Research map / Continual learning / Online and replay-based
Meta-Learning Representations for Continual Learning
Khurram Javed, Martha White
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 |
|---|---|---|---|
| replace Pre-training representation with OML representation for online SGD (ablation) | Online SGD · Split-Omniglot 50 tasks | accuracy (%): 21.16 -> 64.72 | Improved |
| learn from OML representations instead of standard network | MER · Split-Omniglot 50 tasks | accuracy (%): 54.88 -> 76.00 | Improved |
| learn from OML representations instead of standard network | ER-Reservoir · Split-Omniglot 20 tasks | accuracy (%): 24.32 -> 60.92 | Improved |
| learn from OML representations instead of standard network | EWC · Split-Omniglot 50 tasks | accuracy (%): 05.08 -> 64.44 | Improved |
| use fixed Pre-training representation instead of standard network | MER · Split-Omniglot 20 tasks | accuracy (%): 29.02 -> 42.05 | Improved |
| use fixed Pre-training representation instead of standard network | Approx IID · Split-Omniglot 20 tasks | accuracy (%): 48.02 -> 46.02 | No clear effect |
| OML with plain online SGD versus other continual methods on standard networks | OML + Online SGD · Split-Omniglot 50 tasks | accuracy (%): 64.72 vs 54.88 (standard MER) | Improved |
| OML objective versus MAML-Rep (adds online-aware interference minimization) (ablation) | representation meta-learning · Split-Omniglot online | accuracy | Improved |
| learn a representation encoder instead of only a model initialization (ablation) | OML · Incremental sine waves | training error at meta-test | Improved |
| OML versus Pre-training representation instance sparsity | representation · Omniglot training set | instance sparsity (% active): 38% -> 3.8% | Improved |
Closest papers by meaning
- When Meta-Learning Meets Online and Continual Learning: A Survey (TPAMI 2024 · similarity 0.88)
- Learning to Continually Learn with the Bayesian Principle (ICML 2024 · similarity 0.87)
- Look-ahead Meta Learning for Continual Learning (NeurIPS 2020 · similarity 0.87)
- Learning from the Past: Continual Meta-Learning with Bayesian Graph Neural Networks (AAAI 2020 · similarity 0.86)
- Trust Region Continual Learning as an Implicit Meta-Learner (NeurIPS 2026 · similarity 0.85)
- Enhancing Visual Continual Learning with Language-Guided Supervision (CVPR 2024 · similarity 0.85)
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.
