首页> 外文会议>International conference on cognitive science;ICCS 2001 >A New Clustering Algorithm Based On The Relative Gravitation
【24h】

A New Clustering Algorithm Based On The Relative Gravitation

机译:基于相对引力的新聚类算法

获取原文

摘要

This paper presents a new clustering algorithm called G-CLUS based on the relative gravitation. It was motivated from the universal gravitation. Although the law of universal gravitation is the basic formula related to the formation of universe, it has some defects in the aspects of data clustering, we defined a new gravitation force to weaken the attraction among large clusters as follows: RGF=F/m_1~2=G(m_2/r~2m_1), F=G(m_2m_1/r~2) ,where m_1 is the mass of instance considered now, m_2 is the mass of an opponent instance and G is gravitational constant. The instance naturally moves to the direction of attraction. We can decide the direction and magnitude for a given instance by the resultant force coming from all the instances around it. After instances are moved by the attraction, in case the distance is close enough, we can merge them into one instance preserving their history.
机译:本文提出了一种基于相对引力的新聚类算法G-CLUS。它是由万有引力驱动的。尽管万有引力定律是与宇宙形成有关的基本公式,但它在数据聚类方面存在一些缺陷,我们定义了一种新的引力来减弱大星团之间的吸引力,如下所示:RGF = F / m_1〜 2 = G(m_2 / r〜2m_1),F = G(m_2m_1 / r〜2),其中m_1是现在考虑的实例质量,m_2是对手实例的质量,G是重力常数。实例自然向吸引方向移动。我们可以通过来自周围所有实例的合力来确定给定实例的方向和大小。在景点移动实例之后,如果距离足够近,我们可以将它们合并为一个实例,以保留其历史记录。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号