Research map / Continual learning / Online and replay-based
GCR: Gradient Coreset Based Replay Buffer Selection 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 |
|---|---|---|---|
| Use GradApprox gradient-matching coreset selection instead of DER's reservoir sampling | DER-style replay on S-Cifar-100 Class-IL offline · S-Cifar-100 Class-IL offline, K=500 | accuracy: 41.36±1.76 -> 45.91±1.3 | Helped |
| Use GCR rather than DER | Online Class-IL on S-Cifar-10 · S-Cifar-10 Class-IL online, K=500 | accuracy: 45.56±6.02 -> 51.17±3.5 | Helped |
| Use GCR rather than DER as tasks increase to 20 | S-Cifar-100 Class-IL offline · S-Cifar-100 Class-IL offline, 20 tasks, K=200 | accuracy: 15.64±2.59 -> 21.02±1.99 | Helped |
| Add GradApprox coreset selection instead of reservoir sampling (ablation) | GCR on S-Cifar-100 Class-IL offline · S-Cifar-100 Class-IL offline, K=2000 | accuracy: 57.41±1.01 -> 60.09±0.72 | Helped |
| Add GradApprox coreset selection instead of bilevel coresets (ablation) | GCR on S-Cifar-100 Class-IL offline · S-Cifar-100 Class-IL offline, K=2000 | accuracy: 51.11±0.96 -> 60.09±0.72 | Helped |
| Add supervised contrastive loss (ablation) | GCR on S-Cifar-100 Class-IL offline · S-Cifar-100 Class-IL offline, K=2000 | accuracy: 58.35±0.78 -> 60.09±0.72 | Helped |
| Add GradApprox coreset selection to ER (ablation) | ER on S-Cifar-100 Class-IL offline · S-Cifar-100 Class-IL offline, K=2000 | accuracy: 42.80±0.49 -> 47.09±1.08 | Helped |
Closest papers by meaning
- Coreset Selection via Reducible Loss in Continual Learning (ICLR 2025 · similarity 0.92)
- Advancing Prompt-Based Methods for Replay-Independent General Continual Learning (ICLR 2025 · similarity 0.91)
- Mitigating Catastrophic Forgetting in Online Continual Learning by Modeling Previous Task Interrelations via Pareto Optimization (ICML 2024 · similarity 0.90)
- Predicting the Susceptibility of Examples to Catastrophic Forgetting (ICML 2025 · similarity 0.90)
- Layerwise Proximal Replay: A Proximal Point Method for Online Continual Learning (ICML 2024 · similarity 0.89)
- A Unified and General Framework for Continual Learning (ICLR 2024 · similarity 0.89)
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.