摘要
本文首次提出最小一乘聚类中心的数学模型,对于不可微目标函数推导了最优解的条件,同时构造了最小一乘聚类中心的逼近点列,给出了相应的算法,并研究了一个不带类别标记的两类分类问题。最后讨论了最小一乘聚类中心能抵御样本中噪声的数学机理,并通过例子说明了此方法的有效性。
In this paper,we first put forward the mathmatical model of the least norm cluster center and deduced the condiction of the optimal solution to the corresponding nondifferential goal function, and gave an approximating point sequence to the optimal solution, we still discussed the geometric explanation that the least norm cluster center can oppose the sample errors. Finally we studied the classification problem of two classes with non-marked sample points.
出处
《计算机科学》
CSCD
北大核心
2008年第7期195-196,共2页
Computer Science
关键词
最小一乘方法
聚类中心
不可微函数
Least norm method, Cluster center, Nondifferential function
