Research map / Continual learning / Online and replay-based
AFEC: Active Forgetting of Negative Transfer in Continual Learning
Liyuan Wang, Mingtian Zhang, Zhongfan Jia, Qian Li, Chenglong Bao, Kaisheng Ma, Jun Zhu, Yi Zhong
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 active forgetting (expanded-parameter penalty) to EWC, i.e. AFEC vs EWC (ablation) | EWC · CIFAR-100-SC | avg accuracy A20: 51.74 -> 55.24 | Improved |
| add active forgetting to EWC (ablation) | EWC · CIFAR-10 regression two-task LeNet | avg accuracy: 39.91 -> 44.45 | Improved |
| add active forgetting to EWC (ablation) | EWC · CIFAR-10 regression transfer LeNet | avg accuracy: 35.87 -> 40.90 | Improved |
| add EWC weight regularization to fine-tuning | fine-tuning · CIFAR-10 regression transfer LeNet | avg accuracy: 38.93 -> 35.87 | Hurt |
| plug AFEC into MAS | MAS · CIFAR-100-SC | avg accuracy A20: 52.18 -> 54.89 | Improved |
| plug AFEC into SI | SI · ImageNet-100 | avg accuracy A10: 72.97 -> 74.14 | Improved |
| plug AFEC into RWALK | RWALK · CIFAR-100-SC | avg accuracy A20: 49.62 -> 51.76 | Improved |
| plug AFEC into MAS | MAS · CUB-200 w/o PT | avg accuracy A10: 42.56 -> 42.93 | Mixed |
| AFEC vs AGS-CL on large-scale images from scratch | AFEC · ImageNet-100 | avg accuracy A10: 47.20 -> 75.46 | Improved |
| AFEC vs AGS-CL on CUB-200 with pretraining | AFEC · CUB-200 w/ PT | avg accuracy A10: 81.84 -> 82.04 | Improved |
Closest papers by meaning
- Do Your Best and Get Enough Rest for Continual Learning (CVPR 2025 · similarity 0.89)
- Active Learning for Continual Learning: Keeping the Past Alive in the Present (ICLR 2025 · similarity 0.88)
- Predicting the Susceptibility of Examples to Catastrophic Forgetting (ICML 2025 · similarity 0.88)
- Stabilizing Zero-Shot Prediction: A Novel Antidote to Forgetting in Continual Vision-Language Tasks (NeurIPS 2024 · similarity 0.87)
- When to Learn and When to Stop: Quitting at the Optimal Time (Student Abstract) (AAAI 2025 · similarity 0.87)
- STAR: Stability-Inducing Weight Perturbation for Continual Learning (ICLR 2025 · 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.
