A Sequential Algorithm for Training the SOM Prototypes Based on Higher-Order Recursive Equations.pdfVIP

Hindawi Publishing Corporation Advances in Artificial Neural Systems Volume 2010, Article ID 142540, 10 pages doi:10.1155/2010/142540 Research Article A Sequential Algorithm for Training the SOM Prototypes Based on Higher-Order Recursive Equations Mauro Tucci and Marco Raugi Department of Electrical Systems and Automation, University of Pisa, via Diotisalvi 2, 56126 Pisa, Italy Correspondence should be addressed to Mauro Tucci, mauro.tucci@dsea.unipi.it Received 29 July 2010; Accepted 27 November 2010 Academic Editor: Songcan Chen Copyright © 2010 M. Tucci and M. Raugi. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. A novel training algorithm is proposed for the formation of Self-Organizing Maps (SOM). In the proposed model, the weights are updated incrementally by using a higher-order difference equation, which implements a low-pass digital filter. It is possible to improve selected features of the self-organization process with respect to the basic SOM by suitably designing the filter. Moreover, from this model, new visualization tools can be derived for cluster visualization and for monitoring the quality of the map. 1. Introduction matching with the input. The incremental learning algorithm of the SOM has been established heuristically following this The self-organizing map, with its variants, is one of the paradigm, where the matching of the models with the input most popular neural network algorithms in the unsupervised is increased by moving all the vectors in the direction of the learning category