Improving Performance of Semi-Supervised Learning by Adversarial Attacks

TL;DR

This paper introduces SCAR, a framework that enhances semi-supervised learning by using adversarial attacks to select high-confidence unlabeled data, leading to improved image classification performance.

Contribution

The paper proposes a novel framework, SCAR, which leverages adversarial attacks to improve semi-supervised learning by better selecting unlabeled data for labeling.

Findings

Significant improvement in CIFAR10 classification accuracy with SCAR.

Adversarial attacks effectively identify high-confidence unlabeled samples.

Enhanced performance of recent SSL algorithms using the SCAR framework.

Abstract

Semi-supervised learning (SSL) algorithm is a setup built upon a realistic assumption that access to a large amount of labeled data is tough. In this study, we present a generalized framework, named SCAR, standing for Selecting Clean samples with Adversarial Robustness, for improving the performance of recent SSL algorithms. By adversarially attacking pre-trained models with semi-supervision, our framework shows substantial advances in classifying images. We introduce how adversarial attacks successfully select high-confident unlabeled data to be labeled with current predictions. On CIFAR10, three recent SSL algorithms with SCAR result in significantly improved image classification.