Research map / Optimization / Optimizers
Probabilistic Model-Agnostic 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 |
|---|---|---|---|
| Use PLATIPUS with KL weight 0.05 instead of deterministic MAML | MAML · Ambiguous CelebA (5-shot) | Coverage (max=3): 1.00 $\pm$ 0.0 -> 1.59 $\pm$ 0.03 | Helped |
| Use PLATIPUS with KL weight 0.05 instead of deterministic MAML | MAML · Ambiguous CelebA (5-shot) | Accuracy: 89.00 $\pm$ 1.78 -> 88.34 $\pm$ 1.06 | No clear effect |
| Use PLATIPUS with KL weight 0.05 instead of deterministic MAML | MAML · Ambiguous CelebA (5-shot) | Average NLL: 0.73 $\pm$ 0.06 -> 0.67$\pm$ 0.05 | Helped |
| Add gradient noise without variational inference | MAML · Ambiguous CelebA (5-shot) | Coverage (max=3): 1.00 $\pm$ 0.0 -> 1.89 $\pm$ 0.04 | Helped |
| Add gradient noise without variational inference | MAML · Ambiguous CelebA (5-shot) | Accuracy: 89.00 $\pm$ 1.78 -> $84.3 \pm 1.60$ | Hurt |
| Add gradient noise without variational inference | MAML · Ambiguous CelebA (5-shot) | Average NLL: 0.73 $\pm$ 0.06 -> $0.68 \pm 0.05$ | Helped |
| Increase the KL weight from 0.05 to 0.15 (ablation) | PLATIPUS · Ambiguous CelebA (5-shot) | Coverage (max=3): 1.59 $\pm$ 0.03 -> \textbf{1.94 $\pm$ 0.04} | Helped |
| Increase the KL weight from 0.05 to 0.15 (ablation) | PLATIPUS · Ambiguous CelebA (5-shot) | Accuracy: 88.34 $\pm$ 1.06 -> 87.8 $\pm$ 1.03 | Hurt |
| Increase the KL weight from 0.05 to 0.15 (ablation) | PLATIPUS · Ambiguous CelebA (5-shot) | Average NLL: 0.67$\pm$ 0.05 -> \textbf{0.56$\pm$ 0.04} | Helped |
| Use PLATIPUS instead of MAML | MAML · MiniImagenet, 5-way 1-shot | Accuracy: 48.70 \pm 1.84\% -> $50.13 \pm 1.86\%$ | Helped |
| Use PLATIPUS instead of Meta-SGD | Meta-SGD · MiniImagenet, 5-way 1-shot | Accuracy: $50.71 \pm 1.87\%$ -> $50.13 \pm 1.86\%$ | Hurt |
| Condition the learned prior on the task training data (ablation) | PLATIPUS · Few-shot learning experiments | Performance | Helped |
Closest papers by meaning
- Bayesian Model-Agnostic Meta-Learning (NeurIPS 2018 · similarity 0.92)
- Task Agnostic Meta-Learning for Few-Shot Learning (CVPR 2019 · similarity 0.92)
- Efficient Meta Learning via Minibatch Proximal Update (NeurIPS 2019 · similarity 0.90)
- Task-Robust Model-Agnostic Meta-Learning (NeurIPS 2020 · similarity 0.90)
- Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks (ICML 2017 · similarity 0.90)
- Probabilistic Active Meta-Learning (NeurIPS 2020 · similarity 0.89)
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.