Uniform Sampling over Episode Difficulty

TL;DR

This paper introduces a difficulty-based episode sampling method for episodic training in few-shot learning, demonstrating that uniform sampling over episode difficulty improves performance across various algorithms and datasets.

Contribution

Proposes a difficulty-based episode sampling approach and shows that uniform sampling over difficulty outperforms other schemes in few-shot learning.

Findings

Uniform sampling over episode difficulty outperforms curriculum and easy-/hard-mining.

The method improves few-shot learning accuracy across multiple algorithms.

The approach is algorithm agnostic and broadly applicable.

Abstract

Episodic training is a core ingredient of few-shot learning to train models on tasks with limited labelled data. Despite its success, episodic training remains largely understudied, prompting us to ask the question: what is the best way to sample episodes? In this paper, we first propose a method to approximate episode sampling distributions based on their difficulty. Building on this method, we perform an extensive analysis and find that sampling uniformly over episode difficulty outperforms other sampling schemes, including curriculum and easy-/hard-mining. As the proposed sampling method is algorithm agnostic, we can leverage these insights to improve few-shot learning accuracies across many episodic training algorithms. We demonstrate the efficacy of our method across popular few-shot learning datasets, algorithms, network architectures, and protocols.