Research map / Continual learning / Online and replay-based
Balancing Stability and Plasticity through Advanced Null Space in Continual Learning
Yajing Kong, Liu Liu, Zhen Wang, Dacheng Tao
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 shared low-rank null space (ablation) | null space projection · 20-Split CIFAR-100 | BWT (%): -7.67 -> -3.66 | Improved |
| add shared low-rank null space (ablation) | null space projection · 20-Split CIFAR-100 | ACC (%): 75.53 -> 76.31 | Improved |
| add shared low-rank null space (ablation) | null space projection · 25-Split TinyImageNet | ACC (%): 59.20 -> 59.26 | No clear effect |
| obtain shared null space by low-rank approximation instead of random dimensions (ablation) | AdNS without distillation · 10-Split CIFAR-100 | BWT (%): -4.13 -> -2.87 | Improved |
| obtain shared null space by low-rank approximation instead of random dimensions (ablation) | AdNS without distillation · 20-Split CIFAR-100 | ACC (%): 75.70 -> 76.31 | Improved |
| add non-uniform constraint strength (ablation) | null space projection · 25-Split TinyImageNet | BWT (%): -7.69 -> -5.07 | Improved |
| add non-uniform constraint strength (ablation) | null space projection · 20-Split CIFAR-100 | ACC (%): 75.53 -> 75.66 | No clear effect |
| add intra-task distillation (ablation) | null space projection · 20-Split CIFAR-100 | ACC (%): 75.53 -> 77.04 | Improved |
| add intra-task distillation (ablation) | null space projection · 25-Split TinyImageNet | ACC (%): 59.20 -> 60.00 | Improved |
| combine all modules (ablation) | null space projection · 20-Split CIFAR-100 | ACC (%): 75.53 -> 77.33 | Improved |
| replace Adam-NSCL with AdNS | Adam-NSCL · 10-Split CIFAR-100 | ACC (%): 75.03 -> 77.21 | Improved |
| replace Adam-NSCL with AdNS | Adam-NSCL · 25-Split TinyImageNet | BWT (%): -7.19 -> -4.58 | Improved |
Closest papers by meaning
- Training Networks in Null Space of Feature Covariance for Continual Learning (CVPR 2021 · similarity 0.90)
- Achieving a Better Stability-Plasticity Trade-Off via Auxiliary Networks in Continual Learning (CVPR 2023 · similarity 0.90)
- STAR: Stability-Inducing Weight Perturbation for Continual Learning (ICLR 2025 · similarity 0.89)
- Rethinking the Stability-Plasticity Trade-off in Continual Learning from an Architectural Perspective (ICML 2025 · similarity 0.88)
- Re-evaluating Continual Learning with Few-Shot Adaptation (NeurIPS 2026 · similarity 0.87)
- PaLoRA: Paced Low-Rank Adaptation for Continual Learning (NeurIPS 2026 · 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.
