Research map / Continual learning / Online and replay-based
Rethinking Gradient Projection Continual Learning: Stability / Plasticity Feature Space Decoupling
Zhen Zhao, Zhizhong Zhang, Xin Tan, Jun Liu, Yanyun Qu, Yuan Xie, Lizhuang Ma
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 |
|---|---|---|---|
| Apply Space Decoupling to TRGP | TRGP · 20-split-MiniImageNet | ACC (%): +1.98% | Improved |
| Apply Space Decoupling to TRGP | TRGP · 20-split-CIFAR-100 | ACC (%): +3.16% | Improved |
| Apply Space Decoupling to TRGP | TRGP · 10-split-CIFAR-100 | ACC (%): +1.04% | Improved |
| Add I/R Approximation (ablation) | GPM · 20-split-CIFAR-100 | ACC (%): 77.53±0.83 -> 79.99±0.48 | Improved |
| Add I/R Approximation (ablation) | TRGP · 20-split-CIFAR-100 | ACC (%): 80.68±0.7 -> 83.21±0.36 | Improved |
| Add I/R Projection (ablation) | GPM · 20-split-CIFAR-100 | ACC (%): 77.53±0.83 -> 78.41±1.3 | Improved |
| Add I/R Projection (ablation) | GPM · 20-split-CIFAR-100 | BWT (%): -0.97±0.59 -> -1.44±0.65 | Hurt |
| Add I/R Projection (ablation) | TRGP · 20-split-CIFAR-100 | ACC (%): 80.68±0.7 -> 81.15±1.22 | Improved |
| Add I/R Projection (ablation) | TRGP · 20-split-CIFAR-100 | BWT (%): -0.87±0.46 -> -1.24±0.71 | Hurt |
| Add both I/R Approximation and I/R Projection (ablation) | GPM · 20-split-CIFAR-100 | ACC (%): 77.53±0.83 -> 80.71±0.82 | Improved |
| Reduce the feature-space dimension using I/R Approximation (ablation) | GPM · 10-split-CIFAR-100, 20-split-CIFAR-100, 20-split-MiniImageNet | ACC | Improved |
| Construct stability and residual subspaces by manually selecting top singular-value bases instead of intersections (ablation) | GPM+SD and TRGP+SD · 10-split-CIFAR-100, 20-split-CIFAR-100 | ACC: 77.1; 80.44; 78.32; 80.75; 83.92; 80.64 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- TRGP — TRGP: Trust Region Gradient Projection for Continual Learning (ICLR 2022) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- SplitLoRA: Balancing Stability and Plasticity in Continual Learning Through Gradient Space Splitting (ICLR 2026 · similarity 0.90)
- BiLoRA: Almost-Orthogonal Parameter Spaces for Continual Learning (CVPR 2025 · similarity 0.88)
- Addressing Loss of Plasticity and Catastrophic Forgetting in Continual Learning (ICLR 2024 · similarity 0.87)
- LoRanPAC: Low-rank Random Features and Pre-trained Models for Bridging Theory and Practice in Continual Learning (ICLR 2025 · similarity 0.87)
- LoRA Subtraction for Drift-Resistant Space in Exemplar-Free Continual Learning (CVPR 2025 · similarity 0.87)
- An Effective Dynamic Gradient Calibration Method for Continual Learning (ICML 2024 · 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.
