CReST: A Class-Rebalancing Self-Training Framework for Imbalanced Semi-Supervised Learning

TL;DR

CReST introduces a class-rebalancing self-training framework that enhances semi-supervised learning on imbalanced datasets by selectively rebalancing pseudo-labeled samples, significantly improving performance over existing methods.

Contribution

The paper proposes CReST, a novel self-training framework with class-rebalancing and adaptive distribution alignment, to improve SSL on imbalanced data, outperforming prior rebalancing techniques.

Findings

CReST and CReST+ outperform state-of-the-art SSL methods on imbalanced datasets.

The methods outperform other popular rebalancing techniques.

Pseudo-labels for minority classes are highly precise, enabling effective rebalancing.

Abstract

Semi-supervised learning on class-imbalanced data, although a realistic problem, has been under studied. While existing semi-supervised learning (SSL) methods are known to perform poorly on minority classes, we find that they still generate high precision pseudo-labels on minority classes. By exploiting this property, in this work, we propose Class-Rebalancing Self-Training (CReST), a simple yet effective framework to improve existing SSL methods on class-imbalanced data. CReST iteratively retrains a baseline SSL model with a labeled set expanded by adding pseudo-labeled samples from an unlabeled set, where pseudo-labeled samples from minority classes are selected more frequently according to an estimated class distribution. We also propose a progressive distribution alignment to adaptively adjust the rebalancing strength dubbed CReST+. We show that CReST and CReST+ improve state-of-the-art SSL algorithms on various class-imbalanced datasets and consistently outperform other popular rebalancing methods. Code has been made available at https://github.com/google-research/crest.