Research map / Label-efficient and robust learning / Domain adaptation and generalization
Episodic Training for Domain Generalization
Da Li, Jianshu Zhang, Yongxin Yang, Cong Liu, Yi-Zhe Song, Timothy M. Hospedales
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 episodic feature-extractor training (Epi-F) (ablation) | AGG · PACS (AlexNet) | accuracy (%): 1.6% boost over AGG | Improved |
| add episodic classifier training (Epi-C) on top of Epi-F (ablation) | Epi-F · PACS (AlexNet) | accuracy (%): further improves 0.5% | Improved |
| combine all episodic components (Epi-FCR) (ablation) | AGG · PACS (AlexNet) | accuracy (%): 3.3% improvement over vanilla AGG | Improved |
| add Epi-FCR episodic training | AGG · IXMAS | accuracy (%): 90.6 -> 93.0 | Improved |
| add Epi-FCR episodic training | AGG · VLCS | accuracy (%): 71.2 -> 72.9 | Improved |
| add Epi-FCR episodic training | AGG · PACS (ResNet-18) | accuracy (%): 79.1 -> 81.5 | Improved |
| add Epi-R random-classifier episodic training | AGG (concat) · VD-DG | accuracy (%): 38.7 -> 39.7 | Improved |
| add Epi-R random-classifier episodic training (VD score) | AGG (concat) · VD-DG | VD score: 265 -> 304 | Improved |
| accumulate BN statistics from source domains instead of freezing ImageNet statistics | ResNet-18 · PACS | accuracy (%): 79.1 -> 75.7 | Hurt |
| train episodic with ensemble of n+1 AGG models instead (compare) (ablation) | Epi-FCR · PACS (ResNet-18) | accuracy (%): 80.4 vs 81.5 | Improved |
| globally share first two blocks (S2B) (ablation) | Epi-FCR · PACS (ResNet-18) | accuracy (%): 79.1 -> 80.9 | No clear effect |
| split modules at intermediate layers (first/third block) (ablation) | Epi-FCR · PACS (ResNet-18) | accuracy (%): 81.5 -> 80.6 | Hurt |
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.
- MetaReg — MetaReg: Towards Domain Generalization using Meta-Regularization (NeurIPS 2018) · better on 4 of 4 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
- Is Large-scale Pretraining the Secret to Good Domain Generalization? (ICLR 2025 · similarity 0.90)
- PracticalDG: Perturbation Distillation on Vision-Language Models for Hybrid Domain Generalization (CVPR 2024 · similarity 0.89)
- Domain Generalization in CLIP via Learning with Diverse Text Prompts (CVPR 2025 · similarity 0.89)
- Multimodal Unsupervised Domain Generalization by Retrieving Across the Modality Gap (ICLR 2025 · similarity 0.89)
- START: A Generalized State Space Model with Saliency-Driven Token-Aware Transformation (NeurIPS 2024 · similarity 0.88)
- CLIPCEIL: Domain Generalization through CLIP via Channel rEfinement and Image-text aLignment (NeurIPS 2024 · 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.
