Visualizing multivariate network on the surface of a sphere

A multivariate network is a graph whose nodes contain multi-dimensional attributes. We propose a method to visualize such a network using spherical Self-Organizing Map (SOM) and circular layout. The spherical SOM produced an initial graph layout by grouping nodes with similar attributes to adjacent areas on the sphere. The circular layout algorithm fine tunes the position of each node to avoid node-node/node-edge overlap and minimize the number of edge crossings. We also implement an interface to project the spherical image onto a 2D plane such that the user can have a global view of the data set. The user is allowed to select any point of interest to be the center of the 2D projection.