Research map / Continual learning / Online and replay-based
Learning to Remember: A Synaptic Plasticity Driven Framework for Continual Learning
Check 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 |
|---|---|---|---|
| learn connection masks on weights rather than activations (ablation) | DGM · SVHN A10 | accuracy (%): 66.89 -> 74.38 | Helped |
| learn connection masks on weights rather than activations (ablation) | DGM · CIFAR10 A10 | accuracy (%): 51.75 -> 56.21 | Helped |
| learn connection masks on weights rather than activations (ablation) | DGM · MNIST A10 | accuracy (%): 97.92 vs 96.46 | Mixed |
| learn connection masks on weights rather than activations (ablation) | DGM · ImageNet-50 A50 | accuracy (%): 15.16 -> 17.82 | Helped |
| add real samples to the replay loop (ablation) | DGMw · ImageNet-50 top-1 A50 | accuracy (%): 17.82 (r=0) -> 30.48 (r=0.1) | Helped |
| increase the real-sample replay ratio from r=0 to r=0.75 (ablation) | DGMw · ImageNet-50 top-1 A50 | accuracy (%): 17.82 -> 38.22 | Helped |
| increase the real-sample memory budget from K=1000 to K=2000 (ablation) | DGMw · ImageNet-50 top-1 A50 | accuracy (%): 18.84 -> 22.56 | Helped |
| use generated replay without storing or replaying real samples (ablation) | DGMw · ImageNet-50 top-1 A50 | accuracy (%): DGMw (r=0) 17.82; iCarl (K=1000) 28.98 | Hurt |
| learn masks directly on weights instead of activations (ablation) | DGM generator · MNIST, generator network growth | network growth | Helped |
| use the DGMw generator rather than the fixed-size MeRGAN generator | generator on SVHN · SVHN A10 | accuracy (%): 66.78 -> 74.38 | Helped |
Closest papers by meaning
- Recall-Oriented Continual Learning with Generative Adversarial Meta-Model (AAAI 2024 · similarity 0.90)
- Rethinking the Stability-Plasticity Trade-off in Continual Learning from an Architectural Perspective (ICML 2025 · similarity 0.88)
- Task-Free Continual Generation and Representation Learning via Dynamic Expansionable Memory Cluster (AAAI 2024 · similarity 0.87)
- Continual Unsupervised Generative Modelling via Online Optimal Transport (AAAI 2025 · similarity 0.87)
- Self-Normalized Resets for Plasticity in Continual Learning (ICLR 2025 · similarity 0.87)
- Enhancing Visual Continual Learning with Language-Guided Supervision (CVPR 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.