Research map / Continual learning / Online and replay-based
Continual Learning, Fast and Slow
Quang Pham, Chenghao Liu, Steven C. H. Hoi
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 fast learner on top of slow learner (ablation) | DualNet · Split miniImageNet-TA | ACC: 68.33 -> 73.20 | Improved |
| add fast learner on top of slow learner (ablation) | DualNet · Split miniImageNet-TF | ACC: 27.30 -> 36.86 | Improved |
| use Barlow Twins SSL objective instead of classification loss (ablation) | DualNet · Split miniImageNet-TA (Look-ahead) | ACC: 70.96 -> 73.20 | Improved |
| use Look-ahead optimizer instead of SGD for slow learner (ablation) | DualNet (Barlow Twins) · Split miniImageNet-TA | ACC: 70.99 -> 73.20 | Improved |
| use Look-ahead optimizer instead of SGD for slow learner (ablation) | DualNet++ (Barlow Twins) · Split miniImageNet-TA | ACC: 71.10 -> 74.24 | Improved |
| add spatial dropout (DualNet++) to DualNet | DualNet · Split miniImageNet-TA | ACC: 73.20 -> 74.24 | Improved |
| add spatial dropout (DualNet++) to DualNet | DualNet · Split miniImageNet-TA semi-supervised (rho=10%) | ACC: 54.03 -> 58.03 | Improved |
| add spatial dropout (DualNet++) to DualNet | DualNet · Split miniImageNet-TA semi-supervised (rho=25%) | ACC: 62.80 -> 63.96 | Improved |
| add data augmentation to baseline | ER · Split miniImageNet-TA | ACC: 58.24 -> 59.80 | Improved |
| add DualNet slow-fast learners instead of baseline replay | DER++-Aug · Split miniImageNet-TF | ACC: 28.26 -> 36.86 | Improved |
| add pre-trained ImageNet initialization | DualNet · CTrL S- (task-free) | ACC: 65.5 -> 66.7 | Improved |
| increase SSL iterations n from 1 to 20 between supervised updates | DualNet · Split miniImageNet | ACC | Improved |
Closest papers by meaning
- DualNet: Continual Learning, Fast and Slow (NeurIPS 2021 · similarity 0.98)
- Learning Fast, Learning Slow: A General Continual Learning Method based on Complementary Learning System (ICLR 2022 · similarity 0.88)
- Adversarial Continual Learning (ECCV 2020 · similarity 0.87)
- Linear Mode Connectivity in Multitask and Continual Learning (ICLR 2021 · similarity 0.87)
- Expandable and Differentiable Dual Memories with Orthogonal Regularization for Exemplar-free Continual Learning (AAAI 2026 · similarity 0.87)
- Improving Continual Learning Performance and Efficiency with Auxiliary Classifiers (ICML 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.
