C. Erten; "Information Visualization Through Evolving Graph Layouts"

Information Visualization Through Evolving Graph Layouts

Cesim Erten
Postdoctoral Research Assoc./Adjct. Lecturer
Department of Computer Science
University of Arizona
www.cs.arizona.edu/~cesim

Abstract:
As the amount of information available grows rapidly, among the many problems we face is how to present the information in a sensible way Viewing a possible relationship among the data entities as a graph and visualizing the information by laying out the graph has been one of the most widely used methods. There are three primary objectives of a graph visualization algorithm. The output visualization should convey the relationship in the underlying data. The visualization should be aesthetically pleasing. Finally, the algorithm should produce the resulting visualization in a short amount of time.

This talk focuses on the problem of visualizing information through multiple related graphs. Although existing literature provides many useful pointers as to how to achieve the necessary objectives for a single graph, the same problem generalized to multiple related graphs has not received enough consideration. Two main domains give rise to multiple related graphs: A single relationship that evolves through time or multiple relationships defined on the same dataset. Yearly updated collaboration graph of a scientific research network or daily updated control-flow graph of a software project are sample applications of the first domain. Biological pathways induced by different sets of enzymes, or pathways of similar cells from different organisms can be viewed as applications of the second domain.

In this talk I will present the two main issues regarding this problem: The fact that the graphs are large and the fact that multiple relationships are visualized simultaneously. I will describe my approach to addressing these issues via GraphAEL, a system for graph animations with evolving layouts. I will present the novel efficient layout algorithms and visualization models underlying the system.

May 31, 2005, 14:40, 2019