MixMatch: A Holistic Approach to Semi-Supervised Learning

6 May 2019David BerthelotNicholas CarliniIan GoodfellowNicolas PapernotAvital OliverColin Raffel

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... (read more)

PDF Abstract

Evaluation results from the paper

Task Dataset Model Metric name Metric value Global rank Compare
Semi-Supervised Image Classification CIFAR-10, 4000 Labels MixMatch Accuracy 93.76 # 3
Image Classification STL-10 MixMatch Percentage correct 94.41 # 1
Semi-Supervised Image Classification STL-10 MixMatch Accuracy 94.41 # 1
Semi-Supervised Image Classification SVHN, 1000 labels MixMatch Accuracy 96.73 # 2