aggregation for Multi-Source Domain Adaptation (ECCV 2020 · similarity 0.89)
- Invertible Projection and Conditional Alignment for Multi-Source Blended-Target Domain Adaptation (AAAI 2025 · similarity 0.88)
- Cycle Self-Refinement for Multi-Source Domain Adaptation (AAAI 2024 · similarity 0.86)
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.
