Research map / Continual learning / Online and replay-based
Rainbow Memory: Continual Learning With a Memory of Diverse Samples
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 |
|---|---|---|---|
| Add perturbation-uncertainty-based diversity sampling to memory management (ablation) | Rainbow Memory · CIFAR10-Blurry10-Online, K=200 | A5 accuracy: 38.0 $\pm$ 2.2 -> 43.8 $\pm$ 1.2 | Helped |
| Add data augmentation (CutMix+AutoAug) (ablation) | Rainbow Memory · CIFAR10-Blurry10-Online, K=500 | A5 accuracy: 60.8$\pm$0.8 -> 71.1$\pm$0.1 | Helped |
| Add data augmentation (ablation) | Rainbow Memory · MNIST-Blurry10-Online | A5 accuracy: 92.65 $\pm$ 0.33 -> 91.80 $\pm$ 0.69 | Hurt |
| Add data augmentation (ablation) | Rainbow Memory · CIFAR100-Blurry10-Online | A5 accuracy: 34.09 $\pm$ 1.41 -> 41.35 $\pm$ 0.95 | Helped |
| Add data augmentation (ablation) | Rainbow Memory · ImageNet-Blurry10-Online | A10 accuracy: 37.96 -> 50.11 | Helped |
| Add data augmentation (ablation) | Rainbow Memory · CIFAR10-Blurry10-Offline, K=500 | A5 accuracy: 74.58 $\pm$ 0.60 -> 76.86 $\pm$ 0.04 | Helped |
| Use CutMix+AutoAug rather than no augmentation (ablation) | Rainbow Memory · CIFAR10-Blurry10-Online, K=200 | A5 accuracy: 44.4$\pm$1.4 -> 54.6$\pm$1.6 | Helped |
| Use AutoAug rather than no augmentation (ablation) | Rainbow Memory · CIFAR10-Blurry10-Online, K=500 | A5 accuracy: 60.8$\pm$0.8 -> 69.6$\pm$2.9 | Helped |
| Use CutMix rather than no augmentation (ablation) | Rainbow Memory · CIFAR10-Blurry10-Online, K=500 | A5 accuracy: 60.8$\pm$0.8 -> 62.0$\pm$3.5 | Helped |
| Use RandAug rather than no augmentation (ablation) | Rainbow Memory · CIFAR10-Blurry10-Online, K=500 | A5 accuracy: 60.8$\pm$0.8 -> 68.6$\pm$0.7 | Helped |
| Estimate uncertainty using RandAug Monte Carlo perturbations rather than no Monte Carlo (ablation) | Rainbow Memory · CIFAR10-Blurry10-Online, K=500 | A5 accuracy: 58.59 -> 61.27 | Helped |
| Replace uncertainty-based sampling with reservoir sampling (ablation) | Rainbow Memory · CIFAR10-Blurry10-Online, K=1,000 | A5 accuracy: 64.6 $\pm$ 4.2 -> 64.7 $\pm$ 4.1 | Hurt |
Closest papers by meaning
- Gradient Episodic Memory for Continual Learning (NeurIPS 2017 · similarity 0.89)
- Online Continual Learning on a Contaminated Data Stream With Blurry Task Boundaries (CVPR 2022 · similarity 0.89)
- Learning without Prejudices: Continual Unbiased Learning via Benign and Malignant Forgetting (ICLR 2023 · similarity 0.88)
- Task-Free Continual Generation and Representation Learning via Dynamic Expansionable Memory Cluster (AAAI 2024 · similarity 0.87)
- Re-evaluating Continual Learning with Few-Shot Adaptation (NeurIPS 2026 · similarity 0.87)
- Online Task-Free Continual Learning via Dynamic Expansionable Memory Distribution (CVPR 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.