Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
What Matters for Meta-Learning Vision Regression Tasks?
Check 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 FCL to CNP Max | CNP Max on Distractor · Distractor IC | prediction error (pixel): 5.1122 -> 3.7030 | Helped |
| Add FCL to CNP Max | CNP Max on Distractor · Distractor CC | prediction error (pixel): 6.1730 -> 4.6095 | Helped |
| Add data augmentation (ablation) | CNP Max + FCL on Distractor · Distractor IC | prediction error (pixel): 3.7030 -> 2.0039 | Helped |
| Add task augmentation (ablation) | CNP Max on Distractor · Distractor IC | prediction error (pixel): 5.1122 -> 6.1847 | Hurt |
| Add task augmentation (ablation) | CNP CA on Pascal1D · Pascal1D | MSE: 4.6557 (0.7445) -> 1.3638 (0.2459) | Helped |
| Add task augmentation and data augmentation (ablation) | CNP CA on ShapeNet1D · ShapeNet1D IC | prediction error (°): 8.1941 (0.3026) -> 5.8051 (0.2258) | Helped |
| Add domain randomization, data augmentation, and task augmentation (ablation) | CNP CA on ShapeNet2D · ShapeNet2D IC | prediction error: 38.33 (0.33) -> 14.12 (0.14) | Helped |
| Add meta-regularization to domain randomization and data augmentation (ablation) | CNP CA on ShapeNet2D · ShapeNet2D IC | prediction error: 14.26 (0.09) -> 27.89 (0.61) | Hurt |
| Remove CropAndPad from data augmentation (ablation) | CNP CA + DR on ShapeNet2D · ShapeNet2D test | prediction error: 0.1410 -> 0.1368 | Helped |
| Remove Affine from data augmentation (ablation) | CNP CA + DR on ShapeNet2D · ShapeNet2D test | prediction error: 0.1410 -> 0.1743 | Hurt |
| Use context and target sets as FCL positive pairs (ablation) | CNP Max + DA on Distractor · Distractor IC | prediction error (pixel): 2.3036 (0.0374) -> 2.0039 (0.0155) | Helped |
| Use cross-attention rather than Max aggregation (ablation) | CNP on ShapeNet1D with task and data augmentation · ShapeNet1D IC | prediction error (°): 7.6582 (0.1767) -> 5.8051 (0.2258) | Helped |
Closest papers by meaning
- Meta-Learning without Memorization (ICLR 2020 · similarity 0.90)
- Provable Meta-Learning of Linear Representations (ICML 2021 · similarity 0.88)
- Meta Omnium: A Benchmark for General-Purpose Learning-To-Learn (CVPR 2023 · similarity 0.88)
- 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 0.87)
- Unsupervised Meta-Learning via In-Context Learning (ICLR 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.