Visualizing the evolution of genetic algorithm search processes

The paper discusses how visualization techniques can facilitate the development of GA-systems. It demonstrates how visualization techniques can be used for the analysis of search space coverage, of convergence behavior, and of the topology of the explored search space. We describe the features of a GA-visualization environment that uses quadcodes to generate search space coverage maps, that employs 2D-distance maps to visualize convergence, and uses contour maps to visualize fitness. We also describe how these maps are generated. Moreover, we discuss how movies are employed for visualizing the evolution of a GA-system. Finally, we discuss the architecture of our GA-visualization system which is implemented on the top of the Khoros visualization package.