A novel kohonen self-organizing maps using exponential decay average rate of change for color clustering

Abstract

Kohonen Self-Organizing Maps (KSOM) is the most preferred and commonly used algorithm for clustering high-dimensional or multi-dimensional data. Its success in various domains are undeniable but still, it suffers issues on its convergence which is directly affected by its learning rate decay function. This study adopted Average Rate of Change to modify the learning rate decay function of the algorithm and proposes a new Exponential Decay Average Rate of Change (EDARC) as the modified learning rate decay function of the Enhanced Kohonen Self-Organizing Maps (EKSOM) to address its issue on convergence. The enhanced algorithm and the conventional KSOM was applied to color clustering. The results show that EKSOM outperformed the clustering capability of the conventional KSOM algorithm.