Multi-view embedded clustering with unsupervised trace ratio LDA

In many real applications of machine learning and data mining, we are often confronted with high-dimensional data represented by heterogeneous features or views, which describe different perspectives of the data. Efficiently clustering such data is a challenge. To address this problem, we propose a unified and embedded framework referred to as multi-view embedded clustering with trace ratio (MECTR), which performs dimensionality reduction and clustering simultaneously, and adaptively controls the interactions among different views at the same time. Within this framework, we are able not only to obtain multiple discriminative subspaces synchronously, but also keep the clustering results consistent among different views. We also develop an alternate iterative optimization strategy to learn the common clustering indicator, multiple discriminative subspaces and weights for heterogeneous features with convergence. Comprehensive experiments on synthesis dataset and three benchmark datasets demonstrate the superiority of the proposed work.