基于核函数的构造型网络二分覆盖算法

A Kernel-based Classification Algorithm of Binary-Covering Approach in Constructive Neural Networks

该文首先介绍了构造型神经网络中的覆盖算法的特点和性质,以及与支持向量机(SVM)中的核函数法的关系。然后,根据这些性质将两者有机结合起来,提出了一种处理多类问题分类的基于核函数的二分覆盖分类算法(Kf-BCC)。仿真结果表明,这种方法可以达到很好的效果,也为多类问题的分类提供了一种有效途径。

This paper presentes the characteristics,properties and relationship of Covering Algorithms and the kernel function of Support Vector Machines(SVM)firstly.By combining SVM and Covering Algorithm,the paper proposes a kernel -based classification algorithm of binary -covering approach in constructive neural networks(Kf -BCC),which concerns the problems of multi-class classification.The result of experiment shows that it is efficient and one of the right ways of multi-class classification.