Research map / Label-efficient and robust learning / Domain adaptation and generalization
Provable Domain Generalization via Invariant-Feature Subspace Recovery
Haoxiang Wang, Haozhe Si, Bo Li, Han Zhao
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 ISR-Mean post-processing to ERM features | ResNet-50 ERM · Waterbirds | worst-group accuracy: 62.93±5.37 -> 76.10±1.11 | Improved |
| Apply ISR-Cov post-processing to ERM features | ResNet-50 ERM · Waterbirds | worst-group accuracy: 62.93±5.37 -> 82.46±0.55 | Improved |
| Apply ISR-Mean post-processing to GroupDRO features | ResNet-50 GroupDRO · Waterbirds | worst-group accuracy: 90.79±0.47 -> 90.42±0.61 | No clear effect |
| Apply ISR-Cov post-processing to GroupDRO features | ResNet-50 GroupDRO · Waterbirds | worst-group accuracy: 90.79±0.47 -> 91.00±0.45 | Improved |
| Apply ISR-Mean post-processing to ERM features | ResNet-50 ERM · CelebA | worst-group accuracy: 46.39±2.42 -> 55.39±6.13 | Improved |
| Apply ISR-Cov post-processing to ERM features | ResNet-50 ERM · CelebA | worst-group accuracy: 46.39±2.42 -> 79.73±5.00 | Improved |
| Apply ISR-Mean post-processing to reweighting features | ResNet-50 Reweighting · CelebA | worst-group accuracy: 84.44±1.66 -> 90.08±0.50 | Improved |
| Apply ISR-Mean post-processing to GroupDRO features | ResNet-50 GroupDRO · CelebA | worst-group accuracy: 88.22±1.67 -> 90.95±0.32 | Improved |
| Apply ISR-Mean post-processing to ERM features | BERT ERM · MultiNLI | worst-group accuracy: 65.95±1.65 -> 72.60±1.09 | Improved |
| Apply ISR-Cov post-processing to ERM features | BERT ERM · MultiNLI | worst-group accuracy: 65.95±1.65 -> 74.21±2.55 | Improved |
| Apply ISR-Mean post-processing to GroupDRO features | BERT GroupDRO · MultiNLI | worst-group accuracy: 78.43±0.87 -> 78.95±0.95 | No clear effect |
| Apply ISR-Mean to ERM-trained features with only 10% environment labels | Waterbirds ERM · Waterbirds | worst-group accuracy: 73.4%; 10.5% | Improved |
Closest papers by meaning
- Domain Generalization via Conditional Invariant Representations (AAAI 2018 · similarity 0.89)
- A Simple Feature Augmentation for Domain Generalization (ICCV 2021 · similarity 0.89)
- Robust Domain Generalization under Divergent Marginal and Conditional Distributions (NeurIPS 2026 · similarity 0.88)
- Adaptive Methods for Real-World Domain Generalization (CVPR 2021 · similarity 0.88)
- Bridging Domain Invariance and Diversity: A Fine-Grained Risk Bound for Domain Generalization (JMLR 2026 · similarity 0.88)
- Geometric Rate–Distortion Invariance for Domain Generalization (ICML 2026 · 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.
