Multi-view multi-label learning with high-order label correlation

Multi-label learning deals with a kind of problem that the given samples areassociated with multiple labels simultaneously. Recently, multi-label learning has become a popular topic in the literatures of machine learning and has attracted lots of researches. In this paper, we propose a new multi-view multi-label learning method by considering the label correlation, which is called ELSMML. Based on the high-order strategy, we construct a crafted label correlation matrix to describe the relationships among labels. We further utilize multi-view learning and dimension reduction to exploit the high-level latent semantic label information and the latent feature information, so as to build a classifier in the low dimensional space. In addition, we apply manifold regularization terms to make the data samples in the low dimensional space have the same intrinsic structure as the original data. After that, we put forward the accelerated proximal gradient method to optimize the ELSMML model and obtain thepredictive classifier. Besides, we conduct convergence analysis and computational complexity analysis for ELSMML method. In the experiments, the ELSMML method can achieve better performance on the evaluation metrics compared with other baselines. (c) 2022 Elsevier Inc. All rights reserved.

Automated summary

Multi-label learning is a popular topic in machine learning that deals with the simultaneous association of multiple labels with given samples. This paper proposes a new multi-view multi-label learning method called ELSMML, which considers label correlation. The method constructs a crafted label correlation matrix to describe label relationships and utilizes multi-view learning and dimension reduction to exploit latent semantic label information and feature information, building a classifier in a low dimensional space. The ELSMML model is optimized using the accelerated proximal gradient method and achieves better performance compared to other baselines according to evaluation metrics.