Clustering Using Dynamic Growing Hierarchical Self - Organizing Map with Improved Lm Learning

Dr. D. Hari Prasad Professor, Department of Computer Applications, SNR Sons College, Coimbatore Abstract A primary drawback of the traditional SOM is that the size of the mapis fixed and the number of neurons in the map should be determined a priori. This might not be feasible for some applications and result in a significant limitation on the final mapping. Several dynamic SOM models have been proposed recently to reduce the limitations of the fixed network architecture of the traditional SOM. These models rely on an adaptive architecture where neurons and connections are inserted into or removed from the map during their learning process according to the particular requirements of the input data. A major variation of traditional SOM model is proposed in this paper.