SOTAVerified

MixMatch: A Holistic Approach to Semi-Supervised Learning

2019-05-06NeurIPS 2019Code Available1· sign in to hype

David Berthelot, Nicholas Carlini, Ian Goodfellow, Nicolas Papernot, Avital Oliver, Colin Raffel

arXivPDF

Code Available — Be the first to reproduce this paper.

Reproduce

Code

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.

Tasks

Image ClassificationSemi-Supervised Image Classification

Benchmark Results

DatasetModelMetricClaimedVerifiedStatus
CIFAR-10MixMatchPercentage correct95.05—Unverified
CIFAR-100MixMatchPercentage correct74.1—Unverified
STL-10CutOutPercentage correct87.36—Unverified
STL-10IICPercentage correct88.8—Unverified
SVHNMixMatchPercentage error2.59—Unverified

Reproductions