Research map / Label-efficient and robust learning / Domain adaptation and generalization
Reducing the Covariate Shift by Mirror Samples in Cross Domain Alignment
Yin Zhao, Minquan Wang, Longjun Cai
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 mirror loss at both backbone and FC feature layers (ablation) | UDA model · Office-Home | accuracy: 66.5 -> 73.4 | Improved |
| Add mirror loss at both backbone and FC feature layers (ablation) | UDA model · Office-31 | accuracy: 85.5 -> 91.7 | Improved |
| Add mirror loss at backbone feature layer f only (ablation) | UDA model · Office-Home | accuracy: 66.5 -> 71.8 | Improved |
| Add mirror loss at FC feature layer g only (ablation) | UDA model · Office-Home | accuracy: 66.5 -> 71.7 | Improved |
| Add mirror loss at backbone feature layer f only (ablation) | UDA model · Office-31 | accuracy: 85.5 -> 90.0 | Improved |
| Add mirror loss at FC feature layer g only (ablation) | UDA model · Office-31 | accuracy: 85.5 -> 89.7 | Improved |
| Use k=3 instead of k=1 for mirror-neighbor count (ablation) | Mirror alignment · Office-Home | accuracy: 71.7 -> 73.4 | Improved |
| Increase mirror-neighbor count from k=3 to k=7 (ablation) | Mirror alignment · Office-Home | accuracy: 73.4 -> 71.3 | Hurt |
| Increase mirror-neighbor count from k=3 to k=9 (ablation) | Mirror alignment · Office-Home | accuracy: 73.4 -> 71.4 | Hurt |
| Use k=3 instead of k=1 for mirror-neighbor count (ablation) | Mirror alignment · Office-31 | accuracy: 90.2 -> 91.7 | Improved |
| Increase mirror-neighbor count from k=3 to k=7 (ablation) | Mirror alignment · Office-31 | accuracy: 91.7 -> 89.8 | Hurt |
| Vary distance metric and mirror-sample weighting from Euclidean with 1/k weights (ablation) | Mirror alignment · Office-Home | accuracy: 72.7 to 72.3 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- FixBi — FixBi: Bridging Domain Spaces for Unsupervised Domain Adaptation (CVPR 2021) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Maximum Structural Generation Discrepancy for Unsupervised Domain Adaptation (TPAMI 2022 · similarity 0.88)
- Joint Geometrical and Statistical Alignment for Visual Domain Adaptation (CVPR 2017 · similarity 0.87)
- OSAN: A One-Stage Alignment Network To Unify Multimodal Alignment and Unsupervised Domain Adaptation (CVPR 2023 · similarity 0.87)
- Bi-Directional Generation for Unsupervised Domain Adaptation (AAAI 2020 · similarity 0.87)
- RoDA: Robust Domain Alignment for Cross-Domain Retrieval Against Label Noise (AAAI 2025 · similarity 0.86)
- Invertible Projection and Conditional Alignment for Multi-Source Blended-Target Domain Adaptation (AAAI 2025 · similarity 0.85)
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.
