Interpolation Consistency Training for Semi-Supervised Learning

9 Mar 2019  ·  Vikas Verma, Kenji Kawaguchi, Alex Lamb, Juho Kannala, Yoshua Bengio, David Lopez-Paz ·

We introduce Interpolation Consistency Training (ICT), a simple and computation efficient algorithm for training Deep Neural Networks in the semi-supervised learning paradigm. ICT encourages the prediction at an interpolation of unlabeled points to be consistent with the interpolation of the predictions at those points... In classification problems, ICT moves the decision boundary to low-density regions of the data distribution. Our experiments show that ICT achieves state-of-the-art performance when applied to standard neural network architectures on the CIFAR-10 and SVHN benchmark datasets. Our theoretical analysis shows that ICT corresponds to a certain type of data-adaptive regularization with unlabeled points which reduces overfitting to labeled points under high confidence values. read more

PDF Abstract

Datasets


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Semi-Supervised Image Classification CIFAR-10, 1000 Labels ICT (CNN-13) Accuracy 84.52 # 6
Semi-Supervised Image Classification CIFAR-10, 2000 Labels ICT (CNN-13) Accuracy 90.74 # 2
Semi-Supervised Image Classification CIFAR-10, 4000 Labels ICT (CNN-13) Percentage error 7.29 # 22
Semi-Supervised Image Classification CIFAR-10, 4000 Labels ICT (WRN-28-2) Percentage error 7.66 # 23
Semi-Supervised Image Classification SVHN, 1000 labels ICT Accuracy 96.11 # 10
Semi-Supervised Image Classification SVHN, 1000 labels ICT (WRN-28-2) Accuracy 96.47 # 8

Methods


No methods listed for this paper. Add relevant methods here