25.01.15 - 30.01.15, Seminar 15052

Empirical Evaluation for Graph Drawing

Diese Seminarbeschreibung wurde vor dem Seminar auf unseren Webseiten veröffentlicht und bei der Einladung zum Seminar verwendet.


Any application where users visually analyze and interact with relational datasets hinges upon graph drawing techniques. Graph drawing, among other things, provides the algorithmic foundations for network information visualization and is crucial in several different domains ranging from social and natural sciences to information systems and engineering.

Algorithmic experiments and user evaluations are indispensable when it comes to developing practically relevant algorithms that produce easy-to-read graph layouts. Implementation and experimentation have been considered as integral aspects of graph drawing from its very inception, even before many other specializations in algorithmics did. Still, in comparison to other disciplines with a long established tradition of high experimental standards, experimental graph drawing research in general does not yet reach the same level of sophistication.

In this slightly off-beat Dagstuhl Seminar, which could be better described as a hands-on workshop, we want to bring together graph drawing researchers with a clear interest in experimental work with experimentation experts from closely related fields in computer science and also from external fields with more established traditions of designing and analyzing scientific experiments, the latter assuming the roles of trainers. By creating an active learning experience we want to initiate the diffusion and acceptance of advanced approaches to experimentation in graph drawing. We aim at establishing principles and experimental methodology by means of guided knowledge import and appropriate adaptation to the graph drawing context. Ideally, this would result in a position paper drafted by seminar participants.

Another goal is to actually design a set of empirical studies for answering research questions in graph drawing. Within mixed break-out groups consisting of experimentation experts as advisors and graph-drawing researchers as domain experts, the newly acquired knowledge will directly be applied to concrete problems. These will be solicited from the participants beforehand, curated, and then reviewed on site. Following the seminar, groups are encouraged to run the conceived experiments, evaluate them, and publish the results.

According to these goals, the seminar week will start with a series of tutorial lectures given by experts in experiment design followed by the formation of working groups, in which actual graph drawing experiments for selected open research questions are designed under the supervision of the experimentation experts. Towards the end of the week the study designs will be presented and evaluated in a feedback round.