High Dimensional Data Visualization Using Self Organizing Maps

A Self-organizing map is a non-linear unsupervised neural network that is used for data clustering and visualization of high-dimensional data. A Self-organizing map uses U-matrix to visualize the high-dimensional data and the distances between neurons on the map. However the structure of clusters and their shapes are often distorted. For better visualization of high-dimensional data a new approach high dimensional data visualization Self-organizing map (HVSOM) is explained. The HVSOM preserve the inter-neuron distance and better visualizes the differences between the clusters. In HVSOM the distances between input data points on the map resemble same those in the original space.