多视角核K-means聚类算法的收敛性证明

摘要

The Zangwill convergence theorem was utilized to analyze the convergence of the multi-view kernel K-means (MVKKM).The study results showed that,under certain conditions,the iterative sequence generated by MVKKM converges,or there existed at least one subsequence converged to a local minimum or a saddle point of the objective function of the algorithm.And in Matlab,the convergence of the algorithm under different views and different index weight was verified.%用Zangwill收敛性定理对多视角核K-means(MVKKM)的收敛性进行了分析.结果表明,当满足一定的条件时,MVKKM生成的迭代序列收敛或至少存在一个子序列收敛于算法目标函数的局部极小值或鞍点,并在Matlab环境下,通过实验验证了算法在不同视角和不同的权重指数下的收敛性.