Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Free Lunch for Few-shot Learning: Distribution Calibration
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 |
|---|---|---|---|
| Apply distribution calibration with generated feature training | Logistic Regression with DC · miniImageNet 5way1shot | accuracy (%): 56.37 -> 68.57 | Helped |
| Remove Tukey transformation while retaining generated features (ablation) | Distribution calibration · miniImageNet 5way1shot | accuracy (%): 68.57 -> 63.70 | Hurt |
| Remove generated features while retaining Tukey transformation (ablation) | Distribution calibration · miniImageNet 5way1shot | accuracy (%): 68.57 -> 64.30 | Hurt |
| Remove both Tukey transformation and generated features (ablation) | Distribution calibration · miniImageNet 5way1shot | accuracy (%): 68.57 -> 56.37 | Hurt |
| Remove Tukey transformation while retaining generated features (ablation) | Distribution calibration · miniImageNet 5way5shot | accuracy (%): 82.88 -> 82.26 | Hurt |
| Remove generated features while retaining Tukey transformation (ablation) | Distribution calibration · miniImageNet 5way5shot | accuracy (%): 82.88 -> 81.33 | Hurt |
| Use Tukey power λ=0.5 rather than other tested powers (ablation) | Distribution calibration · miniImageNet 5way1shot | accuracy | Helped |
| Increase generated-feature count while below 500 (ablation) | Logistic Regression with DC · miniImageNet 5way1shot | accuracy | Helped |
| Increase retrieved nearest-class features beyond one (ablation) | Nearest-class feature augmentation · miniImageNet 5way1shot | accuracy (%): 62.39 -> 59.73; 58.93; 57.33 | Hurt |
| Replace nearest-class retrieval with calibrated-distribution sampling (ablation) | Feature augmentation · miniImageNet 5way1shot | accuracy (%): 57.33 -> 68.53 | Helped |
| Include the novel support feature when calibrating the mean (ablation) | Distribution calibration · miniImageNet 5way1shot | accuracy (%): 59.38 -> 68.57 | Helped |
| Apply distribution calibration to different feature backbones | Few-shot classifiers on miniImageNet · miniImageNet 5way1shot | accuracy (%): 42.11 -> 54.62; 46.07 -> 57.14; 52.32 -> 61.50; 54.53 -> 64.38; 56.37 -> 68.57 | Helped |
Closest papers by meaning
- Provably Improving Generalization of Few-shot models with Synthetic Data (ICML 2025 · similarity 0.86)
- Divide et Calibra: Multiclass Local Calibration via Vector Quantization (NeurIPS 2026 · similarity 0.85)
- Discriminative Sample-Guided and Parameter-Efficient Feature Space Adaptation for Cross-Domain Few-Shot Learning (CVPR 2024 · similarity 0.85)
- Improving Multi-Class Calibration through Normalization-Aware Isotonic Techniques (ICML 2025 · similarity 0.84)
- Reassessing How to Compare and Improve the Calibration of Machine Learning Models (ICLR 2025 · similarity 0.84)
- FSL-Rectifier: Rectify Outliers in Few-Shot Learning via Test-Time Augmentation (AAAI 2025 · similarity 0.84)
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.