Research map / Architectures and efficiency / RNNs and normalization
Making Scalable Meta Learning Practical
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 |
|---|---|---|---|
| add algorithmic adaptation for Adam (SAMA-NA to SAMA), reweighting only (ablation) | SAMA · WRENCH TREC | accuracy: 74.33 -> 85.73 | Helped |
| add algorithmic adaptation (SAMA-NA to SAMA), reweighting only (ablation) | SAMA · WRENCH ChemProt | accuracy: 60.88 -> 76.89 | Helped |
| add algorithmic adaptation (SAMA-NA to SAMA), reweighting only (ablation) | SAMA · WRENCH Yelp | accuracy: 80.96 -> 93.64 | Helped |
| add label correction on top of reweighting (ablation) | SAMA · WRENCH IMDB | accuracy: 84.31 -> 85.71 | Mixed |
| add label correction on top of reweighting (ablation) | SAMA · WRENCH SemEval | accuracy: 89.67 -> 88.83 | Hurt |
| replace Neumann series with SAMA for meta gradient | meta learning algorithm · AGNews, 1 GPU | memory (MB): 26.0 -> 14.3 | Helped |
| replace conjugate gradient with SAMA | meta learning algorithm · AGNews, 1 GPU | throughput (samples/sec): 82.1 -> 142.0 | Helped |
| scale SAMA from 1 to 4 GPUs with efficient DDP communication (ablation) | SAMA · AGNews | throughput (samples/sec): 142.0 -> 396.7 | Helped |
| add algorithmic adaptation cost (SAMA-NA to SAMA) (ablation) | SAMA · AGNews, 1 GPU | throughput (samples/sec): 144.1 -> 142.0 | No clear effect |
| identity approximation of base Jacobian (DARTS-style) vs iterative differentiation (ablation) | meta gradient · AGNews ablation | memory: 22.94 -> 10.81 | Helped |
| add algorithmic adaptation to SAMA-NA (ablation) | SAMA · AGNews ablation | accuracy: 86.55 -> 89.05 | Helped |
| add algorithmic adaptation to SAMA-NA (ablation) | SAMA · IMDB ablation | accuracy: 81.92 -> 84.31 | Helped |
Closest papers by meaning
- Continuous-Time Meta-Learning with Forward Mode Differentiation (ICLR 2022 · similarity 0.89)
- Dynamic Kernel Selection for Improved Generalization and Memory Efficiency in Meta-Learning (CVPR 2022 · similarity 0.88)
- Scalable Bayesian Meta-Learning through Generalized Implicit Gradients (AAAI 2023 · similarity 0.88)
- Memory-Reduced Meta-Learning with Guaranteed Convergence (AAAI 2025 · similarity 0.87)
- Improving Generalization via Meta-Learning on Hard Samples (CVPR 2024 · similarity 0.87)
- MetaOptimize: A Framework for Optimizing Step Sizes and Other Meta-parameters (ICML 2025 · similarity 0.86)
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.