Summary
International Symposium on Nonlinear Theory and its Applications
2009
Session Number:A1L-B
Session:
Number:A1L-B2
An Experimental Evaluation of Clustering Algorithm using Attractor Selection
Naoko Uemura, Gen Nishikawa , Fukuhito Ooshita , Hirotsugu Kakugawa , Toshimitsu Masuzawa ,
pp.-
Publication Date:2009/10/18
Online ISSN:2188-5079
Summary:
Clustering is a problem to partition a network into small groups of nodes called clusters. Clustering is useful for maintaining large-scale networks because it gives hierarchical structure to the networks. Nishikawa et al. proposed a clustering algorithm based on attractor selection, which is one of the biologically inspired approaches to attain adaptability to dynamics of environments, and they showed the effectiveness of the algorithm by simulations. However, since the simulation settings are based on a theoretical model, the effectiveness in real environments is not evaluated. In this paper, we evaluate Nishikawa’s algorithm in realistic environments. First, we implement Nishikawa’s algorithm in a real sensor network composed of 36 sensor devices SunSPOTs and show that the algorithm is efficient in the environment. Second, we determine realistic simulation settings by analyzing the behavior of the above-mentioned real sensor network. Using the simulation settings, we show that Nishikawa’s algorithm is efficient in large-scale sensor networks.