基于互近邻一致性的近邻传播算法

摘要

Affinity propagation is a new clustering method. It based on the similarity between pairs of data points, through the exchange of information between data points, and finally obtained the final clustering results. This paper presented an im-proved AP clustering based on K-mutual nearest neighbor consistency KMNC-AP. The improved algorithm used the idea of K-mutual nearest neighbor consistency to adjust the similarity between data points. Experiments show that the improved algorithm is more accurate and faster than the original algorithm.%近邻传播(AP)算法是一种新提出的聚类算法,是在数据点的相似度矩阵的基础上进行聚类,通过数据点之间交换信息,最后得到聚类结果.提出了基于互近邻一致性近邻传播算法,即KMNC-AP算法,该算法利用互近邻一致性调整数据点之问的相似度,进而提高聚类效率和精确度.实验结果表明,该算法在处理能力和运算速度上优于原算法.