随着信息获取渠道的增多及测量手段的不断提高，人们可以通过多角度信息对研究对象进行描述。多视角数据在带来更丰富，更完备信息的同时，也伴随着数据结构性差异大，存储运算复杂等问题。合理地利用多视角信息实现精准化、高效化的数据聚类是多源数据融合技术发展中需要迫切解决的基础性问题之一。本项目将采用理论、算法、应用三个方面来研究多视角聚类中存在的问题，拟研究以下三个方向：（1）基于多视角数据结构信息保真，研究基于正则化联合矩阵/张量分解的多视角聚类数学模型及其适定性问题；（2）基于分而治之的分裂思想，研究适用于大规模多视角数据的高性能聚类算法及其收敛特性；（3）基于理论分析和数值模拟的相互验证，研究多视角聚类模型中自主化参数优选机制。上述工作基础上，将进一步研制多视角数据聚类Matlab测试软件包，发布给学术界做进一步研究之用。

With the rapid increase of information and the improvement of measurements, many research objects can be described by information from multiple perspectives. Although the multi-view data can bring us richer and more comprehensive information, it is together with the heterogeneity of data structure and large storage and computation complexity. How to utilize the multi-view information for more precise and efficient data clustering is one of the fundamental problems in the development of multi-source data fusion technologies. With the combination of therotical researches, algorithm studies, and practical applications, this project will tackle the existing theoretical and application problems with multi-view clustering, mainly aiming at addressing the following three aspests: (1) based on preserving the multi-view data structure, we will study the regularized coupled matrix/tensor decomposition based model for multi-view clustering and its well-posed properities; (2) based on the splitting philosophy, we will study the practical high-performance algorithms for multi-view data clustering and its convergence properities; (3) based on the mutual verification of both theoretical analysis and numerical simulations, we will study the autonomous parameters optimization mechanism for multi-view clustering model. On the basis of the aforementioned works, we will further develop the Matlab package for multi-view clustering. The successful implementation of this project will be important for both academical and practical developments in the field of computational mathematics as well as computer science.