Research map / Label-efficient and robust learning / Domain adaptation and generalization
Flatness-Aware Minimization for Domain Generalization
Xingxuan Zhang, Renzhe Xu, Han Yu, Yancheng Dong, Pengfei Tian, Peng Cui
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 |
|---|---|---|---|
| use AdamW instead of Adam as the default DG optimizer (ablation) | ERM (ResNet-50) · Average over 6 DG datasets | accuracy: 65.6 -> 66.0 | Improved |
| use SGD instead of Adam (ablation) | ERM (ResNet-50) · PACS | accuracy: 84.2 -> 79.9 | Hurt |
| use SGD instead of Adam (ablation) | ERM (ResNet-50) · VLCS | accuracy: 77.3 -> 78.1 | Improved |
| optimize zeroth- and first-order flatness jointly (FAD) instead of Adam | ERM (ResNet-50) · Average over 6 DG datasets | accuracy: 65.6 -> 67.6 | Improved |
| use FAD instead of GAM (first-order flatness only) (ablation) | ERM (ResNet-50) · Average over 6 DG datasets | accuracy: 66.6 -> 67.6 | Improved |
| use FAD instead of Adam with SWAD | SWAD · PACS | accuracy: 86.8 -> 88.5 | Improved |
| use FAD instead of Adam with SWAD | SWAD · Average over 5 DG datasets | accuracy: 65.3 -> 66.5 | Improved |
| use FAD instead of Adam with Fishr | Fishr · Average over 5 DG datasets | accuracy: 64.5 -> 65.8 | Improved |
| use FAD instead of Adam with Fishr | Fishr · DomainNet | accuracy: 44.7 -> 43.8 | Hurt |
| train longer (15,000 vs 5,000 iterations) with AdamW (ablation) | AdamW · Average of PACS, VLCS, OfficeHome, NICO++ | accuracy: 76.7 -> 76.4 | No clear effect |
| train FAD longer (15,000 vs 5,000 iterations) (ablation) | FAD · Average of PACS, VLCS, OfficeHome, NICO++ | accuracy: 78.7 -> 78.9 | Mixed |
| train with FAD for only 5,000 iterations versus other optimizers at 15,000 | FAD · PACS and NICO++ | accuracy | Improved |
Closest papers by meaning
- Seeking Consistent Flat Minima for Better Domain Generalization via Refining Loss Landscapes (CVPR 2025 · similarity 0.93)
- Unknown Domain Inconsistency Minimization for Domain Generalization (ICLR 2024 · similarity 0.92)
- Gradient-Guided Annealing for Domain Generalization (CVPR 2025 · similarity 0.90)
- One-Step Generalization Ratio Guided Optimization for Domain Generalization (ICML 2025 · similarity 0.90)
- Sharpness-Aware Model-Agnostic Long-Tailed Domain Generalization (AAAI 2024 · similarity 0.90)
- Set Valued Predictions For Robust Domain Generalization (ICML 2025 · 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.
