Research map / Continual learning / Online and replay-based
Optimizing Reusable Knowledge for Continual Learning via Metalearning
Julio Hurtado, Alain Raymond, Alvaro Soto
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 metalearning KB update to baseline (ablation) | sequential baseline · 20-Split CIFAR-100 | forgetting: -3% v/s -0.25% | Improved |
| add mask-generating functions to Baseline + ML (full MARK) (ablation) | Baseline + ML · 20-Split CIFAR-100 | forgetting (BWT): -3 -> -0.25 | Improved |
| add metalearning to Baseline + Mask (full MARK) (ablation) | Baseline + Mask · 20-Split CIFAR-100 | max average accuracy: 74 -> 78 | Improved |
| add mask functions only to baseline (ablation) | sequential baseline · 20-Split CIFAR-100 | accuracy and forgetting | Mixed |
| add metalearning only to baseline (ablation) | sequential baseline · 20-Split CIFAR-100 | forgetting | Improved |
| retrain each task after sequential learning (ablation) | MARK-Task · 20-Split CIFAR-100 | task accuracy: +1.24% | Improved |
| retrain vs maximum accuracy during training (ablation) | MARK-Task · 20-Split CIFAR-100 | task accuracy: +0.97% | Improved |
| use pretrained ImageNet ResNet-18 features instead of per-task features | MARK · 20-Split CIFAR-100 | accuracy: 78.31 -> 86.29 | Improved |
| use random weights for feature embedding instead of trained (ablation) | MARK · 20-Split MiniImageNet | accuracy: 69.43 -> 33.02 | Hurt |
| MARK-Task vs ACL accuracy | MARK-Task · 20-Split MiniImageNet | accuracy: 62.07 -> 69.43 | Improved |
| MARK-Task vs ACL memory | MARK-Task · 20-Split CIFAR-100 | memory: 134% -> 100% | Improved |
| remove KB mask retraining (No-Retraining variant) (ablation) | MARK · 20-Split MiniImageNet | task accuracy | Hurt |
Closest papers by meaning
- A continual learning survey: Defying forgetting in classification tasks (TPAMI 2021 · similarity 0.88)
- Efficient Continual Learning with Modular Networks and Task-Driven Priors (ICLR 2021 · similarity 0.88)
- Learn more, but bother less: parameter efficient continual learning (NeurIPS 2024 · similarity 0.87)
- Continual Learning with Global Alignment (NeurIPS 2024 · similarity 0.87)
- A Unified and General Framework for Continual Learning (ICLR 2024 · similarity 0.86)
- Separating the 'what' and 'how' of compositional computation to enable reuse and continual learning (NeurIPS 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.
