Adaptive Graph Regularized Concept Factorization With Label Discrimination for Data Clustering

Abstract

The semi-supervised non-negative matrix factorization (NMF) methods have demonstrated their powerful capabilities in fields such as data representation, image clustering, and recommendation systems. However, the semi-supervised NMF methods still have various shortcomings and there is still room for improvement. In this paper, based on concept factorization (CF), a new semi-supervised concept factorization method, called adaptive graph regularized concept factorization with label discrimination (AGCFLD), is proposed. In AGCFLD, the partial label information is embedded into the framework of CF by label discrimination constraint and adaptive graph regularization. Thus, the discriminative abilities of data representations generated by AGCFLD are enhanced in the clustering tasks. Experimental results on several datasets demonstrate the effectiveness of the proposed AGCFLD method compared to the state-of-the-art methods. For the convenience of reproducing the results of this article, the source code can be found on the website: https://github.com/YanfengLi-spec/AGCFLD.