Research map / Continual learning / Class-incremental
Subspace Regularizers for Few-Shot Class Incremental Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| regularize new class weights toward the base-class subspace | Fine-tuning on ResNet-18 · miniImageNet multi-session session 8, -M | weighted average accuracy (%): 26.54 -> 46.79 | Helped |
| use semantic class information to target novel weights | Subspace Regularization · miniImageNet multi-session session 8, -M | weighted average accuracy (%): 46.79 -> 47.55 | Helped |
| retain one example per previous class (+M) | Fine-tuning · miniImageNet multi-session session 8 | weighted average accuracy (%): 26.54 -> 48.13 | Helped |
| use subspace regularization instead of fine-tuning | ResNet-12 classifier · miniImageNet single-session 5-shot | accuracy (%): 66.54 -> 68.88 | Helped |
| use subspace regularization instead of fine-tuning | ResNet-18 classifier · tieredImageNet single-session 1-shot | accuracy (%): 64.42 -> 64.39 | No clear effect |
| remove R_old from Semantic Subspace Reg. (ablation) | Semantic Subspace Reg. · miniImageNet multi-session session 8, -M | weighted average accuracy (%): 47.55 -> 33.25 | Hurt |
| remove R_old from Fine-tuning (ablation) | Fine-tuning · miniImageNet multi-session session 8, -M | weighted average accuracy (%): 26.54 -> 20.39 | Hurt |
| use semantic subspace regularization rather than linear mapping | ResNet-18 classifier · miniImageNet multi-session session 8, -M | weighted average accuracy (%): 46.90 -> 47.55 | Helped |
| use semantic subspace regularization rather than linear mapping | ResNet-12 classifier · miniImageNet single-session 5-shot | accuracy (%): 69.68 (Linear Mapping); 69.75 (label); 68.46 (description) | Mixed |
Closest papers by meaning
- Tripartite Weight-Space Ensemble for Few-Shot Class-Incremental Learning (CVPR 2025 · similarity 0.89)
- Adaptive Subspaces for Few-Shot Learning (CVPR 2020 · similarity 0.89)
- Few-Shot Incremental Learning With Continually Evolved Classifiers (CVPR 2021 · similarity 0.88)
- OrCo: Towards Better Generalization via Orthogonality and Contrast for Few-Shot Class-Incremental Learning (CVPR 2024 · similarity 0.88)
- Boosting Few-Shot Learning via Attentive Feature Regularization (AAAI 2024 · similarity 0.87)
- Adaptive Decision Boundary for Few-Shot Class-Incremental Learning (AAAI 2025 · similarity 0.86)
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.