摘要
提出了一种基于核化技术的模糊核超球感知器分类算法,该算法通过核化技术把样本数据映射到高维特征空间,并利用超球感知器学习寻找高维特征空间的决策超球,从而得到各类样本的决策函数.同时,样本测试中采用的模糊技术有效提高了算法的适应性.该算法学习规则简单,所得特征空间超球在样本空间的分布能很好地反映样本的数据结构,适用于不同类型数据结构样本的学习,并经大量试验显示了算法的有效性.
In this paper the novel fuzzy kernel hyper-ball perceptron based on Mercer kernels is presented. The proposed method first maps the input data into a high-dimensional feature space using some Mercer kernel functions. Then, the decision functions for each group of data are derived by the learning rules of the fuzzy kernel hyper-ball perceptron. The fuzzy memberships are incorporated into its classification algorithm to further improve the perceptron's adaptability and classification accuracy. The experimental results demonstrate the effectiveness and advantage of this new classification algorithm based on the fuzzy kernel hyper-ball perceptron.
出处
《江南大学学报(自然科学版)》
CAS
2004年第3期221-226,共6页
Joural of Jiangnan University (Natural Science Edition)
基金
江苏省自然科学基金项目(BK2003017)资助课题.
关键词
核函数
感知器
超球
分类
kernel function
perceptron
hyper-ball
classification