Label distribution learning with correlation information
Yilin Wu, Yaojin Lin, Wenzhong Guo, Weiping Ding · Engineering Applications of Artificial Intelligence · 2025
Label distribution learning quantifies the label space for each instance and has broad applicability in various fields. However, most existing works primarily focus on label correlation, but they have a deficiency in capturing instance correlation. Meanwhile, traditional label distribution learning operates on individual labels sequentially, which restricts the potential application of common features. Therefore, in this paper, we propose a novel approach for label distribution learning with correlation information, i.e. , instance correlation and label correlation. Specifically, the instance correlation and label correlation are identified by an optimization function and Pearson correlation coefficient, respectively. Besides, we conduct ℓ 2 , 1 regularization to exploit common features. Comprehensive experiments conducted on twelve publicly datasets demonstrate the effectiveness of our proposed approach against other well-established label distribution learning algorithms.