Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results
This addresses the challenge of efficient semi-supervised learning for computer vision tasks, offering incremental improvements over existing methods like Temporal Ensembling.
The paper tackles the problem of semi-supervised learning by proposing Mean Teacher, which averages model weights to improve consistency targets, resulting in error rates of 4.35% on SVHN with 250 labels and reducing error on CIFAR-10 from 10.55% to 6.28% with 4000 labels.
The recently proposed Temporal Ensembling has achieved state-of-the-art results in several semi-supervised learning benchmarks. It maintains an exponential moving average of label predictions on each training example, and penalizes predictions that are inconsistent with this target. However, because the targets change only once per epoch, Temporal Ensembling becomes unwieldy when learning large datasets. To overcome this problem, we propose Mean Teacher, a method that averages model weights instead of label predictions. As an additional benefit, Mean Teacher improves test accuracy and enables training with fewer labels than Temporal Ensembling. Without changing the network architecture, Mean Teacher achieves an error rate of 4.35% on SVHN with 250 labels, outperforming Temporal Ensembling trained with 1000 labels. We also show that a good network architecture is crucial to performance. Combining Mean Teacher and Residual Networks, we improve the state of the art on CIFAR-10 with 4000 labels from 10.55% to 6.28%, and on ImageNet 2012 with 10% of the labels from 35.24% to 9.11%.