Comparison of competitive learning for SOM used in classification of partial discharge

Abstract

This paper shows different competitive learning algorithms for Self Organizing Map (SOM) and are experimentally compared, the characterization of the obtainable results in terms of quality of SOM. The competitive learning algorithms showed to SOM algorithm are Winner-takes-all, Frequency Sensitive Competitive Learning and Rival Penalized Competitive Learning. As a case study: the performance in classification of partial discharge on power cables. © 2012 Springer-Verlag.