核函数的选取是SVM分类器选取的核心问题.核函数的自动选取既可以提高分类器的性能,又可以减少人为的干预.因此如何自动选取核函数已经成为SVM的热点问题,但是这个问题并没有获得很好的解决.近年来对核函数参数的自动选取的研究,特别是对基于梯度的优化算法的研究取得了一定的进展.提出了一种基于梯度的核函数选取的通用算法,并进行了实验. The selection of kernel function is the core problem of SVM classifier selection.It can not only improve the performance of classifier but also reduce the human intervention.How to choose kernel function automatically has become a hot issue of SVM, Has not been well solved.In recent years, the research on the automatic selection of kernel function parameters, especially the research on gradient-based optimization algorithms, has made some progress.A general algorithm based on gradient-based kernel function selection conducted an experiment.