Stats and Layout - Graph Analysis and Visualization - page 118

• Force Atlas —The Force Atlas layouts can produce good layouts
pulling strongly connected nodes together and weakly connected nodes
apart. Force Atlas may require more iterative experimentation with
parameters. If available, try Force Atlas 2, and try adjusting parameters
such as “Prevent Overlap,” “LogMode,” and “Scaling.” Scaling adjusts
the ratio of the attraction force along the links and the repulsion
between nodes. Force Atlas can work well for dense graphs and fully
connected graphs where the edges have weights associated with them.
Force Atlas sometimes is less effective when the graph is very sparse, or
the graph has many components.
• Yifan Hu — Yifan Hu is somewhat similar to Force Atlas. It can be fast
on large data sets. Whereas Force Atlas tends to get different results
tweaking various parameters, Yifan Hu tends to come up with a similar
layout. Yifan Hu does not take into account node sizes, edge weights,
and so on, in the software package Gephi.
To get a sense of how these different algorithms compare, Table 4-3 shows
the results of these algorithms on a few different data sets (all done using
Gephi).