Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Learning Adaptive Classifiers Synthesis for Generalized Few-Shot Learning
Han-Jia Ye, Hexiang Hu, De‐Chuan Zhan
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 the neural dictionary (ablation) | CASTLE- · Heterogeneous GFSL | harmonic mean accuracy: 47.50 -> 49.05 | Improved |
| add the neural dictionary (ablation) | CASTLE- · Office-Home GFSL | harmonic mean accuracy: 33.73 -> 34.48 | Improved |
| adapt head classifiers conditioned on tail examples (aCASTLE) | CASTLE · Heterogeneous GFSL | harmonic mean accuracy: 49.05 -> 52.81 | Improved |
| adapt head classifiers conditioned on tail examples (aCASTLE) | CASTLE · Office-Home GFSL | harmonic mean accuracy: 34.48 -> 39.98 | Improved |
| adapt head classifiers conditioned on tail examples (aCASTLE) | CASTLE · MiniImageNet GFSL 5-shot | harmonic mean accuracy: 76.32 -> 78.33 | Improved |
| adapt head classifiers conditioned on tail examples (aCASTLE) | CASTLE · MiniImageNet GFSL 1-shot | harmonic mean accuracy: 66.22 -> 66.24 | No clear effect |
| add many-shot pre-training of the backbone (ablation) | CASTLE · MiniImageNet 1-shot 5-way GFSL | harmonic mean accuracy: 38.24 -> 66.22 | Improved |
| add many-shot pre-training of the backbone (ablation) | CASTLE · MiniImageNet 1-shot 5-way FSL | mean accuracy: 64.23 -> 66.83 | Improved |
| increase number of classifiers in multi-classifier training from 1 to 256 (ablation) | CASTLE · MiniImageNet 1-shot 5-way GFSL | harmonic mean accuracy: 64.53 -> 66.72 | Improved |
| average synthesized classifiers after synthesis (Post-AVG) instead of averaging embeddings (Pre-AVG) (ablation) | CASTLE · MiniImageNet 5-shot 5-way GFSL | harmonic mean accuracy: 76.32 -> 76.28 | No clear effect |
| add light-weight scale and bias fine-tuning | aCASTLE · MiniImageNet 5-shot 5-way GFSL | harmonic mean accuracy: 78.33 -> 78.93 | Improved |
| add light-weight scale and bias fine-tuning | CASTLE · MiniImageNet 5-shot 5-way GFSL | harmonic mean accuracy: 76.32 -> 76.43 | Improved |
Closest papers by meaning
- Generalized and Incremental Few-Shot Learning by Explicit Learning and Calibration Without Forgetting (ICCV 2021 · similarity 0.89)
- Rethinking Generalization in Few-Shot Classification (NeurIPS 2022 · similarity 0.88)
- Dynamic Few-Shot Visual Learning Without Forgetting (CVPR 2018 · similarity 0.88)
- A Closer Look at Few-shot Classification Again (ICML 2023 · similarity 0.88)
- Making Large Vision Language Models to Be Good Few-Shot Learners (AAAI 2025 · similarity 0.87)
- Rethinking Few-Shot Adaptation of Vision-Language Models in Two Stages (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.
