|IEEE/CVF Conference on Computer Vision and Pattern Recognition 2022
Propagation Regularizer for Semi-supervised Learning with Extremely Scarce Labeled Samples
Abstract : Semi-supervised learning (SSL) is a method to make better models using a large number of easily accessible unlabeled data along with a small number of labeled data obtained at a high cost. Most of existing SSL studies focused on the cases where sufficient labeled samples were available, tens to hundreds labeled samples for class, which still requires a lot of labeling cost. In this paper, we focus on the SSL environment with extremely scarce labeled samples, 1 or 2 labeled samples per class, where most of existing methods failed to learn. We propose a propagation regularizer which can achieve efficient and effective learning with extremely scarce labeled samples by suppressing confirmation bias. In addition, for the realistic model selection in the absence of the validation dataset, we also propose a model selection method based on the propagation regularization in the SSL environment. The proposed methods show 70.9%, 30.3%, and 78.9% accuracy on CIFAR-10, CIFAR100, SVHN dataset with just one labeled sample per class, which are improved by 8.9% to 120.2% compared to the existing approaches.