核函数的选择研究综述

核方法是解决非线性模式分析问题的一种有效方法,是当前机器学习领域的一个研究热点.核函数是影响核方法性能的关键因素,以支持向量机作为核函数的载体,从核函数的构造、核函数中参数的选择、多核学习3个角度对核函数的选择的研究现状及其进展情况进行了系统地概述,并指出根据特定应用领域选择核函数、设计有效的核函数度量标准和拓宽核函数选择的研究范围是其中3个值得进一步研究的方向.

Survey of research on kernel selection

Kernel method is an effective approach for non-linear pattern analysis problems and is also a new research focus in current machine learning community.Kernel function is a key factor to achieve good performance of kernel methods.The research state and advances in kernel selection for support vector machine(SVM) are systematically surveyed from three aspects including kernel construction,kernel parameter selection and multiple kernel learning(MKL).Furthermore,three future research directions,i.e.,how to select an appropriate kernel in practice,how to design an effective kernel evaluation criterion and extend the research scope of kernel selection,are summarized.