Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Bayesian Meta-Learning for the Few-Shot Setting via Deep 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 spectral mixture kernel instead of RBF (ablation) | DKT · Periodic function regression in-range | MSE: 1.38 -> 0.08 | Helped |
| use spectral mixture kernel instead of RBF (ablation) | DKT · Periodic function regression out-of-range | MSE: 2.61 -> 0.10 | Helped |
| add BatchNorm feature centering to cosine kernel (ablation) | DKT · mini-ImageNet 1-shot | accuracy (%): 48.64 -> 49.73 | Helped |
| add BatchNorm feature centering to cosine kernel (ablation) | DKT · CUB 1-shot | accuracy (%): 63.37 -> 62.96 | No clear effect |
| add BatchNorm feature centering to cosine kernel (ablation) | DKT with ResNet-10 · CUB 1-shot | accuracy (%): 70.81 -> 72.27 | Helped |
| transfer kernel across tasks vs training deep kernel from scratch per task (ablation) | DKT · Periodic function regression | MSE | Helped |
| use cosine kernel instead of RBF kernel (ablation) | DKT · CUB 1-shot | accuracy (%): 55.34 -> 63.37 | Helped |
| use cosine kernel instead of polynomial p=2 kernel (ablation) | DKT · mini-ImageNet 5-shot | accuracy (%): 60.26 -> 64.00 | Helped |
| DKT vs best competitor | DKT · CUB 5-shot Conv-4 | accuracy (%): 78.51 -> 77.76 | Hurt |
| DKT vs MAML and feature transfer | DKT · Periodic function regression out-of-range | MSE: 8.45 -> 0.10 | Helped |
| DKT vs best baseline on mini-ImageNet 5-shot | DKT · mini-ImageNet 5-shot | accuracy (%): 66.92 -> 64.00 | Hurt |
Closest papers by meaning
- Meta-Transfer Learning for Few-Shot Learning (CVPR 2019 · similarity 0.91)
- Bridging the Gap Between Practice and PAC-Bayes Theory in Few-Shot Meta-Learning (NeurIPS 2021 · similarity 0.89)
- A Unified Few-Shot Classification Benchmark to Compare Transfer and Meta Learning Approaches (NeurIPS 2021 · similarity 0.89)
- MetaDiff: Meta-Learning with Conditional Diffusion for Few-Shot Learning (AAAI 2024 · similarity 0.88)
- A Hierarchical Bayesian Model for Few-Shot Meta Learning (ICLR 2024 · similarity 0.87)
- Pushing the Limit of Fine-Tuning for Few-Shot Learning: Where Feature Reusing Meets Cross-Scale Attention (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.