基于奇异值分解双支持矩阵分类机

Twin Support Matrix Classification Machine Based on Singular Value Decomposition

针对以矩阵为输入的分类问题,在多秩多线性双支持矩阵分类机的基础上,建立了一个基于奇异值分解双支持矩阵分类机.对于矩阵输入,基于矩阵奇异值分解定义了一个矩阵映射函数,用来处理矩阵输入,降低数据维数并形成一个新的训练集.通过学习新的训练集,分类正确率将会升高,训练时间将会减少.对五组数据集进行训练,通过与其他分类方法相比,基于奇异值分解双支持矩阵分类机是一个有效的分类器.

Currently,the tensor,a form commonly seen,finds increasingly wider application in various kinds of fields.Matrix,as a second-order tensor,can be employed to bridge between a vector and a tensor.High order tensor can also be unfolded into matrix formulation.So,it is of vital significance to research into matrix-input-based classification problems.For matrix-input-based classification problems,based on multi-rank multi-linear twin support matrix classification machine,a twinsupport matrix classification machine is built on the basis of singular value decomposition.A matrix projecting function is defined to handle matrix input on basis of matrix singular value decomposition,reducing the dimensions of matrix input and reformulating a new training set.By learning the new training set,the classification accuracy improves and the training time decreases.Five matrix data sets are then subjected to training and compared with other classification methods,the twin support matrix classification machine based on singular value decomposition is found to be an efficient classification machine.