Research map / Continual learning / Online and replay-based
Mitigating Forgetting in Online Continual Learning via Instance-Aware Parameterization
Hung-Jen Chen, An-Chieh Cheng, Da-Cheng Juan, Wei Wei, Min Sun
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 |
|---|---|---|---|
| use instance-aware path selection with shared meta-graph instead of independent ResNet-18 per task | InstAParam · Split CIFAR-10 | Avg. accuracy (%): 81.5 -> 83.8 | Improved |
| use instance-aware path selection with shared meta-graph instead of independent ResNet-18 per task | InstAParam · Split Tiny-ImageNet | Accuracy (%): 27.2 -> 36.8 | Improved |
| use instance-aware paths versus replay-based A-GEM | InstAParam · Split CIFAR-10 | Avg. accuracy (%): 77.1 -> 83.8 | Improved |
| use smaller single-superkernel meta-graph (15M) instead of full meta-graph (43M) | InstAParam · Split CIFAR-100 | Accuracy (%): 54.4 -> 55.5 | Improved |
| use smaller single-superkernel meta-graph (15M) instead of full meta-graph (43M) | InstAParam · Split Tiny-ImageNet | Accuracy (%): 36.8 -> 33.0 | Hurt |
| blur task boundaries (20% data exchange) in CIFAR-10 | InstAParam · Split CIFAR-10 | Avg. accuracy (%): 83.8 -> 83.5 | No clear effect |
| blur task boundaries (20% data exchange) in CIFAR-10 | EWC · Split CIFAR-10 | Avg. accuracy (%): 73.0 -> 62.1 | Hurt |
| blur task boundaries (20% data exchange) in CIFAR-10 | GSS-Greedy · Split CIFAR-10 | Avg. accuracy (%): 72.8 -> 82.5 | Improved |
Closest papers by meaning
- Do Your Best and Get Enough Rest for Continual Learning (CVPR 2025 · similarity 0.88)
- Predicting the Susceptibility of Examples to Catastrophic Forgetting (ICML 2025 · similarity 0.88)
- STAR: Stability-Inducing Weight Perturbation for Continual Learning (ICLR 2025 · similarity 0.88)
- Federated Continual Learning Goes Online: Uncertainty-Aware Memory Management for Vision Tasks and Beyond (ICLR 2025 · similarity 0.88)
- Train with Perturbation, Infer after Merging: A Two-Stage Framework for Continual Learning (NeurIPS 2025 · similarity 0.88)
- Layerwise Proximal Replay: A Proximal Point Method for Online Continual Learning (ICML 2024 · similarity 0.88)
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.
