Extended Few-Shot Learning: Exploiting Existing Resources for Novel Tasks

TL;DR

This paper introduces extended few-shot learning, leveraging auxiliary datasets through a novel masking approach to improve classification accuracy in scenarios with scarce labeled data.

Contribution

It proposes a framework that automatically selects and adapts auxiliary data using a masking module, outperforming naive transfer learning methods.

Findings

Masking module improves accuracy by 4.68 percentage points over naive methods.

Using auxiliary data with the proposed method yields significant performance gains.

Naive approaches only modestly increase accuracy, highlighting the effectiveness of the proposed adaptation.

Abstract

In many practical few-shot learning problems, even though labeled examples are scarce, there are abundant auxiliary datasets that potentially contain useful information. We propose the problem of extended few-shot learning to study these scenarios. We then introduce a framework to address the challenges of efficiently selecting and effectively using auxiliary data in few-shot image classification. Given a large auxiliary dataset and a notion of semantic similarity among classes, we automatically select pseudo shots, which are labeled examples from other classes related to the target task. We show that naive approaches, such as (1) modeling these additional examples the same as the target task examples or (2) using them to learn features via transfer learning, only increase accuracy by a modest amount. Instead, we propose a masking module that adjusts the features of auxiliary data to be more similar to those of the target classes. We show that this masking module performs better than naively modeling the support examples and transfer learning by 4.68 and 6.03 percentage points, respectively.