Deep learning has achieved remarkable successes in many industry applications and scientific research fields. One essential reason is that deep models can learn rich information from large-scale training datasets through supervised learning. It has been well accepted that the robust deep models heavily rely on the quality of data labels. However, current large-scale datasets mostly involve noisy labels, which are caused by sensor errors, human mistakes, or inaccuracy of search engines, and may severely degrade the performance of deep models. In this survey, we summarize existing works on noisy label learning into two main categories, loss correction and sample selection, and present their methodologies, commonly used experimental setups, datasets, and the state-of-the-art results. Finally, we discuss a promising research direction that might be valuable for the future study.

No abstract

No abstract

Samples in large-scale datasets may be mislabeled due to various reasons, and Deep Neural Networks can easily over-fit to the noisy label data. To tackle this problem, the key point is to alleviate the harm of these noisy labels. Many existing methods try to divide training data into clean and noisy subsets in terms of loss values, and then process the noisy label data variedly. One of the reasons hindering a better performance is the hard samples. As hard samples always have relatively large losses whether their labels are clean or noisy, these methods could not divide them precisely. Instead, we propose a Tripartite solution to partition training data more precisely into three subsets: hard, noisy, and clean. The partition criteria are based on the inconsistent predictions of two networks, and the inconsistency between the prediction of a network and the given label. To minimize the harm of noisy labels but maximize the value of noisy label data, we apply a low-weight learning on hard data and a self-supervised learning on noisy label data without using the given labels. Extensive experiments demonstrate that Tripartite can filter out noisy label data more precisely, and outperforms most state-of-the-art methods on five benchmark datasets, especially on real-world datasets.

No abstract