| 一种新的基于动态SOFM的神经网络聚类模型 |
| |
| 引用本文： | 陆宇旻,郭会林,李陶深,苏一丹.一种新的基于动态SOFM的神经网络聚类模型[J].广西师范大学学报(自然科学版),2007,25(4):257-260. |
| |
| 作者姓名： | 陆宇旻 郭会林 李陶深 苏一丹 |
| |
| 作者单位： | 广西大学,计算机与电子信息学院,广西,南宁,530004 |
| |
| 基金项目： | 广西自然科学基金资助项目(桂科自0624006) |
| |
| 摘 要： | 提出一种动态增删自组织映射(DGDSOM)神经网络,给出该模型的聚类算法描述及实现过程。定义了聚类节点信任度,并根据竞争结果、信任度、中心相似度,制定节点的增删策略,通过动态评价并变动节点,提升聚类效果。实际的聚类分析应用结果表明,该算法可以得到准确的聚类结果。
|
| 关 键 词： | DGDSOM 聚类 SOFM 节点信任度 |
| 文章编号： | 1001-6600（2007）04-0257-04 |
| 收稿时间： | 2007-04-25 |
| 修稿时间： | 2007年4月25日 |
New Neural Network Model Based on Dynamic SOFM |
| |
| LU Yu-min,GUO Hui-lin,LI Tao-shen,SU Yi-dan.New Neural Network Model Based on Dynamic SOFM[J].Journal of Guangxi Normal University(Natural Science Edition),2007,25(4):257-260. |
| |
| Authors: | LU Yu-min GUO Hui-lin LI Tao-shen SU Yi-dan |
| |
| Abstract: | A neural network model based on dynamic growth and delete self-organizing maps(DGDSOM) is presented as an extended version of the Self-Organizing Feature Maps(SOFM),which needs to preset the clustering number and tends to cluster fuzzily.This paper details the clustering algorithm and the implementation of the DGDSOM.The clustering node support degree is defined,and the strategy on growing and deleting nodes dynamically is formulated according to the competition result,support degree and center's similarity,which is helpful to improve the clustering results.The actual clustering analysis shows that this algorithm can get accurate clustering results. |
| |
| Keywords: | DGDSOM clustering SOFM node support degree |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
