Research map / Label-efficient and robust learning / Recognition and long-tail
PAC Prediction Sets for Meta-Learning
Sangdon Park, Edgar Dobriban, Insup Lee, Osbert Bastani
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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-PS (meta PAC prediction set) instead of naive pooled PS | prototypical network score function · mini-ImageNet 5-way, epsilon 0.1 | PAC guarantee on prediction set error | Improved |
| use Meta-PS instead of PS-Test with 20 test-distribution shots per class | prototypical network score function · mini-ImageNet | prediction set size | Improved |
| use Meta-PS instead of PAC variant of meta conformal prediction | mini-ImageNet and FewRel · error over random test tasks | prediction set error control | Improved |
| use Meta-PS instead of PS and Meta-CP | prototypical network modified for imbalanced labels · CDC Heart Dataset, epsilon 0.1 | prediction set error | Improved |
| use Meta-PS instead of PS-Test | prototypical network · CIFAR10-C 10-way, epsilon 0.2 | set size and labeling cost | Improved |
| set delta to 0.1 instead of 1e-5 (ablation) | Meta-PS · mini-ImageNet per-calibration-set error | prediction set error per calibration set | Mixed |
Closest papers by meaning
- PAC Prediction Sets Under Label Shift (ICLR 2024 · similarity 0.89)
- PAC Prediction Sets Under Covariate Shift (ICLR 2022 · similarity 0.87)
- More Flexible PAC-Bayesian Meta-Learning by Learning Learning Algorithms (ICML 2024 · similarity 0.86)
- Post-hoc Uncertainty Learning Using a Dirichlet Meta-Model (AAAI 2023 · similarity 0.86)
- Recursive PAC-Bayes: A Frequentist Approach to Sequential Prior Updates with No Information Loss (NeurIPS 2024 · similarity 0.83)
- Learning via Surrogate PAC-Bayes (NeurIPS 2024 · similarity 0.83)
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.
