一种AP算法的改进:M-AP聚类算法

Affinity Propagation(AP)聚类算法将所有数据点作为潜在的聚类中心,在相似度矩阵的基础上通过消息传递进行聚类.与传统聚类方法相比,对于大规模数据集,AP是一种快速、有效的聚类方法.但是AP算法在聚类结构复杂的(非团状)数据集上得到的效果并不是很好.因此,在AP的基础上加入一个merge过程,将AP算法改进为M-AP算法,可以有效地解决这种问题.而当样本数目比较大时,将CVM压缩算法融入其中,可以有效地解决大样本问题.

Improved AP Algorithm:M-AP Clustering Algorithm

Affinity propagation(AP) clustering simultaneously considers all data points as potential exemplars.It takes similarity between pairs of data points as input measures,and clusters gradually during the message-passing procedure.But the result of AP clustering algorithm in the data set of complex structure(non-group)is not very good.Therefore,we proposed a new clustering algorithm by adding a merge process on the basis of AP clustering algorithm,called M-AP algorithm which can effectively solve this kind of problem.When the number of samples is very large, the problem of large sample can be effectively solved by using CVM compression algorithm.