Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Z-Score Normalization, Hubness, and Few-Shot Learning
Nanyi Fei, Yizhao Gao, Zhiwu Lu, Tao Xiang
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 z-score normalization | ProtoNet · miniImageNet 1-shot (ResNet-12) | 5-way accuracy (%): 63.38 -> 66.06 | Improved |
| add z-score normalization | IMP · miniImageNet 1-shot (ResNet-12) | 5-way accuracy (%): 63.70 -> 65.01 | Improved |
| add z-score normalization | IEPT · miniImageNet 1-shot (ResNet-12) | 5-way accuracy (%): 64.05 -> 67.35 | Improved |
| add z-score normalization | ProtoNet · miniImageNet 5-shot (ResNet-12) | 5-way accuracy (%): 81.22 -> 81.73 | Improved |
| add z-score normalization | ProtoNet · tieredImageNet 1-shot (Conv4-64) | 5-way accuracy (%): 53.56 -> 56.70 | Improved |
| add z-score normalization | IEPT · CUB 1-shot (Conv4-64) | 5-way accuracy (%): 68.27 -> 73.54 | Improved |
| add z-score normalization to pre-trained features (ablation) | pre-trained model · miniImageNet 1-shot | 5-way accuracy (%): 49.89 -> 52.80 | Improved |
| add z-score normalization to pre-trained features (ablation) | pre-trained model · tieredImageNet 1-shot | 5-way accuracy (%): 51.12 -> 53.76 | Improved |
| add z-score normalization to pre-trained features (ablation) | pre-trained model · miniImageNet 1-shot skewness | skewness of k-occurrence (lower is better): 3.219 -> 2.360 | Improved |
| use ZN instead of SimpleShot centering (ablation) | pre-trained model · miniImageNet 1-shot | 5-way accuracy (%): 53.21 -> 52.80 | Hurt |
| use ZN instead of SimpleShot centering (ablation) | pre-trained model · tieredImageNet 1-shot | 5-way accuracy (%): 52.89 -> 53.76 | Improved |
| use ZN instead of l2 normalization (ablation) | ProtoNet episodic training · miniImageNet 1-shot | 5-way accuracy (%): 53.99 -> 55.16 | Improved |
Closest papers by meaning
- Few-Shot Learning With a Strong Teacher (TPAMI 2022 · similarity 0.89)
- Class Normalization for (Continual)? Generalized Zero-Shot Learning (ICLR 2021 · similarity 0.88)
- Pushing the Limit of Fine-Tuning for Few-Shot Learning: Where Feature Reusing Meets Cross-Scale Attention (AAAI 2024 · similarity 0.87)
- Scaling Few-Shot Learning for the Open World (AAAI 2024 · similarity 0.86)
- Boosting Few-Shot Learning via Attentive Feature Regularization (AAAI 2024 · similarity 0.86)
- FSL-Rectifier: Rectify Outliers in Few-Shot Learning via Test-Time Augmentation (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.
