our research




  • Lauro Lins (previously at SCI and NYU/Poly) joined our group.
  • An updated version of the visualization of the Tree of Life is here. A NYTimes article discussed an NSF funded visualization project of the Tree of Life.
  • A demo of Twitterscope based on Graphviz GvMap is online.
  • Emden Gansner is on the PC of the 2012 Graph Drawing Symposium
  • Large-scale network visualization and multidimensional scaling by Approximate Low Rank Stress Majorization to be presented at Eurovis 2012. We're delighted that Marc Khoury, an intern in 2011, was a co-author with Yifan Hu, Shankar Krishnan and Carlos Scheidegger from our group. First sub-quadratic-memory algorithm.
  • Summer interns in the vis group for 2012 include Xiaotong Liu, Mehmet Adil Yalcin and Sohaib Ghani.
  • Carlos Scheidegger is on the PC for ICDM 2012, and giving an invited talk at Interface Symposia 2012 in May.
  • Jim Klosowski is on the PC for the International Symposium of Visual Computing (ISVC 2012) and the International Conference on Computer Graphics, Visualization, and Computer Vision (WSCG 2012).
  • North is serving on the PC for IEEE Infovis, VizSec, and ED2012 at the Diagrams Symposium.
  • Yifan Hu will give a talk at NIST on "Visualizing Data with Graphs and Maps", May 7, 2012.
  • Yifan Hu visited Prof. Xiaoru Yuan at Beijing University and gave a talk "Visualizing Large Graphs and Clusters", Mar 5, 2012.
  • Yifan Hu presented a paper,  "A Maxent-Stress Model for Graph Layout", by Gansner, Hu & North at IEEE PacificVis2012, Feb 28-Mar 2, Songdou, Korea.
  • Yifan Hu served on the PC for IEEE PacificVis2012.


  • Yifan Hu gave a talk at DIMACS on "Getting the Big Picture: Visualizing Large Graphs and Clusters", Oct 17.
  • Yifan Hu, who co-maintains the University of Florida Matrix Collection, recently has to layout graphs in the Collection that have billions of edges and hundreds of millions of nodes. At this scale, the first problem is that the usual 32-bit version sfdp simply can not store all the edges. A 64-bit version is able to layout such a graph (given a day or two), but rendering the layout became an issue. Writing out the graph in postscript took 100GB of disk space, and conversion to a bit-map based format using the usual tools failed. In the end he resorted to rending using OpenGL in a streaming fashion, and save to a bit-map format. The result is seen, for example, here.  The next challenge: how to interact with and make sense out of such a huge graph?
  • 600,000 views in StumbleUpon of Yifan Hu's Music Map! Interesting reactions to an Infovis technique from a broad, non-technical audience.
  • Accepted paper at ESA 2011: Approximating Minimum Manhattan Networks in Higher Dimensions by Aparna Das, Emden R. Gansner, Michael Kaufmann, Stephen Kobourov, Joachim Spoerhase, and Alexander Wolff.
  • Accepted paper at KDD 2011: Unsupervised Clustering of Multidimensional Distributions using Earth Mover Distance by David Applegate, Tamraparni Dasu, Shankar Krishnan and Simon Urbanek.
  • Marc Khoury and Nivan Ferreira are working with us as interns for the summer of 2011.
  • Yifan Hu attended PacificVis 2011 conference and presented two papers,  Multilevel Agglomerative Edge Bundling for Visualizing Large Graph and Visualizing Dynamic Data with Map.
  • A visualization of the tree of life by Yifan Hu was mentioned in Wired Science.
  • Yifan Hu created an interactive visualization of tweets about TED talks. It uses Graphviz GvMap. (Requires HTML5 canvas such as in Firefox or Safari; does not work with Microsoft Internet Explorer.)
  • Jim Klosowski and Shankar Krishnan presented Real-Time Image Deconvolution on the GPU at SPIE 2011.

About Us

What visualization means to us:

Scale. Many problems today involve millions or billions of objects. Visualization is a key aspect of ``Big Data'' because it enables human insight and domain knowledge to be combined with algorithms. We need to cope with scale in visualization, computation and human interaction.


Automation. Visualization needs to be repeatable, generalizable, and reusable as a fundamental service. As far as possible it should just work without unnecessary manual operation or tweaking.


Optimization. Applied algorithms are at the heart of producing aesthetically beautiful, readable visual encodings of large, complex data sets. Our goal is to approach the quality of the best designs while satisfying the need for scale and automation. Reliance on other aspects of visualization, such as an overall knowledge discovery process, or on high interaction, can never take the place of excellent visualization itself.


We admire the design of beautiful graphical metaphors, but also believe it is smart to look at the past 150 years of design to inform the work that we do. The design of novel metaphors is stimulating and interesting research but often not the only or even the best approach to solve a visualization problem.

Questions we work on

  • How can we visually explore huge networks? Can we browse them directly? Can we extract subgraphs that show and explain interesting relationships?
  • What is the next generation of software systems to support high performance, interactive data analysis?
  • How can we visualize billions of transactions and events? How should we engineer systems to provide near-instantaneous access to overviews, mid-level views, and details?
  • How can high performance graphics, algorithms and visualization enable future communication services?