Graphics Reference
In-Depth Information
the outliers are. At the heart of all force-directed algorithms are the forces.
In most cases, you have control over these settings.
Note
Getting a nice force-directed layout almost always requires some
experimentation with different algorithms and different settings.
Force Settings and Other Settings
Force-directed layouts are, by definition, controlled by their forces. Getting
a good layout often involves going beyond the defaults and adjusting some
of the force attributes, such as attraction, repulsion, and gravity.
•
Attraction/repulsion
—Adjusting the ratio between attraction/
repulsion is one important consideration. This may be set as a single
parameter (for example, Relative Strength in Yifan Hu, Scaling in Force
Atlas, or both in Gephi) or as multiple parameters (for example,
Attraction and Repulsion in Force-Directed layout in Cytoscape). If all
the nodes are in a blob, increasing repulsion can help push the nodes
apart.
•
Gravity
—Gravity is similar to attraction, but it applies to all nodes,
even when not connected. Turning up gravity can bring very distant
nodes back toward the center, but turning up gravity too high will create
a packed circle.
•
Edge weight
—The strength of the attraction can be the same for all
nodes, or it can be based on weight of the edges. Strong weights will pull
nodes closer together. Edge weight is an attribute provided with the
data. This is perhaps the most important additional attribute to include
when you prepare your data.
Other Layout Settings
Beyond basic control over forces, each layout algorithm will have a
bewildering set of additional settings. Some are self-explanatory (such as
no-overlap, which is a useful setting to turn on). Others may be vague.
Experimentation with these settings is required.
Many other possible settings are available, including linear versus
logarithmic, temperature, and so on. For example, in the previous
