Research map / Optimization / Optimizers
Efficient Meta Learning via Minibatch Proximal Update
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 |
|---|---|---|---|
| replace MAML inner gradient step with minibatch proximal update (Meta-MinibatchProx) | MAML · miniImageNet 5-way 1-shot | accuracy: 46.21 ± 1.76 -> 48.51 ± 0.92 | Helped |
| same replacement | MAML · miniImageNet 20-way 5-shot | accuracy: 18.34 ± 0.33 -> 33.61 ± 0.41 | Helped |
| same replacement | MAML · tieredImageNet 5-way 5-shot | accuracy: 66.58 ± 1.78 -> 68.30 ± 0.91 | Helped |
| same replacement with transduction | MAML + Transduction · tieredImageNet 5-shot 5-way | accuracy: 70.30 ± 1.75 -> 71.45 ± 0.94 | Helped |
| use first-order approximation (FOMAML) instead of full MAML | MAML · tieredImageNet 5-shot 5-way | accuracy: 66.58 ± 1.78 -> 64.07 ± 1.72 | Hurt |
| use Reptile instead of MAML | MAML · tieredImageNet 5-shot 5-way | accuracy: 66.58 ± 1.78 -> 65.99 ± 0.42 | Hurt |
| use l2,1-norm robust regularizer under 5% outlier images | Meta-MinibatchProx vs MAML · miniImageNet with outliers | accuracy | Helped |
| add transduction | Meta-MinibatchProx · miniImageNet 5-way 1-shot | accuracy: 48.51 ± 0.92 -> 50.77 ± 0.90 | Helped |
Closest papers by meaning
- Unraveling Model-Agnostic Meta-Learning via The Adaptation Learning Rate (ICLR 2022 · similarity 0.90)
- Probabilistic Model-Agnostic Meta-Learning (NeurIPS 2018 · similarity 0.90)
- Meta-Learning with Adaptive Hyperparameters (NeurIPS 2020 · similarity 0.90)
- Memory-Reduced Meta-Learning with Guaranteed Convergence (AAAI 2025 · similarity 0.89)
- Provable Meta-Learning with Low-Rank Adaptations (NeurIPS 2025 · similarity 0.88)
- More Flexible PAC-Bayesian Meta-Learning by Learning Learning Algorithms (ICML 2024 · similarity 0.88)
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.