An Adaptive Deep Metric Learning Loss Function for Class-Imbalance Learning via Intraclass Diversity and Interclass Distillation

• Published in: IEEE transaction on neural networks and learning systems 2024-11, Vol.35 (11), p.15372-15386
• Main Authors: Du, Jie, Zhang, Xiaoci, Liu, Peng, Vong, Chi-Man, Wang, Tianfu
• Format: Article
• Language: English
• Subjects: Class-imbalance learning (CIL); Correlation; deep metric learning (DML); diverse and discriminant feature; Face recognition; Feature extraction; Learning systems; loss function with adaptive weights; semantic correlations between classes; Semantics; Task analysis; Ultrasonic imaging
• ISSN: 2162-237X; 2162-2388; 2162-2388
• DOI: 10.1109/TNNLS.2023.3286484

Deep metric learning (DML) has been widely applied in various tasks (e.g., medical diagnosis and face recognition) due to the effective extraction of discriminant features via reducing data overlapping. However, in practice, these tasks also easily suffer from two class-imbalance learning (CIL) problems: data scarcity and data density, causing misclassification. Existing DML losses rarely consider these two issues, while CIL losses cannot reduce data overlapping and data density. In fact, it is a great challenge for a loss function to mitigate the impact of these three issues simultaneously, which is the objective of our proposed intraclass diversity and interclass distillation (IDID) loss with adaptive weight in this article. IDID-loss generates diverse features within classes regardless of the class sample size (to alleviate the issues of data scarcity and data density) and simultaneously preserves the semantic correlations between classes using learnable similarity when pushing different classes away from each other (to reduce overlapping). In summary, our IDID-loss provides three advantages: 1) it can simultaneously mitigate all the three issues while DML and CIL losses cannot; 2) it generates more diverse and discriminant feature representations with higher generalization ability, compared with DML losses; and 3) it provides a larger improvement on the classes of data scarcity and density with a smaller sacrifice on easy class accuracy, compared with CIL losses. Experimental results on seven public real-world datasets show that our IDID-loss achieves the best performances in terms of G-mean, F1-score, and accuracy when compared with both state-of-the-art (SOTA) DML and CIL losses. In addition, it gets rid of the time-consuming fine-tuning process over the hyperparameters of loss function.