...
  • 主题
高级搜索  >
历史搜索 清空历史
首页> 外文期刊>Soft Computing - A Fusion of Foundations, Methodologies and Applications >Analytical and numerical evaluation of the suppressed fuzzy c-means algorithm: a study on the competition in c-means clustering models
【24h】

Analytical and numerical evaluation of the suppressed fuzzy c-means algorithm: a study on the competition in c-means clustering models

机译：抑制型模糊c均值算法的分析和数值评估：c均值聚类模型的竞争研究

获取原文
获取原文并翻译 | 示例
           
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Suppressed fuzzy c-means (s-FCM) clustering was introduced in Fan et al. (Pattern Recogn Lett 24:1607–1612, 2003) with the intention of combining the higher speed of hard c-means (HCM) clustering with the better classification properties of fuzzy c-means (FCM) algorithm. The authors modified the FCM iteration to create a competition among clusters: lower degrees of memberships were diminished according to a previously set suppression rate, while the largest fuzzy membership grew by swallowing all the suppressed parts of the small ones. Suppressing the FCM algorithm was found successful in the terms of accuracy and working time, but the authors failed to answer a series of important questions. In this paper, we clarify the view upon the optimality and the competitive behavior of s-FCM via analytical computations and numerical analysis. A quasi competitive learning rate (QLR) is introduced first, in order to quantify the effect of suppression. As the investigation of s-FCM’s optimality did not provide a precise result, an alternative, optimally suppressed FCM (Os-FCM) algorithm is proposed as a hybridization of FCM and HCM. Both the suppressed and optimally suppressed FCM algorithms underwent the same analytical and numerical evaluations, their properties were analyzed using the QLR. We found the newly introduced Os-FCM algorithm quicker than s-FCM at any nontrivial suppression level. Os-FCM should also be favored because of its guaranteed optimality. Keywords Fuzzy c-means algorithm - Suppressed fuzzy c-means algorithm - Competitive clustering - Alternating optimization - Learning rate
机译：Fan等人介绍了抑制的模糊c均值（s-FCM）聚类。 （Pattern Recogn Lett 24：1607–1612，2003），其目的是将较高速度的硬C均值（HCM）聚类与模糊C均值（FCM）算法的更好分类特性相结合。作者修改了FCM迭代以在集群之间建立竞争：较低的隶属度根据先前设置的抑制率而降低，而最大的模糊隶属度则通过吞噬所有较小的被抑制部分而增长。在精度和工作时间方面抑制FCM算法是成功的，但是作者未能回答一系列重要问题。在本文中，我们通过分析计算和数值分析阐明了关于s-FCM的最优性和竞争行为的观点。首先引入准竞争学习率（QLR），以量化抑制的效果。由于对s-FCM最优性的研究无法提供精确的结果，因此提出了另一种最优抑制的FCM（Os-FCM）算法作为FCM和HCM的混合体。抑制的和最优抑制的FCM算法都经过了相同的分析和数值评估，并使用QLR分析了它们的性质。我们发现，在任何非平凡的抑制级别上，新引入的Os-FCM算法都比s-FCM更快。 Os-FCM还应确保其最优性，因此受到青睐。关键词模糊c均值算法抑制模糊c均值算法竞争聚类交替优化学习率

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