AuxMix: Semi-Supervised Learning with Unconstrained Unlabeled Data
It addresses a practical limitation in SSL for scenarios with unconstrained unlabeled data, though it is incremental by building on existing SSL and self-supervised methods.
The paper tackles the problem of semi-supervised learning performance degradation when unlabeled data comes from a different distribution than labeled data, proposing AuxMix to mask dissimilar auxiliary data and improve accuracy by 5% on CIFAR10 with ResNet-50.
Semi-supervised learning (SSL) has seen great strides when labeled data is scarce but unlabeled data is abundant. Critically, most recent work assume that such unlabeled data is drawn from the same distribution as the labeled data. In this work, we show that state-of-the-art SSL algorithms suffer a degradation in performance in the presence of unlabeled auxiliary data that does not necessarily possess the same class distribution as the labeled set. We term this problem as Auxiliary-SSL and propose AuxMix, an algorithm that leverages self-supervised learning tasks to learn generic features in order to mask auxiliary data that are not semantically similar to the labeled set. We also propose to regularize learning by maximizing the predicted entropy for dissimilar auxiliary samples. We show an improvement of 5% over existing baselines on a ResNet-50 model when trained on CIFAR10 dataset with 4k labeled samples and all unlabeled data is drawn from the Tiny-ImageNet dataset. We report competitive results on several datasets and conduct ablation studies.