New self-organizing algorithms for topological maps

Abstract

Self-organizing Maps (SOM) were developed by Kohonen in late 1980's in response to the growing amount of large data sets required to represent a problem for thorough analysis with standard tools and techniques. SOMs are widely used in many applications including machine vision and image analysis, speech analysis and recognition, robotics, telecommunications, signal processing and radar measurement, process control, and so on. Despite their popularity in engineering applications and statistical data analysis, the basic SOM and its variants possess some intrinsic weaknesses that limit their dominance in traditional and contemporary learning tasks such as clustering, classification and density estimation.