Research map / Probabilistic and scientific modeling / Molecules and proteins
Meta-Learning Probabilistic Inference for Prediction
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 Versa instead of Non-Amortized VI | few-shot classification · Omniglot 5-way 1-shot | NLL: 0.144 ± 0.005 -> 0.010 ± 0.005 | Helped |
| Use Versa instead of Non-Amortized VI | few-shot classification · Omniglot 20-way 5-shot | NLL: 0.078 ± 0.002 -> 0.031 ± 0.004 | Helped |
| Use Versa instead of Amortized VI | few-shot classification · miniImageNet 5-way 5-shot | NLL: 1.165 ± 0.010 -> 0.859 ± 0.015 | Helped |
| Use Versa instead of Reptile | few-shot classification · miniImageNet 5-way 5-shot | accuracy (%): 65.99 ± 0.58 -> 67.37 ± 0.86 | Helped |
| Use Versa instead of Relation Net | few-shot classification · Omniglot 20-way 1-shot | accuracy (%): 97.64 ± 0.30 -> 97.66 ± 0.29 | No clear effect |
| Use Versa instead of the C-VAE | ShapeNet view reconstruction · ShapeNet 1-shot | MSE: 0.0269 -> 0.0108 | Helped |
| Use Versa instead of the C-VAE | ShapeNet view reconstruction · ShapeNet 1-shot | SSIM: 0.5705 -> 0.7893 | Helped |
| Increase the observed views from 1-shot to 5-shot | Versa · ShapeNet view reconstruction | MSE: 0.0108 -> 0.0069 | Helped |
| Increase the observed views from 1-shot to 5-shot | Versa · ShapeNet view reconstruction | SSIM: 0.7893 -> 0.8483 | Helped |
| Replace test-time gradient optimization with amortized forward-pass inference | Versa vs. MAML · miniImageNet test tasks | speed: 302.9 seconds -> 53.5 seconds | Helped |
Closest papers by meaning
- Probabilistic Model-Agnostic Meta-Learning (NeurIPS 2018 · similarity 0.89)
- Meta-Amortized Variational Inference and Learning (AAAI 2020 · similarity 0.87)
- Bayesian Meta-Learning for the Few-Shot Setting via Deep Kernels (NeurIPS 2020 · similarity 0.87)
- Efficient Meta Learning via Minibatch Proximal Update (NeurIPS 2019 · similarity 0.87)
- Amortized Bayesian Meta-Learning (ICLR 2019 · similarity 0.87)
- More Flexible PAC-Bayesian Meta-Learning by Learning Learning Algorithms (ICML 2024 · similarity 0.85)
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.