English

Few-Shot Learning via Learning the Representation, Provably

Machine Learning 2021-03-31 v2 Optimization and Control Machine Learning

Abstract

This paper studies few-shot learning via representation learning, where one uses TT source tasks with n1n_1 data per task to learn a representation in order to reduce the sample complexity of a target task for which there is only n2(n1)n_2 (\ll n_1) data. Specifically, we focus on the setting where there exists a good \emph{common representation} between source and target, and our goal is to understand how much of a sample size reduction is possible. First, we study the setting where this common representation is low-dimensional and provide a fast rate of O(C(Φ)n1T+kn2)O\left(\frac{\mathcal{C}\left(\Phi\right)}{n_1T} + \frac{k}{n_2}\right); here, Φ\Phi is the representation function class, C(Φ)\mathcal{C}\left(\Phi\right) is its complexity measure, and kk is the dimension of the representation. When specialized to linear representation functions, this rate becomes O(dkn1T+kn2)O\left(\frac{dk}{n_1T} + \frac{k}{n_2}\right) where d(k)d (\gg k) is the ambient input dimension, which is a substantial improvement over the rate without using representation learning, i.e. over the rate of O(dn2)O\left(\frac{d}{n_2}\right). This result bypasses the Ω(1T)\Omega(\frac{1}{T}) barrier under the i.i.d. task assumption, and can capture the desired property that all n1Tn_1T samples from source tasks can be \emph{pooled} together for representation learning. Next, we consider the setting where the common representation may be high-dimensional but is capacity-constrained (say in norm); here, we again demonstrate the advantage of representation learning in both high-dimensional linear regression and neural network learning. Our results demonstrate representation learning can fully utilize all n1Tn_1T samples from source tasks.

Keywords

Cite

@article{arxiv.2002.09434,
  title  = {Few-Shot Learning via Learning the Representation, Provably},
  author = {Simon S. Du and Wei Hu and Sham M. Kakade and Jason D. Lee and Qi Lei},
  journal= {arXiv preprint arXiv:2002.09434},
  year   = {2021}
}

Comments

ICLR2021

R2 v1 2026-06-23T13:49:43.077Z