Learning with Non-Uniform Label Noise: A Cluster-Dependent Semi-Supervised Approach

Mengtian Zhang, Bo Jiang, Yuye Ling, Xinbing Wang, Chenghu Zhou

Keywords: Non-uniform label noise, Cluster-dependent sample selection mechanism, Semi-supervised training.

TL;DR: For the robust learning with non-uniform label noise, we propose a cluster-dependent sample selection algorithm followed by a semi-supervised training mechanism.

Abstract: Learning with noisy labels is a challenging task in machine learning. Most existing methods explicitly or implicitly assume uniform label noise across all samples. In reality, label noise can be highly non-uniform in the feature space, with higher error rate for more difficult samples. Some recent works consider instance-dependent label noise but they require additional information such as some cleanly labeled data and confidence scores, which are usually unavailable or costly to obtain. In this paper, we consider learning with non-uniform label noise that requires no such additional information. we propose a cluster-dependent sample selection algorithm followed by a semi-supervised training mechanism based on the cluster-dependent label noise. The proposed self-adaptive multi-scale sample selection method increases the consistency of sample space by forcing the selection of clean samples from the entire feature space. Despite its simplicity, the proposed method can distinguish clean data from the corrupt ones more precisely and achieve state-of-the-art performance on image classification benchmarks, especially when the number of training samples is small and the noise rate is large.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

Submission Guidelines: Yes

Please Choose The Closest Area That Your Submission Falls Into: Deep Learning and representational learning