Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Meta-Learning without Memorization
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 weight meta-regularization to MAML | MAML · Sinusoid regression, 5-shot | test MSE: 0.46 (0.04) -> 0.16 (0.04) | Helped |
| Add weight meta-regularization to CNP | CNP · Sinusoid regression, 5-shot | test MSE: 0.91 (0.10) -> 0.11 (0.02) | Helped |
| Add weight meta-regularization to MAML | MAML · Sinusoid regression, 10-shot | test MSE: 0.13 (0.01) -> 0.06 (0.01) | Helped |
| Add weight meta-regularization to CNP | CNP · Sinusoid regression, 10-shot | test MSE: 0.92 (0.05) -> 0.09 (0.01) | Helped |
| Replace weight regularization with activation regularization (ablation) | MR-MAML · Pose prediction | meta-test MSE | Mixed |
| Add weight meta-regularization | MAML · Pose prediction | meta-test MSE: 5.39 (1.31) -> 2.26 (0.09) | Helped |
| Add weight meta-regularization | CNP · Pose prediction | meta-test MSE: 8.48 (0.12) -> 2.89 (0.18) | Helped |
| Add weight decay to all CNP weights (ablation) | CNP · Pose prediction | meta-test MSE: 8.48 (0.12) -> 6.86 (0.27) | Helped |
| Add Bayes-by-Backprop regularization to all CNP weights (ablation) | CNP · Pose prediction | meta-test MSE: 8.48 (0.12) -> 7.73 (0.82) | Helped |
| Use weight meta-regularization instead of standard CNP regularization | CNP · Pose prediction | meta-test MSE: 8.48 (0.12) -> 2.89 (0.18) | Helped |
| Add weight meta-regularization to MAML | MAML · NME Omniglot, 20-way 1-shot | meta-test accuracy: 7.8 (0.2)% -> 83.3 (0.8)% | Helped |
| Add weight meta-regularization to MAML | MAML · NME Omniglot, 20-way 5-shot | meta-test accuracy: 50.7 (22.9)% -> 94.1 (0.1)% | Helped |
Closest papers by meaning
- Task Agnostic Meta-Learning for Few-Shot Learning (CVPR 2019 · similarity 0.91)
- Bridging Multi-Task Learning and Meta-Learning: Towards Efficient Training and Effective Adaptation (ICML 2021 · similarity 0.91)
- Unsupervised Learning via Meta-Learning (ICLR 2019 · similarity 0.90)
- What Matters for Meta-Learning Vision Regression Tasks? (CVPR 2022 · similarity 0.90)
- Unsupervised Meta-Learning via In-Context Learning (ICLR 2025 · similarity 0.90)
- Zero-Shot Task Adaptation with Relevant Feature Information (AAAI 2024 · similarity 0.90)
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.