中文摘要：

核方法在机器学习、数据挖掘等领域取得了重要的应用,而核函数及其参数的选择直接决定着核方法的预期效果,是核方法研究及应用中不可回避的困难问题。分析评述已有核选择研究成果是兼具理论价值和实践意义的重要研究内容。文章从新的核函数构造、单个核函数的参数优化以及多核学习算法三个方面分析了核选择研究的成果,以期为核参数选择提供技术指导。此外,还指出了核选择研究的几个有价值的深入方向。

英文摘要：

Kernel methods have recently gained a significant attention in machine learning, data mining and other fields. The performance of kernel methods greatly depends on the selection of kernel function and its parameters. However,the selection of the right kernel function and proper setting of its parameters is a challenging problem. Taxonomy of kernel selection methods is given, and it is of great theoretical value and practical significance in the usage of kernel methods. The existing kernel selection methods are identified and explained,and they are divided into three major categories：the construction of new kernel, the optimi- zation of a signal kernel function,and multiple kernel learning algorithms. Finally, some valuable develop- ment orientation on kernel selection are proposed.