Research map / Bandits and online learning / Online convex optimization and games
Memory Efficient Online 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 |
|---|---|---|---|
| replace stored task history (FTML) with fixed-size state vector (MOML) | online meta learning on S-MNIST · S-MNIST | CTM test accuracy: 85.82 vs 80.62 | Helped |
| replace stored task history (FTML) with fixed-size state vector (MOML) | online meta learning on 5 way-CIFAR-100 · 5 way-CIFAR-100 | LTM test accuracy: 60.78 vs 54.50 | Helped |
| replace stored task history with fixed-size state vector on miniImageNet | online meta learning on 5 way-miniImageNet · 5 way-miniImageNet | CTM test accuracy: 56.23 vs 56.77 | No clear effect |
| add a 10-task FIFO buffer to MOML (ablation) | MOML on 5 way-CIFAR-100 · 5 way-CIFAR-100 | LTM test accuracy: 60.78 -> 65.24 | Helped |
| add a 10-task buffer to MOML (ablation) | MOML on 5 way-CIFAR-100 · 5 way-CIFAR-100 | CTM test accuracy: 55.83 -> 56.70 (about 1 point) | Helped |
| add a 10-task buffer to MOML (ablation) | MOML on S-MNIST · S-MNIST | LTM test accuracy: 87.49 -> 87.77 | Helped |
| add a 10-task buffer to MOML (ablation) | MOML on S-MNIST · S-MNIST | CTM test accuracy: 85.82 -> 84.33 (lower) | No clear effect |
| use latest-B buffer instead of random-B buffer (ablation) | B-MOML · 5 way-CIFAR-100 and S-MNIST | LTM test accuracy: 65.24 vs 64.37 on CIFAR-100 (B=10); 87.77 vs 86.97 on S-MNIST (B=10) | Mixed |
| sub-linear regret / memory-free MOML vs FTML memory growth | MOML · theory | regret / memory | Helped |
Closest papers by meaning
- Online Meta-Learning (ICML 2019 · similarity 0.90)
- Online Meta-Learning via Learning with Layer-Distributed Memory (NeurIPS 2021 · similarity 0.89)
- Memory-Reduced Meta-Learning with Guaranteed Convergence (AAAI 2025 · similarity 0.89)
- Online Constrained Meta-Learning: Provable Guarantees for Generalization (NeurIPS 2023 · similarity 0.89)
- Online-Within-Online Meta-Learning (NeurIPS 2019 · similarity 0.89)
- Adaptive Gradient-Based Meta-Learning Methods (NeurIPS 2019 · 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.