自组织拓扑映射与主曲线学习  

Self-organized Topological Mapping and Principal Curve Learning

作　　者：倪劲松[1] 李玉珍[2] 王宜怀[2] 

机构地区：[1]苏州大学数学科学学院 [2]计算机科学与技术学院,苏州215006

出　　处：《计算机科学》2006年第3期151-154,共4页Computer Science

基　　金：江苏省教育厅自然科学基金(02KJD52001)

摘　　要：本文利用自组织拓扑映射方法设计了一种简易主曲线学习的算法,该算法继承了 HS 主曲线算法和 K 主曲线算法的主要优点.同时降低了一般主曲线算法的难度,使其变得更简洁明了。We use the method of self-organized topological mapping to design a learning algorithm of principal curves. The algorithm is composed by two parts. In the first part, based on the theory of generalized vector guantizer, we have achieved refined GL-algorithm, and use it to compute polygonal line principal curves for every finite discrete data set, which is like Kegl has done. In the second part, we combine the method of self-organized topological mapping with re fined GL-algorithm to design an algorithm of principal curves（Algorithm C）, which is simpler than HS-type and K- type. This algorithm has inherited main virtues of the principal curves of HS-type and K-type.

关 键 词：向量量化器 自组织拓扑映射 Voronoi邻域 主曲线 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TU452[自动化与计算机技术—计算机科学与技术]