Research map / Learning theory and algorithms / Kernels and regression theory
Bridging the Gap Between Practice and PAC-Bayes Theory in Few-Shot Meta-Learning
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 |
|---|---|---|---|
| PACMAML (first-order PAC-Bayes, subsampled inner set) vs MAML | ANIL miniImageNet · miniImageNet 1-shot, m_i=80 | accuracy: 48.1 -> 50.1 | Helped |
| train base-learner on a subsample S' instead of all data S (PACMAML vs PACOH) (ablation) | meta-learner · miniImageNet 1-shot, m_i=10 | accuracy: 31.2 -> 47.8 | Helped |
| increase observed-task examples m_i far beyond target m (PACOH/Reptile) | Sinusoid regression · Sinusoid, m=5 | RMSE | Mixed |
| increase observed-task examples m_i (PACMAML/MAML) | Sinusoid regression · Sinusoid, m=5 | RMSE | Helped |
| increase m_i from 10 to 80 | PACMAML · miniImageNet 1-shot ANIL | accuracy: 47.8 -> 50.1 | Helped |
| PACMAML vs first-order FOMAML | ANIL miniImageNet · m_i=10 | accuracy: 41.8 -> 47.8 | Helped |
| PACMAML vs MAML on BERT NLI | BERT-base · 12 NLI tasks, k=4 | avg test accuracy: 48.21 -> 51.58 | Helped |
| PACMAML (first-order) instead of MAML high-order gradients | BERT-base, v=6 · TPU memory | memory: 120G -> 33G | Helped |
Closest papers by meaning
- More Flexible PAC-Bayesian Meta-Learning by Learning Learning Algorithms (ICML 2024 · similarity 0.92)
- PACOH: Bayes-Optimal Meta-Learning with PAC-Guarantees (ICML 2021 · similarity 0.91)
- Recursive PAC-Bayes: A Frequentist Approach to Sequential Prior Updates with No Information Loss (NeurIPS 2024 · similarity 0.89)
- Fast Rate Bounds for Multi-Task and Meta-Learning with Different Sample Sizes (NeurIPS 2025 · similarity 0.88)
- Learning via Surrogate PAC-Bayes (NeurIPS 2024 · similarity 0.87)
- A Hierarchical Bayesian Model for Few-Shot Meta Learning (ICLR 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.