Distance-weighted Sinkhorn loss for Alzheimer’s disease classification

• Published in: iScience Vol. 27; no. 3; p. 109212
• Main Authors: Wang, Zexuan, Zhan, Qipeng, Tong, Boning, Yang, Shu, Hou, Bojian, Huang, Heng, Saykin, Andrew J., Thompson, Paul M., Davatzikos, Christos, Shen, Li
• Format: Journal Article
• Language: English
• Published: United States Elsevier Inc 15.03.2024; Elsevier
• Subjects: Biocomputational method; Classification of bioinformatical subject; Machine learning; Medical informatics; Neural networks; Classification of bioinformatical subject; Neural networks; Machine learning; Medical informatics; Biocomputational method
• ISSN: 2589-0042; 2589-0042
• DOI: 10.1016/j.isci.2024.109212

Traditional loss functions such as cross-entropy loss often quantify the penalty for each mis-classified training sample without adequately considering its distance from the ground truth class distribution in the feature space. Intuitively, the larger this distance is, the higher the penalty should be. With this observation, we propose a penalty called distance-weighted Sinkhorn (DWS) loss. For each mis-classified training sample (with predicted label A and true label B), its contribution to the DWS loss positively correlates to the distance the training sample needs to travel to reach the ground truth distribution of all the A samples. We apply the DWS framework with a neural network to classify different stages of Alzheimer’s disease. Our empirical results demonstrate that the DWS framework outperforms the traditional neural network loss functions and is comparable or better to traditional machine learning methods, highlighting its potential in biomedical informatics and data science. [Display omitted] •Our loss function aims to learn the data-wise label distribution•Our loss function is theoretically based on Wasserstein distance•Our method better classifies Alzheimer’s disease stages•The code is publicly available on GitHub Medical informatics; Biocomputational method; Classification of bioinformatical subject; Neural networks; Machine learning