TL;DRThis paper proposes a method to improve the robustness of deep learning models in the presence of noisy labels by utilizing unsupervised learning and cluster regularization.
Abstract
Supervised learning of deep neural networks heavily relies on large-scale datasets annotated by high-quality labels. In contrast, mislabeled samples can significantly degrade the generalization of models and result in memorizing samples, further learning erroneous associations of data contents to incorrect annotations. To this end, this paper proposes an eff