Combating Semantic Contamination in Learning with Label Noise

TL;DR

This paper identifies Semantic Contamination caused by label refurbishment in noisy label learning and proposes Collaborative Cross Learning, a semi-supervised approach that improves robustness against label noise by better preserving semantic relationships.

Contribution

The paper introduces Collaborative Cross Learning, a novel semi-supervised method that effectively mitigates Semantic Contamination in label refurbishment for noisy label learning.

Findings

Outperforms existing methods on synthetic noisy datasets

Effectively reduces Semantic Contamination impacts

Improves robustness of deep neural networks to label noise

Abstract

Noisy labels can negatively impact the performance of deep neural networks. One common solution is label refurbishment, which involves reconstructing noisy labels through predictions and distributions. However, these methods may introduce problematic semantic associations, a phenomenon that we identify as Semantic Contamination. Through an analysis of Robust LR, a representative label refurbishment method, we found that utilizing the logits of views for refurbishment does not adequately balance the semantic information of individual classes. Conversely, using the logits of models fails to maintain consistent semantic relationships across models, which explains why label refurbishment methods frequently encounter issues related to Semantic Contamination. To address this issue, we propose a novel method called Collaborative Cross Learning, which utilizes semi-supervised learning on refurbished labels to extract appropriate semantic associations from embeddings across views and models. Experimental results show that our method outperforms existing approaches on both synthetic and real-world noisy datasets, effectively mitigating the impact of label noise and Semantic Contamination.