English

Embedding Adaptation is Still Needed for Few-Shot Learning

Machine Learning 2021-04-16 v1 Computer Vision and Pattern Recognition

Abstract

Constructing new and more challenging tasksets is a fruitful methodology to analyse and understand few-shot classification methods. Unfortunately, existing approaches to building those tasksets are somewhat unsatisfactory: they either assume train and test task distributions to be identical -- which leads to overly optimistic evaluations -- or take a "worst-case" philosophy -- which typically requires additional human labor such as obtaining semantic class relationships. We propose ATG, a principled clustering method to defining train and test tasksets without additional human knowledge. ATG models train and test task distributions while requiring them to share a predefined amount of information. We empirically demonstrate the effectiveness of ATG in generating tasksets that are easier, in-between, or harder than existing benchmarks, including those that rely on semantic information. Finally, we leverage our generated tasksets to shed a new light on few-shot classification: gradient-based methods -- previously believed to underperform -- can outperform metric-based ones when transfer is most challenging.

Keywords

Cite

@article{arxiv.2104.07255,
  title  = {Embedding Adaptation is Still Needed for Few-Shot Learning},
  author = {Sébastien M. R. Arnold and Fei Sha},
  journal= {arXiv preprint arXiv:2104.07255},
  year   = {2021}
}

Comments

In submission

R2 v1 2026-06-24T01:11:15.121Z