On the Importance of Distractors for Few-Shot Classification

TL;DR

This paper introduces a contrastive learning-based finetuning method for few-shot classification that leverages unlabelled distractor examples from unrelated classes, significantly improving generalization especially across domains.

Contribution

The paper presents a novel finetuning approach using distractors in contrastive learning, which is label-agnostic and enhances few-shot classification performance.

Findings

Up to 12% accuracy improvement in cross-domain tasks.

Up to 5% accuracy gain in unsupervised prior-learning.

Demonstrates the effectiveness of distractors in boosting few-shot generalization.

Abstract

Few-shot classification aims at classifying categories of a novel task by learning from just a few (typically, 1 to 5) labelled examples. An effective approach to few-shot classification involves a prior model trained on a large-sample base domain, which is then finetuned over the novel few-shot task to yield generalizable representations. However, task-specific finetuning is prone to overfitting due to the lack of enough training examples. To alleviate this issue, we propose a new finetuning approach based on contrastive learning that reuses unlabelled examples from the base domain in the form of distractors. Unlike the nature of unlabelled data used in prior works, distractors belong to classes that do not overlap with the novel categories. We demonstrate for the first time that inclusion of such distractors can significantly boost few-shot generalization. Our technical novelty includes a stochastic pairing of examples sharing the same category in the few-shot task and a weighting term that controls the relative influence of task-specific negatives and distractors. An important aspect of our finetuning objective is that it is agnostic to distractor labels and hence applicable to various base domain settings. Compared to state-of-the-art approaches, our method shows accuracy gains of up to $12\%$ in cross-domain and up to $5\%$ in unsupervised prior-learning settings.