Logical Symmetry Based K-means Algorithm with Self-adaptive Distance Metric

摘要

In this paper, we propose a modified version of the K-means clustering algorithm with distance metric. The proposed algorithm adopts a novel weighted Euclidean distance measure based on the idea of logical symmetry of points to its candidate clusters, which challenges the common assumption that the point similarity can only be determined by their physical distance to the cen-troids of the clusters. This kind of logical symmetry distance can be adaptively applied to many practical data clustering scenarios such as social network analysis and computer vision, in which the logical relationship of the clustering objectives is an important consideration in the design of the clustering algorithm. Several data sets are used to illustrate its effectiveness.