Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Overview of Data Visualization 2

Martin Smith
October 08, 2012
290

Overview of Data Visualization 2

Martin Smith

October 08, 2012
Tweet

Transcript

  1. Visualization? • Practically any technique using images or diagrams for

    communication • Not new – cartography is old • Not old – computer graphics are young • Wide applications: architecture, product design, education, communication • A Periodic Table of Visualization Methods
  2. From social structure, to data, to visualization • People &

    relationships as data • Mathematical graphs as social networks • Network analysis as applied graph theory
  3. Side note: ‘Storage’ is important • Consider advantages and disadvantages

    of particular data structures: – Typically breaks down into lists and matrices – Incidence (edge) and Adjacency (vertex) • Rank algorithms by how well they: – Enumerate nodes, test adjacency, etc
  4. Graph theory – Graph drawing • Leonhard Euler, functions, f(x),

    e, i, Σ, and Seven Bridges of Königsberg, Prussia
  5. Graph layout • Hierarchical – source to sink • Tree

    – roots without cycles • Symmetric – create visual symmetry • Orthogonal – min. area & edge crossing • Spectral – eigenvectors of a matrix • Force/MDS – springs and electric charges
  6. Evaluating graph drawings • Is it fast to compute? Can

    we parallelize it? • Does the algorithm converge? How quickly? • Does it show obvious symmetries? • Can we adapt it with parameters? • Does it minimize edge crossings? Size? • Does vertex nearness reflect adjacency? • Are sizes, distances and shapes distributed uniformly? • Election 2008 visualization – let’s evaluate
  7. Force-directed algorithms • Attractive and repulsive forces simulated as a

    physical system • Forces can be gravity (Newton), springs (Hooke), charged particles (Coulomb), magnetism (Maxwell?) • Advantages: quality, flexibility, interactivity • Disadvantages: can be slow, hurt by local minima or initial conditions
  8. Two common force-directed algorithms • Fruchterman-Reingold: Nodes as steel rings,

    edges as springs, electrical repulsive force, step width using a global cooling temperature definition • Kamada-Kawai: Same as above, but instead of a temperature, minimize force equations with some initial node criteria
  9. What can we do to make graph drawing better for

    social networks? • Modify the graph, improve data collection • Add constraints to force-directed algorithm • Get better initial conditions, more iterations • Deal with orientations and coordinate systems • 3D -> 2D, or use more interesting forces • Apply multiple layout algorithms • Curved lines and uniform distributions • Come up with better algorithms in general
  10. EgoNet (egonet.sf.net) • Java, Swing – hosted at SourceForge •

    Java Universal Network/Graph Framework • Currently defaults to F-R, but places isolates regularly instead of randomly • We’d like to explore other optimizations
  11. Thank you! • Slides will posted (where?) • Martin Smith

    [email protected] • EgoNet: http://egonet.sf.net/ • JUNG: http://jung.sourceforge.net/
  12. References • 1. Freeman, L., Visualizing Social Networks. Journal of

    Social Structure 1(1), Carnegie-Mellon, 2000. • 2. Giuseppe Di Battista, Peter Eades, Roberto Tamassia, Ioannis G. Tollis. Algorithms for Drawing Graphs: an Annotated Bibliography. Computational Geometry: Theory and Applications 4:235-282, 1994. • 3. Tamassia, R. Advances in the Theory and Practice of Graph Drawing. Theoretical Computer Science 217 (2), 1999. • 4. Fruchterman, T. M. J., & Reingold, E. M. Graph Drawing by Force- Directed Placement. Software: Practice and Experience, 21(11), 1991. • 5. Kamada, T. & Kawai, S. (1989). An algorithm for drawing general undirected graphs. Information Processing Letters, 31, 7-15. • 6. Network Workbench Community Wiki at https://nwb.slis.indiana.edu/community/?n=VisualizeData.HomePage