关键词: 仿射传播, 聚类, 大数据, 多代表点

Abstract: The traditional affinity propagation (AP) is difficult to handle with multi-exemplar clustering, and has large space and time complexity, so it is not suitable for large data sets. To address these problems, this paper proposes a multi-exemplar affinity propagation using fast reduced set density estimator (FRSMEAP). At the beginning of clustering, fast reduced set density estimator (FRSDE) is introduced to preprocess large-scale data sets, and then the condensed set fully representing sample properties can be obtained. Multi-exemplar affinity propagation (MEAP) algorithm is used to cluster the condensed set, which can find better decision boundaries than AP. So the accuracy of clustering is improved. In order to get the final data partition, the K-nearest neighbor (KNN) is used to assign the remained data. The simulation results on synthetic and standard data sets show that the proposed algorithm can not only cluster on large-scale data sets, but also has the advantage of high precision and fast speed.

Key words: affinity propagation, clustering, large data sets, multi-exemplar