Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
PAC-Bayes Meta-Learning With Implicit Task-Specific Posteriors
Cuong Nguyen, Thanh-Toan Do, Gustavo Carneiro
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 SImPa implicit posterior instead of ABML Gaussian posterior | SImPa vs ABML · Omniglot 5-way 1-shot | accuracy: 97.281 -> 98.352 | Improved |
| use SImPa instead of PLATIPUS | SImPa vs PLATIPUS · mini-ImageNet 5-way 1-shot (4-block CNN) | accuracy: 50.13 -> 51.72 | Improved |
| use SImPa instead of ABML | SImPa vs ABML · mini-ImageNet 5-way 1-shot (4-block CNN) | accuracy: 45.00 -> 51.72 | Improved |
| use SImPa instead of Prototypical nets (trained 30-way) | SImPa vs Prototypical nets · mini-ImageNet 5-way 5-shot (4-block CNN) | accuracy: 68.20 -> 63.49 | Hurt |
| use SImPa with extracted features instead of LGM-Net | SImPa vs LGM-Net · mini-ImageNet 5-way 1-shot (non-standard) | accuracy: 69.13 -> 62.85 | Hurt |
| use SImPa with extracted features instead of LEO | SImPa vs LEO · tiered-ImageNet 5-way 1-shot | accuracy: 66.33 -> 70.26 | Improved |
| use SImPa instead of MetaOptNet | SImPa vs MetaOptNet · tiered-ImageNet 5-way 5-shot | accuracy: 81.75 -> 80.15 | Hurt |
| use SImPa instead of LEO | SImPa vs LEO · mini-ImageNet 5-way 5-shot (non-standard) | accuracy: 77.59 -> 77.65 | No clear effect |
| use SImPa instead of ABML/PLATIPUS/BMAML/MAML for calibration | SImPa · mini-ImageNet 1-shot (4-block CNN) | ECE and MCE | Improved |
| use SImPa instead of MAML, PLATIPUS, ABML | SImPa · multi-modal regression | MSE | Improved |
| scale MAML to ResNet10 (about 5 million parameters) | MAML · mini-ImageNet | accuracy: 89% train -> 42% test | Hurt |
Closest papers by meaning
- Bridging the Gap Between Practice and PAC-Bayes Theory in Few-Shot Meta-Learning (NeurIPS 2021 · similarity 0.92)
- Scalable PAC-Bayesian Meta-Learning via the PAC-Optimal Hyper-Posterior: From Theory to Practice (JMLR 2023 · similarity 0.91)
- More Flexible PAC-Bayesian Meta-Learning by Learning Learning Algorithms (ICML 2024 · similarity 0.91)
- PACOH: Bayes-Optimal Meta-Learning with PAC-Guarantees (ICML 2021 · similarity 0.90)
- A Unified View on PAC-Bayes Bounds for Meta-Learning (ICML 2022 · similarity 0.90)
- Recursive PAC-Bayes: A Frequentist Approach to Sequential Prior Updates with No Information Loss (NeurIPS 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.
