Efficient Adaptive Label Refinement for label noise learning

doi:10.1016/j.neucom.2025.130305

Efficient Adaptive Label Refinement for label noise learning

2025-07-01

0

OA

AI

DOI：10.1016/j.neucom.2025.130305

Original

Save

View PDF

Abstract

En 中文

Deep neural networks are highly susceptible to overfitting noisy labels, which leads to degraded performance. Existing methods address this issue by employing complex manually designed strategies, aiming to achieve optimal partitioning in each iteration to avoid fitting noisy labels while thoroughly learning clean samples. However, this often results in models that are overly sensitive to hyperparameters and challenging to train. To address this issue, we decouple the tasks of avoiding fitting incorrect labels and thoroughly learning clean samples, proposing a simple yet highly applicable method called Adaptive Label Refinement (ALR). First, inspired by label refurbishment techniques, we update the original hard labels to soft labels using the model's predictions to reduce the risk of fitting incorrect labels. Then, by introducing entropy loss, the model adaptively 'harden' the high-confidence soft labels based on the confidence differences, guiding the model to focus more on learning from clean samples. This approach is simple and efficient, requiring no auxiliary datasets and minimal hyperparameter tuning, making it more adaptable than existing methods that require careful adjustments for different datasets. We validate ALR's effectiveness through experiments on benchmark datasets with artificial label noise (CIFAR-10/100) and real-world datasets with inherent noise (ANIMAL-10N, Clothing1M, WebVision). The results show that ALR outperforms state-of-the-art methods.

Keywords:

Deep neural network

Supervised learning

Label noise

Label refinement

Efficient Adaptive Label Refinement for label noise learning

Abstract

AI Summary

Journal

Organization