Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Bi-Level Meta-Learning for Few-Shot Domain Generalization
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 TSA evaluation to 2LM | 2LM · Meta-Dataset, all domains | average accuracy: 75.7 -> 79.5 | Helped |
| Add TSA evaluation to 2LM | 2LM · Meta-Dataset, out-of-domain | average accuracy: 69.4 -> 77.4 | Helped |
| Add domain-specific subspaces to vanilla multi-domain training (ablation) | MDL · Meta-Dataset, in-domain | average accuracy: 77.05 -> 78.35 | Helped |
| Add domain-specific subspaces to vanilla multi-domain training (ablation) | MDL · Meta-Dataset, out-of-domain | average accuracy: 62.53 -> 67.13 | Helped |
| Add the inter-domain prior subspace (ablation) | 2LM_S · Meta-Dataset, all domains | average accuracy: 74.04 -> 75.73 | Helped |
| Regularize unseen-domain subspace adaptation toward the prior subspace (ablation) | 2LM · Meta-Dataset, out-of-domain | average accuracy | Helped |
| Increase the rank of domain-specific subspaces (ablation) | 2LM · Meta-Dataset | accuracy | Mixed |
| Use 2 lower-layer gradient-update steps (ablation) | 2LM · Meta-Dataset | accuracy | Helped |
| Use prior-subspace regularization and domain-specific projections rather than vanilla multi-domain learning (ablation) | MDL · Meta-Dataset, all domains | average accuracy: 71.38 -> 75.73 | Helped |
Closest papers by meaning
- Open Domain Generalization with Domain-Augmented Meta-Learning (CVPR 2021 · similarity 0.91)
- Learning to Generalize: Meta-Learning for Domain Generalization (AAAI 2018 · similarity 0.89)
- Meta-Learning With Differentiable Convex Optimization (CVPR 2019 · similarity 0.89)
- Advancing Open-Set Domain Generalization Using Evidential Bi-Level Hardest Domain Scheduler (NeurIPS 2024 · similarity 0.87)
- Meta-Exploiting Frequency Prior for Cross-Domain Few-Shot Learning (NeurIPS 2024 · similarity 0.87)
- Pushing the Limit of Fine-Tuning for Few-Shot Learning: Where Feature Reusing Meets Cross-Scale Attention (AAAI 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.