Information Technology Reference
In-Depth Information
an open source 2D physics library popular for game development. The Open Graph
Drawing Framework (OGDF)
2
is usedtoprovidegraph data structures used by the ap-
plication, saving and loadinggraphs at runtime, as well as for layout metrics. Vertex
data is stored in an OpenGL Vertex Buffer Object (VBO). Edge data references the
VBO and is rendered using an Index array. This reduces the amount of data transferred
to the graphics card each frame and improves rendering performance by reducing the
number of draw calls needed. Further enhancements were needed to improve rendering
times on on a multi-projector display. A naive approach would simply involve rendering
the graph in its entirety for each projector. Instead, we use a deferred rendering tech-
nique. The entire display is first rendered to an off-screen Framebuffer Object (FBO).
Following this step, a portion of the FBO is rendered to each projector. This approach
scales much better as the number of projectors increases, as the cost of each projector
is just a single textured polygon.
8Con lu ion
In this paper we presented GION, a new interactive graph layout techniqueoflarge
graph structures. GION is based on a physics engine to provide smooth and under-
standable animations to update the graph layout while the user moves a cluster. The
results of a user study comparing GION with moving asingle cluster at a time found
the use of physics engine produced graphs with less stress, fewer edge crossings, and
less mouse movement. Participants preferred the GION techniquetomoving asingle
cluster during the experiment.
We applied two standard quality layout metrics: stress and crossings. With GION,
users constructed graph layouts that did not show significantly less stress or signifi-
cantly fewer edge crossings, in comparison with the Fruchterman-Reingold algorithm.
These results from our experiments lead ustoquestion the validity of these two stan-
dard metrics for large graphs in the context of human layout improvement, and ourwork
raises the questionastowhatquality metrics should be applied instead. We conjecture
that measures like the precision of neighborhood preservation [5] will be better suited
in this context than standard metrics for small graphs.
Acknowledgment.
This work was supported in part by a grant from the Australian
Research Council - Discovery Grant DP120100248, Linkage Grant H2814 A4421, Tom
Sawyer Software, and NewtonGreen Technologies.
References
1. Ball, R., North, C., Bowman, D.: Move to improve: promoting physical navigation to increase
user performance with largedisplays.In:Proc. of the SIGCHI Conference on Human Factors
in Computing Systems, pp. 191-200. ACM (2007)
2. Catherine, R., Sudarshan, S.: Graph clustering for keyword search. In: Chawla, S., Karla-
palem, K., Pudi, V. (eds.) COMAD. Computer Society of India (2009)
