Pseudo-Labeling and Confirmation Bias in Deep Semi-Supervised Learning
This work addresses a key limitation in semi-supervised learning for image classification, demonstrating that pseudo-labeling can outperform consistency regularization, which was previously assumed inferior.
The paper tackles the problem of confirmation bias in pseudo-labeling for semi-supervised image classification, showing that mixup augmentation and a minimum labeled sample per batch reduce this bias, achieving state-of-the-art results on CIFAR-10/100, SVHN, and Mini-ImageNet.
Semi-supervised learning, i.e. jointly learning from labeled and unlabeled samples, is an active research topic due to its key role on relaxing human supervision. In the context of image classification, recent advances to learn from unlabeled samples are mainly focused on consistency regularization methods that encourage invariant predictions for different perturbations of unlabeled samples. We, conversely, propose to learn from unlabeled data by generating soft pseudo-labels using the network predictions. We show that a naive pseudo-labeling overfits to incorrect pseudo-labels due to the so-called confirmation bias and demonstrate that mixup augmentation and setting a minimum number of labeled samples per mini-batch are effective regularization techniques for reducing it. The proposed approach achieves state-of-the-art results in CIFAR-10/100, SVHN, and Mini-ImageNet despite being much simpler than other methods. These results demonstrate that pseudo-labeling alone can outperform consistency regularization methods, while the opposite was supposed in previous work. Source code is available at https://git.io/fjQsC.