Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
On sensitivity of meta-learning to support data
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 |
|---|---|---|---|
| Select worst-case rather than randomly sampled support examples for adaptation | MetaOptNet-SVM · CIFAR-FS 1-shot | accuracy: 70.79 ± 0.69% -> 5.27 ± 2.82% | Hurt |
| Select best-case rather than randomly sampled support examples for adaptation | MetaOptNet-SVM · CIFAR-FS 1-shot | accuracy: 70.79 ± 0.69% -> 87.65 ± 5.76% | Helped |
| Select worst-case rather than randomly sampled support examples for adaptation | MetaOptNet-SVM · FC100 1-shot | accuracy: 38.19 ± 0.48% -> 5.29 ± 1.57% | Hurt |
| Select best-case rather than randomly sampled support examples for adaptation | MetaOptNet-SVM · FC100 1-shot | accuracy: 38.19 ± 0.48% -> 60.14 ± 6.12% | Helped |
| Select worst-case rather than randomly sampled support examples for adaptation | MetaOptNet-SVM · miniImageNet 1-shot | accuracy: 59.03 ± 0.62% -> 4.82 ± 3.03% | Hurt |
| Select best-case rather than randomly sampled support examples for adaptation | MetaOptNet-SVM · miniImageNet 1-shot | accuracy: 59.03 ± 0.62% -> 80.38 ± 5.40% | Helped |
| Train using worst-case support examples instead of random examples (ablation) | R2D2 · CIFAR-FS train tasks | worst-case accuracy: 13.83 ± 9.35 % -> 46.34 ± 14.93% | Helped |
| Train using worst-case support examples instead of random examples (ablation) | R2D2 · CIFAR-FS test tasks | worst-case accuracy: 6.14 ± 2.89% -> 6.76 ± 2.69% | No clear effect |
| Train using worst-case support examples instead of random examples (ablation) | MetaOptNet-SVM · FC100 train tasks | worst-case accuracy: 18.05 ± 12.64% -> 90.37 ± 12.18% | Helped |
| Train using worst-case support examples instead of random examples (ablation) | MetaOptNet-SVM · FC100 test tasks | worst-case accuracy: 5.29 ± 1.57% -> 6.08 ± 1.78% | No clear effect |
| Increase the number of support examples from 1-shot to 10-shot | Meta-learning algorithms · CIFAR-FS | worst-case accuracy: MetaOptNet-SVM: 5.27 ± 2.82% -> 22.24 ± 10.13% | Helped |
Closest papers by meaning
- Unraveling Meta-Learning: Understanding Feature Representations for Few-Shot Tasks (ICML 2020 · similarity 0.90)
- Meta-Dataset: A Dataset of Datasets for Learning to Learn from Few Examples (ICLR 2020 · similarity 0.90)
- Meta-Album: Multi-domain Meta-Dataset for Few-Shot Image Classification (NeurIPS 2022 · similarity 0.89)
- Data Augmentation for Meta-Learning (ICML 2021 · similarity 0.88)
- Meta-Learning without Memorization (ICLR 2020 · similarity 0.88)
- Provable Meta-Learning with Low-Rank Adaptations (NeurIPS 2025 · similarity 0.86)
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.