Powering Finetuning in Few-shot Learning: Domain-Agnostic Feature Adaptation with Rectified Class Prototypes
Powering Finetuning in Few-shot Learning: Domain-Agnostic Feature Adaptation with Rectified Class Prototypes
R. Tao,Han Zhang,Yutong Zheng,M. Savvides
2022 · DOI: 10.48550/arXiv.2204.03749
arXiv.org · 1 Citations
TLDR
This paper proposes Distribution Calibration Module (DCM) and Selected Sampling (SS) to reduce, and achieves state-of-the-art results on Meta-Dataset with consistent performance boosts over ten datasets from different domains.