Computing Sammon´s Projection of Social Networks by Differential Evolution

Visualization of complex real-world data is an essential part of network processing. Complex high-dimensional or networked data ought to be presented in a form suitable for machine and human analysis. Therefore, advanced methods of dimension reduction or projection to low-dimensional spaces are investigated. In this work we use Differential Evolution as a real-parameter optimization metaheuristic algorithm to minimize the error function used in Sammon´s projection and compare its results with the results obtained by a traditional heuristic algorithm for Sammon´s projection. The metaheuristic algorithm achieves lower projection error and its results are demonstrated on a 2D visualization of real-world data from the domain of social networks.