Research map / Continual learning / Online and replay-based
Online Fast Adaptation and Knowledge Accumulation (OSAKA): a New Approach to Continual 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 |
|---|---|---|---|
| add pretraining (ablation) | Continual-MAML · Synbols alpha=0.98 | online cumulative accuracy (total): 74.4 -> 78.4 | Helped |
| add pretraining (ablation) | Continual-MAML · Tiered-ImageNet alpha=0.98 | online cumulative accuracy (total): 61.4 -> 59.1 | Hurt |
| add update modulation (UM) (ablation) | Continual-MAML + pre. · Synbols alpha=0.98 | online cumulative accuracy (total): 78.4 -> 74.8 | Hurt |
| add update modulation (UM) (ablation) | Continual-MAML + pre. · Tiered-ImageNet alpha=0.98 | online cumulative accuracy (total): 59.1 -> 66.7 | Helped |
| add update modulation (UM) (ablation) | Continual-MAML + pre. · Tiered-ImageNet alpha=0.90 | online cumulative accuracy (total): 57.8 -> 59.7 | Helped |
| add prolonged adaptation phase (PAP) (ablation) | Continual-MAML + pre. + UM · Synbols alpha=0.98 | online cumulative accuracy (total): 74.8 -> 86.3 | Helped |
| add prolonged adaptation phase (PAP) (ablation) | Continual-MAML + pre. + UM · Tiered-ImageNet alpha=0.98 | online cumulative accuracy (total): 66.7 -> 69.1 | Helped |
| add prolonged adaptation phase (PAP) (ablation) | Continual-MAML + pre. + UM · Tiered-ImageNet alpha=0.90 | online cumulative accuracy (total): 59.7 -> 53.4 | Hurt |
| use Continual-MAML (dynamic representation) instead of static-representation MAML | C-MAML + pre. + UM + PAP · Synbols font classification alpha=0.98 | font classification accuracy: 37.9 -> 72.0 | Helped |
| remove BGD (MetaBGD -> C-MAML) | MetaBGD · Synbols alpha=0.98 | online cumulative accuracy (total): 72.5 -> 74.4 | Helped |
| use Continual-MAML instead of MAML in low non-stationarity | C-MAML + pre. + UM + PAP · Tiered-ImageNet alpha=0.90 | online cumulative accuracy (total): 60.4 -> 53.4 | Hurt |
Closest papers by meaning
- Learning to Continually Learn via Meta-learning Agentic Memory Designs (NeurIPS 2026 · similarity 0.90)
- Learning to Continually Learn with the Bayesian Principle (ICML 2024 · similarity 0.89)
- A Unified and General Framework for Continual Learning (ICLR 2024 · similarity 0.89)
- Learn more, but bother less: parameter efficient continual learning (NeurIPS 2024 · similarity 0.88)
- Mitigating Catastrophic Forgetting in Online Continual Learning by Modeling Previous Task Interrelations via Pareto Optimization (ICML 2024 · similarity 0.88)
- STAR: Stability-Inducing Weight Perturbation for Continual Learning (ICLR 2025 · 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.