This project will be used for research in semi-supervised learning (SSL). Semi-supervised learning means learning where only part of training data has labels. Some of the open problems within SSL that we aim to study are e.g. improving training times in SSL, uncertainty predicitons, and dealing with unknown classes in the the unlabeled training set. The research is done mainly on open datasets for computer vision, such as CIFAR-10, CIFAR-100, and ImageNet, but we also plan to study other domains such as radar.