Seminar 15052 – Empirical Evaluation for Graph Drawing

Janet Siegmund
Universität Passau
Seminar 15052 – Empirical Evaluation for Graph Drawing
• Janet Siegmund
• Software Product-Line Group
Janet Siegmund
Software Product-Line Group
Ariane-V Rocket
$370 mio
Janet Siegmund
Software Product-Line Group
Consistent use of
background colors can
improve program
comprehension
Confounding
factors for
program
comprehension
Questionnaire to
measure programming
experience
Janet Siegmund
Software Product-Line Group
Differences between
expert and novice
program comprehension
Automatic extraction of
collaboration networks
of developers
Understanding
programmers' brains
with functional
magnetic resonance
imaging (fMRI)
Janet Siegmund
Software Product-Line Group
Bettina Speckmann
TU Eindhoven
Seminar 15052 – Empirical Evaluation for Graph Drawing
Algorithms for geovisualization and automated cartography
Bettina Speckmann
TU Eindhoven
Curved Schematization
Area preserving
circular arcs
Bézier curves
circular arcs – different styles – degree of curviness
User study
Aesthetics
Recognizability
Simplicity
Matthias F. Stallmann
North Carolina State University
Seminar 15052 – Empirical Evaluation for Graph Drawing
Dagstuhl 15052 Introduction
Matthias (Matt) Stallmann
spent first 7 years
in Bad Nauheim,
Hessen
last 40 at NCSU, Raleigh, NC
father grew up
in Königsberg
Is Computer Science a Science?
Challenges:
•  implementation takes time
•  need infrastructure
•  too many variables
•  what to measure
•  inconclusive?
Observation
Theory
Analysis
Experiments
Random initial arrangements, in parallel
Crossing minimization in
layered graphs, for example.
No communication
get optimum on many
“large” instances
(~300 nodes, edges) using
dfs+mod_bary+sifting
Best
different initial
arrangements
(e.g., 1024)
- 
- 
- 
- 
expected solution quality
probability of optimum or close to it
better to randomize input or heuristic?
better to run longer or use more
parallelism?
Bottleneck crossing minimization
total crossings = 44 (optimum)
one edge has 9 crossings
total crossings = 57
no edge has more than 5 crossings
Application to crosstalk/delay in circuits, not much is known
ILP, SDP? Good known heuristics
Michael Stoffel
Universität Konstanz
Seminar 15052 – Empirical Evaluation for Graph Drawing
Michael Stoffel
University of Konstanz
Chair of Political Methodology
Some other projects …
MP incentives
to cater to
party vs. constituency
EU spending efficiency
subject to
local and national
gov’t interests
Dorothea Wagner
KIT - Karlsruher Institut für Technologie
Seminar 15052 – Empirical Evaluation for Graph Drawing
Dorothea Wagner
Algorithm Engineering
Route Planning
me
peri nt
Ex
Analyze
Design
Algorithmics
Implement
Graph Clustering
Graph Drawing
Social Network Analysis
Sensor Networks
Energy Networks
...
1
Dorothea Wagner – Dagstuhl 2015
Institute of Theoretical Informatics
Algorithmics Group
Route Planning
fast shortest path computations (on-board
and off-board)
practical applications (Google, Microsoft,
Nokia, telenav, ...)
time dependence and dynamic traffic data
flexible objectives (e.g. , travel time vs. cost)
and constraints (e.g. , vehicle size, weight,
toll roads)
public transport
multimodal route planning
alternative routes
intelligent point-of-interest search
theory
Dorothea Wagner – Dagstuhl 2015
Institute of Theoretical Informatics
Algorithmics Group
Graph Clustering
network analysis and exploration
(social, biological, technical, ...)
compute densely connected
subgraphs (clusters)
=
evaluate structure and function of
clusters
new
cluster
G0
Dorothea Wagner – Dagstuhl 2015
G1
dynamic networks
temporal tracing of
clusters
G2
Institute of Theoretical Informatics
Algorithmics Group
Graph Drawing
algorithms for visualizing graphs in the plane
applications: sociology, biology, UML-diagrams, ...
examples of our research:
constrained planar graph drawings (e.g., point set
embedding, extending partial drawings)
alternative drawing conventions
argument maps
+
Dorothea Wagner – Dagstuhl 2015
?
Institute of Theoretical Informatics
Algorithmics Group
Alexander Wolff
Universität Würzburg
Seminar 15052 – Empirical Evaluation for Graph Drawing
alexander.wolff @ informatik . uni-wuerzburg . de
Drawing metro maps
[with Martin2 , Herman, Max, GD’12]
alexander.wolff @ informatik . uni-wuerzburg . de
Drawing metro maps
Boundary labeling
[with Martin2 , Herman, Max, GD’12]
[with Michael, Antonis, Michalis, Andr´
e,... GD’04, WADS’13, GD’14]
alexander.wolff @ informatik . uni-wuerzburg . de
Drawing metro maps
Boundary labeling
[with Martin2 , Herman, Max, GD’12]
[with Michael, Antonis, Michalis, Andr´
e,... GD’04, WADS’13, GD’14]
todonotes.sty
alexander.wolff @ informatik . uni-wuerzburg . de
Drawing metro maps
[with Martin2 , Herman, Max, GD’12]
[with Michael, Antonis, Michalis, Andr´
e,... GD’04, WADS’13, GD’14]
Boundary labeling
luatodonotes.sty
Kai Xu
Middlesex University
Seminar 15052 – Empirical Evaluation for Graph Drawing
Kai Xu
Lecturer,
London
Bachelor,
Shanghai
PhD, Brisbane
Research Scientist,
Hobart & Canberra
Postdoc,
Sydney
User Study – Curved Edges (InfoVis 2012)
Making Sense of Graphs (InfoVis 2013)
How to Evaluate?
How to Evaluate Dense (and Large) Graphs?
• What is the salient
information?
• Connectivity?
• Structure (such as
clustering)?
• Shape?
• How to evaluate?
End
Seminar 15052 – Empirical Evaluation for Graph Drawing