SEEDING METHOD FOR K-MEANS CLUSTERING AND OTHER CLUSTERING ALGORITHMS

摘要

Techniques for determining a collection of seeds for use in a clustering algorithm. After one or more seeds are selected by any of various methods, additional seeds are selected by randomly selecting a point from a set of data points with the probability of selecting a given point being proportional to the distance from the given point to the nearest previously selected seed. An abundant number of seeds may be initially selected, and the collection of seeds culled by an iterative process.