Clustering by Self-Updating Process

Abstract

Clustering by Self-Updating Process Ting-Li Chen1 Institute of Statistical Science, Academia Sinica, Taiwan R.O.C. ? We introduce a simple, intuitive, yet powerful algorithm for clustering analysis. This algorithm stands from the viewpoint of elements to be clustered, and simulates the process of how they perform self-clustering. The algorithm is therefore named Self-Updating Process (SUP). We discover the algorithm's ability to simultaneously isolate noise while performing clustering, which enables the algorithm to produce good clustering results even when the level of noise in the data is high. We present simulation studies to demonstrate the performance of this algorithm. Applications to gene expression data and image segmentation are provided.