Centroid neural network with Bhattacharyya Kernel for GPDF data clustering

Abstract

A clustering algorithm for GPDF data called Centroid Neural Network with Bhattacharyya Kernel (BK-CNN) is proposed in this paper. The proposed BK-CNN is based on the unsupervised competitive centroid neural network (CNN) and employs a kernel method for data projection. In order to cluster the GPDF data, the Bhattacharyya kernel is used to measure the distance between two probability distributions for data projection. When applied to GPDF data in an image classification model, the experiment results show that the proposed BKCNN algorithm is more efficient than other conventional algorithms such as k-means algorithm, SOM and CNN with Bhattacharyya distance. © Springer-Verlag Berlin Heidelberg 2007.