基于可能性C-均值的鲁棒多视角聚类算法

Robust multi-view clustering algorithm based on possibilistic C-means

目前多数多视角聚类算法不考虑噪声问题,为了更有效地分析含有噪声数据的聚簇结构,提出了一种基于可能性C-均值的鲁棒多视角聚类(PCM-RMVC)算法,该算法同时利用多个视角空间中的特征信息,最小化每个视角空间中数据对象与聚簇中心的距离.推导出数据隶属度和每个视角权重的迭代更新规则,设计出聚类过程的迭代算法.实验表明:PCM-RMVC算法对噪声具有较强的鲁棒性,并且聚类效果优于五种有代表性的多视角聚类算法.

Most exiting multi-view clustering algorithms ignore the problem of noise.In order to analysis the cluster structure of data which containing noise more efficiently,a robust multi-view clustering algorithm based on possibilistic C-means（PCM-RMVC）was proposed.The feature information in multi-view space was simultaneously incorporated and the distances between objects and cluster centers in each view space were minimized.The update rules for fuzzy memberships of objects and weights of feature spaces were derived,and then an iterative algorithm was designed for the clustering process.Experimental studies show that the PCM-RMVC algorithm is robust to noise and the qualities of clustering results of PCM-RMVC are superior to five representative algorithms.