Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
536342 | Pattern Recognition Letters | 2005 | 12 Pages |
Abstract
To overcome the limitation of requiring the cluster threshold with the parametric approach, this paper presents a clustering constraint which first considers an estimate of the global distribution. The clustering process moves from local clusters identifying the data globally to larger clusters with a specified density function. Merging then occurs to provide a statistically supported representation of the data. A hashing-based sequential clustering algorithm is introduced which utilizes the initial and merging constraints. Experimental data shows the methods effectiveness at classifying varying cluster shapes and sizes when compared to recent clustering techniques.
Related Topics
Physical Sciences and Engineering
Computer Science
Computer Vision and Pattern Recognition
Authors
José J. Amador,