ELBG implementation

Giuseppe Patanè and Marco Russo

Abstract

In this paper we describe the implementation of the ELBG, a clustering technique we developed as an improvement on the traditional LBG algorithm. It tries to solve the problem of the local minima deriving from a bad choice of the initial conditions. In a previous paper, we described in depth this technique and some points were highlighted. (a) It performs better than or equal to all of the other algorithms we considered. (b) The final result is virtually independent of the initial conditions. (c) No parameters have to be tuned manually (d) Fast convergence. (e) Low overhead with respect to the traditional LBG. The aim of this paper is to describe the particular solutions we adopted to obtain such results at the cost of a low overhead with respect to the traditional LBG.

Keywords: Clustering, Unsupervised Learning, GLA, LBG, ELBG

[MyKES00] G.Patanè and M.Russo, ELBG implementation. International Journal of Knowledge based Intelligent Engineering Systems, vol. 4 no. 2, pp 94-109, April 2000.

Download: [ps.gz] [pdf]