Soon: self organising oscillator networks for use in clustering problems

Abstract

The self-organising oscillator network (SOON) is a comparatively new clustering algorithm [H.F.M.B.H. Rhouma, February 2001], that has received relatively little attention so far. The SOON is distance based, meaning that clustering behaviour is different in a number of ways that can be beneficial. This paper examines the effect of adjusting the control parameters of the SOON with two widely different datasets which represent two different types of real-world data; the first is a communications signal dataset representing one modulation scheme under a variety of noise conditions. The second is a biological dataset taken from microarray experiments on the cell-cycle of yeast. The modulation scheme data is relatively easy to cluster at high SNR, however at lower SNR, the clustering problem becomes much more difficult as the separation between the cluster reduces. The paper demonstrates that the SOON is a viable tool to analyse these problems, and can add many useful insights to the data, that may not always be available using other clustering methods