Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
MetaFun: Meta-Learning with Iterative Functional Updates
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 |
|---|---|---|---|
| use iterative functional updates with MetaFun rather than parameter-space MAML | Sinusoid regression · Sinusoid 5-shot | MSE: 0.205±0.013 -> 0.040±0.008 | Helped |
| use MetaFun rather than parameter-space MAML | Sinusoid regression · Sinusoid 10-shot | MSE: 0.059±0.010 -> 0.017±0.005 | Helped |
| use MetaFun-DFP rather than LEO | miniImageNet classification · miniImageNet 5-way 1-shot, without augmentation | accuracy: 61.76±0.08% -> 62.12±0.30% | Helped |
| use MetaFun-DFP rather than MetaOptNet-SVM | miniImageNet classification · miniImageNet 5-way 5-shot, with augmentation | accuracy: 80.00±0.45% -> 80.82±0.17% | Helped |
| use MetaFun-DFP rather than LEO | tieredImageNet classification · tieredImageNet 5-way 1-shot | accuracy: 66.33±0.05% -> 67.72±0.14% | Helped |
| use MetaFun-KFP rather than MetaOptNet-SVM | tieredImageNet classification · tieredImageNet 5-way 5-shot | accuracy: 81.75±0.58% -> 83.28±0.12% | Helped |
| replace the learned neural local update with a gradient-based update (ablation) | MetaFun · miniImageNet 1-shot | accuracy: 62.12±0.30% -> 59.63±0.19% | Hurt |
| replace the learned neural local update with a gradient-based update (ablation) | MetaFun · miniImageNet 5-shot | accuracy: 78.20±0.16% -> 75.84±0.04% | Hurt |
| replace the learned neural local update with a gradient-based update (ablation) | MetaFun · tieredImageNet 1-shot | accuracy: 67.72±0.14% -> 62.55±0.10% | Hurt |
| replace the learned neural local update with a gradient-based update (ablation) | MetaFun · tieredImageNet 5-shot | accuracy: 82.81±0.15% -> 78.18±0.09% | Hurt |
| replace the deep kernel with a squared exponential kernel (ablation) | MetaFun · miniImageNet 1-shot | accuracy: 61.16±0.15% -> 60.04±0.19% | Hurt |
| replace the deep kernel with a squared exponential kernel (ablation) | MetaFun · tieredImageNet 5-shot | accuracy: 83.28±0.12% -> 79.70±0.20% | Hurt |
Closest papers by meaning
- Function Contrastive Learning of Transferable Meta-Representations (ICML 2021 · similarity 0.90)
- MARS: Meta-learning as Score Matching in the Function Space (ICLR 2023 · similarity 0.88)
- MetaDiff: Meta-Learning with Conditional Diffusion for Few-Shot Learning (AAAI 2024 · similarity 0.87)
- Provable Meta-Learning with Low-Rank Adaptations (NeurIPS 2025 · similarity 0.85)
- Improving Memory Efficiency for Training KANs via Meta Learning (ICML 2025 · similarity 0.85)
- Memory-Reduced Meta-Learning with Guaranteed Convergence (AAAI 2025 · similarity 0.85)
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.