MixMatch: A Holistic Approach to Semi-Supervised Learning

NeurIPS 2019 David 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
TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK RESULT LEADERBOARD
Image Classification CIFAR-10 MixMatch Percentage correct 95.05 # 37
Image Classification CIFAR-100 MixMatch Percentage correct 74.1 # 38
Semi-Supervised Image Classification CIFAR-10, 1000 Labels MixMatch Accuracy 92.25 # 1
Semi-Supervised Image Classification CIFAR-10, 2000 Labels MixMatch Accuracy 92.97 # 1
Semi-Supervised Image Classification CIFAR-10, 250 Labels MixMatch Accuracy 88.92 # 5
Semi-Supervised Image Classification CIFAR-10, 4000 Labels MixMatch Accuracy 93.76 # 6
Semi-Supervised Image Classification CIFAR-10, 500 Labels MixMatch Accuracy 91.35 # 1
Semi-Supervised Image Classification STL-10, 1000 Labels MixMatch Accuracy 89.82 # 4
Semi-Supervised Image Classification STL-10, 5000 Labels MixMatch Accuracy 94.41 # 1
Image Classification SVHN MixMatch Percentage error 2.59 # 23
Semi-Supervised Image Classification SVHN, 1000 labels MixMatch Accuracy 96.73 # 5
Semi-Supervised Image Classification SVHN, 2000 Labels MixMatch Accuracy 96.96 # 1
Semi-Supervised Image Classification SVHN, 250 Labels MixMatch Accuracy 96.22 # 3
Semi-Supervised Image Classification SVHN, 4000 Labels MixMatch Accuracy 97.11 # 1
Semi-Supervised Image Classification SVHN, 500 Labels MixMatch Accuracy 96.36 # 1