Research map / Learning theory and algorithms / Kernels and regression theory
Meta-Learning with Neural Tangent Kernels
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 Meta-RKHS-I instead of Reptile | few-shot meta-learning · Mini-ImageNet 5-way 1-shot | accuracy: 49.70 ± 1.83 % -> 51.10 ± 1.82 % | Helped |
| Use Meta-RKHS-I instead of Reptile | few-shot meta-learning · Mini-ImageNet 5-way 5-shot | accuracy: 65.91 ± 0.84 % -> 66.19 ± 0.80 % | Helped |
| Use Meta-RKHS-II instead of Reptile | few-shot meta-learning · FC-100 5-way 1-shot | accuracy: 38.40 ± 1.94 % -> 41.20 ± 2.17% | Helped |
| Use Meta-RKHS-I instead of Reptile | few-shot meta-learning · FC-100 5-way 5-shot | accuracy: 50.50 ± 0.87 % -> 51.47 ± 0.86% | Helped |
| Use Meta-RKHS-II instead of Bayesian TAML | out-of-distribution few-shot adaptation · CUB 5-way 1-shot, trained on Mini-ImageNet | accuracy: 41.57± 0.60% -> 45.36± 0.87 % | Helped |
| Use Meta-RKHS-II instead of Bayesian TAML | out-of-distribution few-shot adaptation · VGG Flower 5-way 1-shot, trained on Mini-ImageNet | accuracy: 58.56± 0.66 % -> 60.80± 1.02 % | Helped |
| Use Meta-RKHS-II instead of Bayesian TAML | out-of-distribution few-shot adaptation · CUB 5-way 5-shot, trained on Mini-ImageNet | accuracy: 61.78± 0.56% -> 65.21 ± 0.64 % | Helped |
| Use Meta-RKHS-II instead of Bayesian TAML | out-of-distribution few-shot adaptation · VGG Flower 5-way 5-shot, trained on Mini-ImageNet | accuracy: 77.95±0.46 % -> 78.25 ± 0.49 % | No clear effect |
| Set adaptation time to t=10 rather than t=∞ (ablation) | Meta-RKHS-II · Mini-ImageNet 5-way 1-shot | accuracy: 48.70 ± 2.28 % -> 50.53 ± 2.09 | Helped |
| Set adaptation time to t=10 rather than t=∞ (ablation) | Meta-RKHS-II · Mini-ImageNet 5-way 5-shot | accuracy: 64.95± 0.96 % -> 65.40± 0.91 % | No clear effect |
| Set adaptation time to t=10 rather than t=∞ (ablation) | Meta-RKHS-II · FC-100 5-way 1-shot | accuracy: 37.60 ± 2.13 % -> 41.20 ± 2.17 % | Helped |
| Set adaptation time to t=10 rather than t=∞ (ablation) | Meta-RKHS-II · FC-100 5-way 5-shot | accuracy: 49.48 ± 0.98 % -> 51.36 ± 0.96 | Helped |
Closest papers by meaning
- Convergence of Meta-Learning with Task-Specific Adaptation over Partial Parameters (NeurIPS 2020 · similarity 0.90)
- Unraveling Model-Agnostic Meta-Learning via The Adaptation Learning Rate (ICLR 2022 · similarity 0.90)
- Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks (ICML 2017 · similarity 0.90)
- On Fast Adversarial Robustness Adaptation in Model-Agnostic Meta-Learning (ICLR 2021 · similarity 0.89)
- Memory-Reduced Meta-Learning with Guaranteed Convergence (AAAI 2025 · similarity 0.88)
- MetaDiff: Meta-Learning with Conditional Diffusion for Few-Shot 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.