Information Technology Reference
In-Depth Information
GION: Interactively Untangling Large Graphs
on Wall-Sized Displays
Michael R. Marner 1 , Ross T. Smith 1 ,Bruce H. Thomas 1 , Karsten Klein 2 ,
Peter Eades 2 , and Seok-Hee Hong 2
1 University of South Australia, Adelaide, Australia
{ michael.marner,ross.smith,bruce.thomas } @unisa.edu.au
2 The University of Sydney, Sydney, Australia
{ karsten.klein,peter.d.eades,seokhee.hong } @sydney.edu.au
Abstract. Data sets of very large graphs are now commonplace; the scale of
these graphs presents considerable difficulties for graph visualization methods.
The use of interactive techniques and large screens have been proposed as two
possible avenues to address these difficulties.This paper presents GION, a new
skeletal animation technique for interacting with large graphs on wall-sized dis-
plays. Our technique is based on a physical simulation, and aims to enhance the
users' ability to efficiently interact with the graph visualization for exploratory
analysis. We conducted a user study to evaluate our techniqueagainst standard
operations available in most graph layout editors, and the study shows that the
new techniqueproduces layouts with less stress, and fewer edge crossings. GION
is preferred by users, and requires significantly less mouse movement.
1
Introduction
Graphs provide a versatile model for data from a large variety of application domains,
including biology, finance, telecommunication, software engineering, and social sci-
ences. Graph visualization helps scientists and engineers to understand critical issues
in these domains. However, the depth of understanding depends on the quality of the
drawing.Automatic graph layout methods are developed for computational efficiency
and quality, i.e. readability. These methods however can only optimize a few criteria in
combination, and it is impossible to define a quality measure that allows to create opti-
mal layouts for all graphs, tasks, and observers. Moreover, the size of relevant data sets
for analysis has grown exponentially over the last years. For example, data from social
networks, biology, and finace continuetogrow at a rate that is not accommodated by
current methodologies.
While some layoutalgorithms are capable of laying out graphs with hundreds of
thousands of nodes in a few seconds [4], data sizes from practice are still a challengein
anumber of ways:
- There is a trade-off between computational resources and layoutquality. For exam-
ple, algebraic methods runquickly but in many cases give poor results [6], while
stress minimization [5] gives good quality layouts but is too slow for interactive
work on large graphs.
Search WWH ::




Custom Search