Clustering Spherical Shells by a Mini-Max Information Algorithm
We focus on spherical shells clustering by a mini-max information (MMI) clustering algorithm based on mini-max optimization of mutual information (MI). The minimization optimization leads to a mass constrained deterministic annealing (DA) approach, which is independent of the choice of the initial data configuration and has the ability to avoid poor local optima. The maximization optimization provides a robust estimation of probability soft margin to phase out outliers. Furthermore, a novel cluster validity criteria is estimated to determine an optimal cluster number of spherical shells for a given set of data. The effectiveness of MMI algorithm for clustering spherical shells is demonstrated by experimental results.
KeywordsCluster Algorithm Mutual Information Spherical Shell Cluster Number Structural Risk Minimization
Unable to display preview. Download preview PDF.
- 2.Blahut, R.E.: Princinple and practice of information theory. Addison-Wesley, Reading (1987)Google Scholar
- 8.Rose, K.: Deterministic annealing for clustering, compression, classification, regression, and related optimization problems. Proc. of IEEE, 86, 2210–2239 (1998)Google Scholar