Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
How to Distribute Data across Tasks for Meta-Learning?
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 |
|---|---|---|---|
| Allocate the same amount of data to every task | MAML on homogeneous tasks · Mixed linear regression | test loss | Helped |
| Choose the number of data points per task to balance task count against per-task data | MAML with a fixed data budget · Mixed linear regression | test loss | Helped |
| Allocate more data per task to hard tasks when training easy and hard tasks separately | MAML · CIFAR-FS, class hierarchy and noisy labels | accuracy | Helped |
| Allocate more data to easy tasks than to hard tasks | MAML trained jointly on easy and hard tasks · CIFAR-FS, class hierarchy and noisy labels | accuracy | Helped |
| Use 5 adaptation steps during meta-training instead of 1 (ablation) | MAML · Few-shot image classification | performance stability | Helped |
| Increase the number of adaptation steps beyond 5 (ablation) | MAML · Few-shot image classification | performance | No clear effect |
| Increase the task mini-batch size further (ablation) | MAML · Few-shot image classification | performance | No clear effect |
Closest papers by meaning
- Meta-Learning without Memorization (ICLR 2020 · similarity 0.89)
- Bridging Multi-Task Learning and Meta-Learning: Towards Efficient Training and Effective Adaptation (ICML 2021 · similarity 0.88)
- Unsupervised Learning via Meta-Learning (ICLR 2019 · similarity 0.88)
- Meta-Learning with Fewer Tasks through Task Interpolation (ICLR 2022 · similarity 0.88)
- Meta-learning for Mixed Linear Regression (ICML 2020 · similarity 0.88)
- Any-Way Meta Learning (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.