Research map / Label-efficient and robust learning / Domain adaptation and generalization
Domain Generalization with Small Data
Kecheng Chen, Elena Gal, Hong Yan, Haoliang Li
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 Bayesian layers (probabilistic embeddings) to Unet (ablation) | Unet · Spinal cord GM segmentation (site2) | DSC: 0.7223 -> 0.7934 | Improved |
| add Bayesian classifier on top of Bayesian layers (ablation) | Unet + Bayesian layers · Spinal cord GM segmentation (site2) | DSC: 0.7934 -> 0.8268 | Improved |
| add local alignment (ablation) | Unet + Bayesian layers + Bayesian classifier · Spinal cord GM segmentation (site2) | DSC: 0.8268 -> 0.8364 | Improved |
| add global alignment (P-MMD) (ablation) | Unet + Bayesian layers + Bayesian classifier · Spinal cord GM segmentation (site2) | DSC: 0.8268 -> 0.8371 | Improved |
| add both local and global alignment (ablation) | Unet + Bayesian layers + Bayesian classifier · Spinal cord GM segmentation (site2) | DSC: 0.8364 -> 0.8485 | Improved |
| add probabilistic embeddings (Bayesian layer) to deterministic ResNet18 baseline (ablation) | SWAD baseline · Epithelium-stroma classification (avg) | accuracy (%): 79.62 -> 80.63 | Improved |
| add probabilistic embeddings to deterministic baseline (ablation) | SWAD baseline · Epithelium-stroma classification (NKI) | accuracy (%): 74.84 -> 74.90 | No clear effect |
| replace pairwise/local-only alignment of BDIL with probabilistic framework plus global P-MMD | Ours vs BDIL · Epithelium-stroma classification (NKI) | accuracy (%): 71.89 -> 76.71 | Improved |
| train on 80/60/40% of source samples with the proposed method instead of BDIL | Ours vs BDIL · Skin lesion MSK (reduced data) | accuracy: 0.6528 -> 0.6732 | Improved |
| use proposed method instead of DNA with 40/30/20 samples per class | Ours vs DNA · Skin lesion MSK (few samples per class) | accuracy: 0.5196 -> 0.6183 | Improved |
| replace mean embedding with kernel mean embedding (P-MMD) for alignment (ablation) | Ours · Epithelium-stroma classification (NKI) | accuracy | Improved |
| increase number of Monte Carlo samples T (ablation) | Ours · Epithelium-stroma classification (NKI) | accuracy | Mixed |
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.
- SWAD — SWAD: Domain Generalization by Seeking Flat Minima (NeurIPS 2021) · better on 3 of 3 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
- Bridging Domain Invariance and Diversity: A Fine-Grained Risk Bound for Domain Generalization (JMLR 2026 · similarity 0.89)
- Exploiting Domain-Specific Features to Enhance Domain Generalization (NeurIPS 2021 · similarity 0.89)
- Domain Generalization via Conditional Invariant Representations (AAAI 2018 · similarity 0.88)
- Robust Domain Generalization under Divergent Marginal and Conditional Distributions (NeurIPS 2026 · similarity 0.88)
- Towards Generalizing to Unseen Domains with Few Labels (CVPR 2024 · similarity 0.87)
- Rethinking Multi-domain Generalization with A General Learning Objective (CVPR 2024 · similarity 0.87)
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.
