Extending the Self-Organizing Maps Algorithm with Using Gaussian Kernel

Abstract

SOM algorithm is used for high dimensional data visualization. In other words, this algorithm has map high dimensional data to low dimensions space. However, basic SOM algorithm due to using Euclidean distance has bugs in resolving real issues. Paper main aim is the extension of SOM algorithm.  For this purpose in Euclidean distance calculation phase, Gaussian kernel used. For measure the accuracy and quality of clustering, F1 measure is used. With examination results as observed, Kernel SOM algorithm have higher F1 than the basic SOM algorithm.