Mao Lin Huang

1
Visual Representations of Data and Knowledge

• Mao Lin Huang
• University of Technology, Sydney,

2
Rendering Effective Route Maps
3
General Idea

• Automatically generate a route map that has the
  same properties as a hand drawn map.
• Hand drawn maps
• Exaggerated Lengths (non-constant scale factor)
• No irrelevant information

4
More Specifically

• Constant scale factor
• Road lengths on a conventional map vary in
  several orders of magnitude gt small roads and
  neighborhoods are hard to navigate with large
  maps
• Information irrelevant to navigation
• Names of locations, places, cities, etc. that are
  all far away from the route
• Takes up space that would be otherwise useful for
  showing crossroads and relevant landmarks

5
Generalization Techniques

• Generalize Length
• Use more space for short roads, less for longer
  ones. Distribute based on importance, not
  physical length
• Generalize Angle
• Align roads or make room for others
• Generalize Shape
• Navigator doesnt need to know roads shape.
• Simpler roads are easier to differentiate on a
  map.

6
Demo at mapblast.com
7
Simple Visualization Model
Data
View Port
Visual Mapping
8
Film Data Table Example
Attributes
9
Visual Mapping

• Define a Space
• Map data ? marks
• Map data attributes ? graphical mark attributes
• Year ? X
• Length ? Y
• Popularity ? size
• Subject ? color
• Award? ? shape

10
Example FilmFinder
38
11
Example FilmFinder
39
12

• Use of graphical time scales as an approach to
  visualize histories. Time Scale History
  Intuitive

13
(No Transcript)
14
(No Transcript)
15
Patient Records
16
Galaxies

• Projection of clustering algorithms into 2D
• Galaxies are clusters of related data
• Proximity of galaxies is relevant
• Designed to add temporal patterns to clustering

17
Galaxies
18
3D Visualization VR Techniques
19
3D Cone Tree
16
20
3D Cone Trees
17
research.microsoft.com/ggr/gi97.ppt
21
Perspective Wall
18
research.microsoft.com/ggr/gi97.ppt
22
Example 3D-Room (The Exploratory)
20
Robertson, Card, and Mackinlay (1989)
23
3D Navigation Task (Hallway)
research.microsoft.com/ggr/gi97.ppt
21
24
3D GUI for Web Browsing
22
25
3D GUI for Web Browsing
http//research.microsoft.com/ui/TaskGallery/index
.htm
23
26
Web Forager
http//research.microsoft.com/ui/TaskGallery/index
.htm
24
27
WebBook
research.microsoft.com/ggr/gi97.ppt
25
28
3D GUI for Desktop
http//research.microsoft.com/ui/TaskGallery/index
.htm
26
29
(No Transcript)
30
ThemeScape

• Abstract 3D landscape of information
• Reduce cognitive load using terrain
• Elevation, colour encode theme strength
  redundantly
• Landscape metaphor translates well
• Peaks are easy to recognize
• Interesting characteristics include ridges and
  valleys

31
ThemeScape
32
ThemeScape
33
Calendar Based Visualization

• Using 3 dimensions
• X-axis Time of day
• Y-axis Days of data period
• Z-axis Univariate data samples

34
Calendar Based Visualization
35
Calendar Based Visualization
36
Graph-Driven Visualization of Relational Data
Graph Visualization
An example of visualizing relational data. This
is the visualization of a family tree (graph).
Here each image node represents a person and the
edges represent relationships among these people
in a large family.
37
Classical Graph Layouts

• Link-node diagrams
• Layout algorithms (graph drawing)
• Geometric positioning of nodes edges
• Small amount of nodes
• Avoid node overlaps
• Reduce edge crossings

radial layout
symmetric
force-directed
hierarchical
orthogonal
38
Using a very large virtual page
The virtual page technique predefines the drawing
of the whole graph, and then provides a small
window and scroll bar to allow the user to
navigate through it (by changing the viewing
area).
39
Fish-eye views
The fish-eye technique can keep a detailed
picture of a part of a graph as well as the
global context of the graph. It changes the
zoomed focus point.
40
3D Graph Drawing
SGI fsn file-system viewer Image
from http//www.sgi.com/fun/images/fsn.map2.jpg
41
Trees
42
2 Approaches
A

• Connection (node link)
• Enclosure (node in node)
• Structure vs. attributes
• Attributes only (multi-dimensional viz)
• Structure only (1 attribute, e.g. name)
• Structure attributes

C
B
A
B
C
43
Containment Approach
44
Treemaps (Shneiderman)

• Slice and Dice
• Alternate horizontal andvertical cuts for levels
• Node area ? node attribute
• Zoom onto nodes
• Space-Filling
• Structure 3 attributes
• Area, color, label

45
Treemaps
46
Balanced trees
47
Treemaps

• 1000 nodes
• Quantitative attributes
• Good combination of structure attributes
• For unbalanced trees, structure more difficult
• Learning time 20 min
• Evaluation major performance boost over outliner
• Bad aspect ratios long narrow rectangles
• Large scale or deep causes solid black

48
Treemap Algorithm

• Calculate sizes
• Recurse to children
• My size sum children sizes
• Draw Treemap (node, space, direction)
• Draw node rectangle in space
• Alternate direction
• For each child
• Calculate child space as of node space using
  size and direction
• Draw Treemap (child, child space, direction)

49
Cushion Treemaps
50
Squared Treemaps
51
Treemaps on the Web

• Map of the Market http//www.smartmoney.com/mark
  etmap/
• People Map http//www.truepeers.com/
• Coffee Map http//www.peets.com/tast/11/coffee_s
  elector.asp

52
DiskMapper

• http//www.miclog.com/dmdesc.htm

53
2D Tree Drawing (web sitemap)
MosiacG System Zyers and Stasko Image
from http//www.w3j.com/1/ayers.270/paper/270.htm
l
54
PDQ Trees

• OverviewDetail of 2D layout
• Dynamic Queries on each level for pruning

55
Space-Optimized Tree Layout
A large data set of approximately 50 000 nodes
My Unix root with approx. 3700 directories and
files
56
Hyperbolic tree
The hyperbolic browser technique performs
fish-eye viewing with animated transitions to
preserve the users mental map. It changes both
the viewing area and the zoomed focus point.
57
H3
Image from http//graphics.stanford.edu/papers/h3
/fig/nab0.gif