Large-Graph Layout Algorithms at Work: An Experimental Study

Abstract

In the last decade several algorithms that generate straight-line drawings of general large graphs have been invented. In this paper we investigate some of these methods that are based on force-directed or algebraic approaches in terms of running time and drawing quality on a big variety of artificial and real-world graphs. Our experiments indicate that there exist significant differences in drawing qualities and running times depending on the classes of tested graphs and algorithms.

Graph visualization with latent variable models

“…They are all at least an order of magnitude faster than Fruchterman and Reingold's grid-variant algorithm, and generally produce much better layouts [7].…”

“…The classical MDS (CMDS) algorithm by Kruskal and Seery [14], one of our comparison methods, uses the shortest path pairwise distance matrix. Spectral methods are significantly faster than the fastest force-based methods, but they have a tendency to produce layouts with many overlapping nodes [7]. Civril et al, who recently independently rediscovered the algorithm, found that CMDS suffers less from this tendency than other popular spectral methods [2].…”

“…Since Walshaw published his algorithm, several other multilevel methods have been developed; see [7] for a comparison. They are all at least an order of magnitude faster than Fruchterman and Reingold's grid-variant algorithm, and generally produce much better layouts [7].…”

“…It is hard to say what a visualization based on this local principle will reveal about global structural properties, unless the structure of the graph is very simple and regular, as in the case of a grid or a mesh. Perhaps because of this, authors of graph drawing methods have traditionally tested their methods on graphs with simple and regular structures (see, e.g., [5,7,29]). We will see that conventional graph drawing methods do fail to reveal global structure in more complex graphs.…”

Interaction in the Visualization of Multivariate Networks

“…Network layout can be calculated solely in dependence of network structure [26], only in dependence on node properties [6] or a combination of both network structure and network attributes [37,57]. The type of layout depends on the user task.…”