Research map / Continual learning / Online and replay-based
Learning Fast, Learning Slow: A General Continual Learning Method based on Complementary Learning System
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 |
|---|---|---|---|
| Use CLS-ER's dual semantic memories and replay alignment instead of DER++ | continual learning · S-MNIST, buffer 200 | accuracy: 85.61 ±1.40 -> 89.54 ±0.21 | Helped |
| Use CLS-ER's dual semantic memories and replay alignment instead of DER++ | continual learning · S-CIFAR-10, buffer 200 | accuracy: 64.88 ±1.17 -> 66.19 ±0.75 | Helped |
| Use CLS-ER's dual semantic memories and replay alignment instead of DER++ | continual learning · S-Tiny-ImageNet, buffer 200 | accuracy: 10.96 ±1.17 -> 23.47 ±0.80 | Helped |
| Use CLS-ER's dual semantic memories and replay alignment instead of DER++ | continual learning · R-MNIST, buffer 200 | accuracy: 90.43 ±1.87 -> 92.26 ±0.18 | Helped |
| Use CLS-ER's dual semantic memories and replay alignment instead of DER++ | continual learning · P-MNIST, buffer 200 | accuracy: 83.58 ±0.59 -> 84.63 ±0.40 | Helped |
| Use dual semantic memories instead of one semantic memory (ablation) | CLS-ER · S-Tiny-ImageNet, buffer 200 | accuracy: 17.68 ±1.65 -> 23.47 ±0.80 | Helped |
| Use dual semantic memories instead of one semantic memory (ablation) | CLS-ER · S-CIFAR-10, buffer 200 | accuracy: 61.88 ±2.43 -> 66.19 ±0.75 | Helped |
| Use dual semantic memories instead of one semantic memory (ablation) | CLS-ER · S-MNIST, buffer 200 | accuracy: 88.32 ±0.65 -> 89.54 ±0.21 | Helped |
| Use dual semantic memories instead of one semantic memory (ablation) | CLS-ER · R-MNIST, buffer 200 | accuracy: 92.10 ±1.07 -> 92.26 ±0.18 | Helped |
| Use dual semantic memories instead of one semantic memory (ablation) | CLS-ER · P-MNIST, buffer 200 | accuracy: 83.28 ±0.68 -> 84.63 ±0.40 | Helped |
| Evaluate with the stable model rather than the working model (ablation) | CLS-ER · S-CIFAR-10, buffer 200 | accuracy: 50.09 ±1.48 -> 66.19 ±0.75 | Helped |
| Evaluate with the stable model rather than the plastic model (ablation) | CLS-ER · S-CIFAR-10, buffer 200 | accuracy: 62.68 ±1.94 -> 66.19 ±0.75 | Helped |
Closest papers by meaning
- A Unified and General Framework for Continual Learning (ICLR 2024 · similarity 0.90)
- Do Your Best and Get Enough Rest for Continual Learning (CVPR 2025 · similarity 0.89)
- Predicting the Susceptibility of Examples to Catastrophic Forgetting (ICML 2025 · similarity 0.89)
- Layerwise Proximal Replay: A Proximal Point Method for Online Continual Learning (ICML 2024 · similarity 0.87)
- Understanding the Forgetting of (Replay-based) Continual Learning via Feature Learning: Angle Matters (ICML 2025 · similarity 0.87)
- Mitigating Catastrophic Forgetting in Online Continual Learning by Modeling Previous Task Interrelations via Pareto Optimization (ICML 2024 · similarity 0.87)
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.