基于欧式距离的一种新的核函数

支持向量机表现的好坏很大程度上取决于核函数的选取，因此最近几年关于核函数的研究有许多。越来越多的核函数也被提了出来!但是选取合适的核函数往往却不容易，因为数据的特征往往不知道。文中利用函数的Taylor展开思想，提出了一种新的核函数，叫T—KMOD，基于KMOD提出的。该核函数的灵活性更好，可以处理很多分类的问题。用网络入侵的数据对该核函数进行了仿真，从仿真的结果可以看出，和一些常用的核函数相比，它的鲁棒性更好，有更强的分类能力。同时该函数的分类效果更好。所以该核函数和一般常用的核函数相比，可能更具有一般选择性。

A New Kind of Kernel Function Based on Euclidean Distance

The behaviors of SVM largely depends on its adopted kernel function, so there are many researches on kernel function in recent years. More and more kernel functions are presented. It＇s very difficult to select a proper kernel function mentioned above,because the na- ture of the date is usually unknown. It presents a new kind of kernel function by the thought of function＇s Taylor expansion, called T- KMOD,based on KMOD. Its flexibility is better,so it may deal with lots of mapping problems. From the performance of T-KMOD in- vestigated on network intrusion dates,obtain that it is robust and has stronger mapping ability comparing with commonly applied kernel functions. And it can obtain better generalization performance as well. So the proposed kernel function may be served as a generic alterna tive for some commonly applied kernel functions.