Affinity Propagation聚类算法的扩展及改进研究
首发时间:2010-01-04
摘要:当前在有限区域内分布的稀疏不均的、具有一定分布结构的海量数据点集的聚类分析是一个尚未圆满解决的难题,针对Affinity Propagation (AP)聚类算法的不足之处,提出了两个扩展及改进型的聚类算法。在这两个算法中,基于“单元网格”的AP聚类算法是在“单元网格”层次上对AP聚类算法的一种扩展,而基于近邻点集的AP聚类算法是试图在时间上对AP聚类算法做一些改进。为实现这两个算法介绍了“单元网格”、近邻点集等几个比较重要的概念及思想方法。最后给出了一点颇有价值的研究方向。
关键词: 相似性度量 Affinity Propagation 单元网格 近邻点集
For information in English, please click here
An Expanded and Improved Research for Affinity Propagation
Abstract:It is still an unsatisfactorily resolved problem clustering of the massive data points set in limited areas with an uneven and sparse distribution. Facing to the shortcomings of Affinity Propagation(AP) clustering algorithm, it presents two expansion and improved algorithms. In these two algorithms, the AP clustering algorithm based on the “grid cell” is an extension of the AP on the level of “grid cell”, while the AP clustering algorithm based on near neighbor point set is try to make some improvements in time complexity. To implement these two algorithms, it introduced several important concepts and thinking ways, such as “grid cell”, near neighbor point set, and so on. Finally, a valuable research direction is presented.
Keywords: Dissimilarity Measure Affinity Propagation Grid Cell Near Neighbour Point Set
基金:
论文图表:
引用
No.3842822051126259****
同行评议
共计0人参与
勘误表
Affinity Propagation聚类算法的扩展及改进研究
评论
全部评论0/1000