Cluster Analysis Via Dynamic Self-organizing Neural Networks
The paper presents dynamic self-organizing neural networks with one-dimensional neighbourhood that can be efficiently applied to complex, multidimensional cluster-analysis problems. The proposed networks in the course of learning are able to disconnect their neuron chains into sub-chains, to reconnect some of the sub-chains again, and to dynamically adjust the overall number of neurons in the system; all of that – to fit in the best way the structures “encoded” in data sets. The operation of the proposed technique has been illustrated by means of three synthetic data sets, and then, this technique has been tested with the use of two real-life, complex and multidimensional data sets (Optical Recognition of Handwritten Digits Database and Image Segmentation Database of Statlog Databases) available from the ftp-server of the University of California at Irvine (ftp.ics.uci.edu).
KeywordsCluster Distribution Kohonen Network Epoch Number Additional Neuron Optical Recognition
Unable to display preview. Download preview PDF.
- 3.Gorzałczany M.B., Rudziński F.: Application of dynamic self-organizing neural networks to WWW-document clustering, International Journal of Information Technology and Intelligent Computing (to appear); Rutkowski, L., Tadeusiewicz, R., Zadeh, L.A., Żurada, J.M. (eds.): ICAISC 2006. LNCS (LNAI), vol. 4029. Springer, Heidelberg (2006)Google Scholar
- 5.Machine Learning Database Repository, University of California at Irvine, ftp.ics.uci.eduGoogle Scholar