Using Smoothed Data Histograms for Cluster Visualization in Self-Organizing Maps

Several methods to visualize clusters in high-dimensional data sets using<br> the Self-Organizing Map (SOM) have been proposed. However, most of these<br> methods only focus on the information extracted from the model vectors of<br> the SOM.<br> This paper introduces a novel method to visualize the clusters of a SOM<br> based on data histograms smoothened by ranking the membership of a data<br> item to a unit according to its distance. The method is illustrated using a<br> simple 2-dimensional data set and similarities to other SOM based<br> visualizations and to the posterior probability distribution of the<br> Generative Topographic Mapping are discussed. Furthermore, the method is<br> evaluated on a real world data set consisting of pieces of music.<br> <br>