Empirical Evaluation for Graph Drawing

Summary

Graph Drawing provides, among other things, the algorithmic foundations for network information visualization. It has considered implementation and experimentation as integral aspects from its very inception and recent research has demonstrated varying approaches to empirical evaluation. Experimental standards, however, have never been established, and little progress toward higher levels of sophistication can be observed.

The seminar was a community effort organized as a hands-on training event. It brought together experts on experimentation from fields with an established experimental tradition (referred to as "trainers"), and a group of graph drawing researchers expected to act as exponents and multipliers ("participants"). After two days of invited lectures on experimental methodology in different disciplines and a problem selection session, participants spent three days in working groups designing experiments. Trainers moving between groups and intermittent reporting session facilitated knowledge dissemination.

Participant feedback in the Dagstuhl survey indicates that the inclusion of trainers was highly appreciated. A number of experimental designs for a broad range of problems have been developed, and it is expected that many of them will be implemented and carried out in collaborative follow-up work.

As everyone who has ever been to Schloss Dagstuhl knows, Dagstuhl seminars are the ideal forum for achieving such goals. The fact that a considerable part of the graph drawing community came together for a week to focus on experimentation is expected to lead to a rapid diffusion of the seminar results and foster the acceptance of new methodology and criteria within the community.

On behalf of all participants, the organizer express their sincere gratitude to the Dagstuhl staff for their outstanding service and support.