Optimized Generic Feature Learning for Few-shot Classification across Domains

TL;DR

This paper introduces a hyper-parameter optimization strategy that leverages cross-domain, cross-task data as validation to learn features that generalize well across various tasks and domains, significantly improving few-shot image classification.

Contribution

It proposes a novel hyper-parameter optimization approach using cross-domain data as validation, enhancing feature generalization for few-shot learning.

Findings

Features learned outperform previous few-shot methods

Effective across multiple domains and tasks

Hyper-parameter optimization improves generalization

Abstract

To learn models or features that generalize across tasks and domains is one of the grand goals of machine learning. In this paper, we propose to use cross-domain, cross-task data as validation objective for hyper-parameter optimization (HPO) to improve on this goal. Given a rich enough search space, optimization of hyper-parameters learn features that maximize validation performance and, due to the objective, generalize across tasks and domains. We demonstrate the effectiveness of this strategy on few-shot image classification within and across domains. The learned features outperform all previous few-shot and meta-learning approaches.