Research map / Continual learning / Online and replay-based
An Efficient Dataset Condensation Plugin and Its Application to Continual Learning
Enneng Yang, Li Shen, Zhenyi Wang, Tongliang Liu, Guibing Guo
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 low-rank plugin at same memory (SM) | DM · CIFAR10 10 img/cls | accuracy: 48.9 -> 59.8 | Improved |
| add low-rank plugin at same memory (SM) | DSA · CIFAR10 10 img/cls | accuracy: 51.1 -> 56.5 | Improved |
| add low-rank plugin at same memory (SM) | DC · CIFAR10 10 img/cls | accuracy: 44.9 -> 50.8 | Improved |
| low-rank plugin with same number of images (SI), 8x less storage (ablation) | DM · CIFAR10 10 img/cls | accuracy: 48.9 -> 46.0 | Mixed |
| increase rank r from 1 to 8 (same number of images) (ablation) | LoDM(SI) · CIFAR10 10 img/cls | accuracy | Improved |
| larger learning rate (1 or 5 vs 0.1) for synthetic data (ablation) | LoDM · CIFAR10 10 img/cls | accuracy | Improved |
| use low-rank plugin to cut iterations | DM · CIFAR10 10 img/cls | accuracy: 2,000 vs 20,000 iterations | Improved |
| add low-rank plugin to MTT | MTT · CIFAR10 1 img/cls | accuracy: 46.3% -> 58.7% | Improved |
| add low-rank plugin to IDC | IDC · CIFAR10 1 img/cls | accuracy: 50.6% -> 57.2% | Improved |
| add low-rank plugin to HaBa | HaBa · CIFAR10 1 img/cls | accuracy: 48.3% -> 66.1% | Improved |
| add low-rank plugin to RememberThePast | RememberThePast · CIFAR10 1 img/cls | accuracy: 66.4% -> 68.4% | Improved |
| replace DM with LoDM as GDumb memory content | GDumb · CIFAR100 5-task class-incremental | accuracy: 33.81% -> 37.92% | Improved |
Closest papers by meaning
- Dataset Condensation with Gradient Matching (ICLR 2021 · similarity 0.90)
- Dataset Condensation via Efficient Synthetic-Data Parameterization (ICML 2022 · similarity 0.89)
- Elucidating the Design Space of Dataset Condensation (NeurIPS 2024 · similarity 0.88)
- M3D: Dataset Condensation by Minimizing Maximum Mean Discrepancy (AAAI 2024 · similarity 0.88)
- Fetch and Forge: Efficient Dataset Condensation for Object Detection (NeurIPS 2024 · similarity 0.86)
- Multisize Dataset Condensation (ICLR 2024 · similarity 0.85)
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.
