关于核函数选取的方法

On the method of Kernel-function selection in support vector machine

在支持向量机技术中,核函数选取的好坏直接影响支持向量机的性能.目前关于核函数的研究在理论和应用两方面均取得了一定的成果,但还未深入到足以指导核函数的选取.本文从混合核函数着手研究,建立若干选取规则,得到关于核函数选取的方法.采用平衡约束规划(MPEC)模型来优化选取参数,解决了参数的选取问题.

In the technology of Support Vector Machine （SVM）, it is important to select an optimal kernel function in order to enhance the performance of the SVM. At present, research in theory and application of kernel function has achieved some results, but not yet enough in depth to guide the selection of kernel function. In this paper, we give some regulations about how to select a kernel function by using the mixtures of kernels and a model of mathematical programming with equilibrium constraints for selecting the parameter to resolve the problem of how to select parameters.