Comparing and Evaluating Layout Algorithms within GraphEd

Graph drawing and layout algorithms gain growing interest and importance for the visualization of complex data structures. But, despite many algorithms and concepts, there is no satisfactory solution to the central problem on the criteria for good and readable layouts. We approach this problem and evaluate layout algorithms by comparing their eeect on a large set of sample graphs. In each run we compute statistical data which is collected and evaluated. Our experiments show that traditional layout criteria, such as minimal area or maximal edge length or straight line edges, are not as important as they may appear. Balance is often better than optimization, and displaying the intended and inherited structure of a graph is often more important than these formal cost criteria.