Research map / Continual learning / Class-incremental
Overcoming Catastrophic Forgetting in Incremental Few-Shot Learning by Finding Flat Minima
Guangyuan SHI, Jiaxin Chen, Wenlong Zhang, Li-Ming Zhan, Xiao-Ming Wu
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 flat-minima training and clamped fine-tuning (F2M) over frozen-base Baseline | incremental few-shot learner · CIFAR-100, session 9 | accuracy (%): 43.83 -> 44.67 | Improved |
| F2M over frozen-base Baseline | incremental few-shot learner · miniImageNet, session 9 | accuracy (%): 43.88 -> 44.65 | Improved |
| F2M over frozen-base Baseline | incremental few-shot learner · CUB-200-2011, session 11 | accuracy (%): 58.12 -> 60.26 | Improved |
| freeze embedding and never adapt (Baseline) versus complex IFL/IL methods | incremental few-shot learner · CIFAR-100, session 9 | accuracy (%): 40.08 (joint) / 42.22 (FSLL*) -> 43.83 | Improved |
| add parameter clamping to the flat region during incremental sessions (ablation) | F2M · CIFAR-100, session 9 | accuracy (%): 36.01 -> 44.67 | Improved |
| add prototype fixing term to base objective (ablation) | F2M · CIFAR-100, session 9 | accuracy (%): 44.08 -> 44.67 | Improved |
| add prototype normalization (ablation) | F2M · CIFAR-100, session 9 | accuracy (%): 44.43 -> 44.67 | Improved |
| add flat-minima noise injection in base training (with clamping only vs full) (ablation) | F2M · CIFAR-100, session 9 | accuracy (%): 42.55 -> 44.67 | Improved |
| add parameter clamping to a model without flat-minima search (ablation) | Baseline-style training · CIFAR-100, session 9 | accuracy (%): 06.45 -> 42.55 | Improved |
| increase flat region bound b from 0.02 to 0.08 (ablation) | F2M · CIFAR-100, session 9 | accuracy (%): 44.67 -> 43.04 | Hurt |
| decrease flat region bound b from 0.01 to 0.0025 (ablation) | F2M · CIFAR-100, session 9 | accuracy (%): 44.58 -> 44.16 | Hurt |
| noise-based flat minima search versus standard SGD training (ablation) | base training · CIFAR-100 flatness on training set | flatness indicator I (lower is flatter): 0.2993 -> 0.0506 | Improved |
Closest papers by meaning
- Generalized and Incremental Few-Shot Learning by Explicit Learning and Calibration Without Forgetting (ICCV 2021 · similarity 0.90)
- Pre-trained Vision and Language Transformers Are Few-Shot Incremental Learners (CVPR 2024 · similarity 0.88)
- Re-evaluating Continual Learning with Few-Shot Adaptation (NeurIPS 2026 · similarity 0.87)
- Adaptive Decision Boundary for Few-Shot Class-Incremental Learning (AAAI 2025 · similarity 0.87)
- Long-Tail Class Incremental Learning via Independent Sub-prototype Construction (CVPR 2024 · similarity 0.87)
- Tripartite Weight-Space Ensemble for Few-Shot Class-Incremental Learning (CVPR 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.
