Visualizing [Dense, Dynamic, Complex] Networks

Jean-Daniel Fekete is Senior Research Scientist (DR1) at INRIA, Scientific Leader of the INRIA Project Team AVIZ that he founded in 2007. He received his PhD in Computer Science in 1996 from Université Paris-Sud. He was recruited by INRIA in 2002 as a confirmed researcher and became Senior Research Scientist in 2006. During 2015, Jean-Daniel was on Sabbatical at NYU-Poly, then at Harvard. His main Research areas are Visual Analytics, Information Visualization and Human Computer Interaction.

Jean-Daniel Fekete is the chair of the IEEE Information Visualization Conference Steering Committee, member of the EuroGraphics EuroVis Steering Committee, and member of the EuroGraphics publication board. He was the General Chair of the IEEE VIS Conference in 2014, in Paris. He was President of the French-Speaking HCI Association (AFIHM) 2009-2013, Conference Chair of the IEEE InfoVis Conference in 2011, Paper Co-Chair of the IEEE Pacific Visualization conference in 2011.

Résumé :

Visualization is about making sense of potentially large and complex structures by finding an adequate graphical representation that our visual and cognitive systems can process effectively. For visualizing networks, the graph drawing community has focused on the node-link representation for decades, trying to address multiple important, difficult, and interesting issues related to 2D embeddings under some optimality criteria (planar drawing, minimizing crossings, graph decompositions, and many more).

Visualizing a graph structure using its adjacency matrix is much less common, although it has been shown to be more efficient than the node-link representation when the graph becomes dense, for important low-level tasks. I will present the line of research I conducted over more than ten years that shows matrix-based visualization of graphs is effective, from simple graphs, social networks, to dynamic weighted networks.

One main question to address in matrix-based visualization is the computation of the vertices order. This problem is known with multiple names: linear ordering, seriation, reordering. With a proper ordering, a visualized matrix reveals important patterns and structures of the graph. I will briefly explain how the problem has been formalized in the past, and show some results revealing sometimes unexpected information.
Finally, there are important questions related to assessing the effectiveness of reordering methods, that I will present as food for thought and challenges, in between the domains of combinatorial optimization, human-computer interaction, perception and cognition.