MixMatch: A Holistic Approach to Semi-Supervised Learning

TL;DR

MixMatch introduces a unified semi-supervised learning algorithm that effectively leverages unlabeled data through label guessing and data mixing, achieving state-of-the-art results and improved privacy-accuracy trade-offs.

Contribution

It unifies existing semi-supervised methods into a new algorithm, MixMatch, combining label guessing and MixUp to enhance learning from unlabeled data.

Findings

Significantly reduces error rates on CIFAR-10 and STL-10 datasets.

Achieves state-of-the-art performance across multiple datasets.

Improves privacy-accuracy trade-offs in differential privacy settings.

Abstract

Semi-supervised learning has proven to be a powerful paradigm for leveraging unlabeled data to mitigate the reliance on large labeled datasets. In this work, we unify the current dominant approaches for semi-supervised learning to produce a new algorithm, MixMatch, that works by guessing low-entropy labels for data-augmented unlabeled examples and mixing labeled and unlabeled data using MixUp. We show that MixMatch obtains state-of-the-art results by a large margin across many datasets and labeled data amounts. For example, on CIFAR-10 with 250 labels, we reduce error rate by a factor of 4 (from 38% to 11%) and by a factor of 2 on STL-10. We also demonstrate how MixMatch can help achieve a dramatically better accuracy-privacy trade-off for differential privacy. Finally, we perform an ablation study to tease apart which components of MixMatch are most important for its success.