Exploring High-D Spaces with Multiform Matrices and Small Multiples

1
Never forget your primary wealth, your and your
familys health, it will be your hope and your
familys hope for ever.
2
Exploring High-D Spaces with Multiform Matrices
and Small Multiples by Alan MacEachren etc.

• Presenter
• Xu Liu
• Ming Luo

3
Outline

• Contribution
• Main ideas
• Demos
• Critiques

4
Contribution

• GeoVISTA development environment
• A general framework, ready for combination of any
  information visualization forms
• Dynamically visualize information with user
  interaction
• Explore high dimension by univariate and
  bivariate relations between attributes
• Find most significant relations between attributes

5
Multiform Bivariate small Multiple
6
Multiform Bivariate small Multiple
O
O
O
7
Multiform Bivariate Matrix
8
Multiform Bivariate Small Multiple and Matrix

• Multiple and matrix are designed as generic
  JavaBean components, providing a Java interface
  through which any bivariate representation form
  (instantiated as a JavaBean) can communicate.
• Visualization is updated dynamically with the
  users interaction

9
Find most significant relations between attributes

• Maximum conditional entropy between each pair of
  attributes
• Generate an ordering of attributes, which keeps a
  hierarchical clustering
• Select subspace interactively or automatically

10
Generate an ordering of attributes

• A node is an attribute
• An edge is the maximum conditional entropy
  between two attributes

11
Generate an ordering of attributes

• A node is an attribute
• An edge is the maximum conditional entropy
  between two attributes

12
Select subspace
13
Critiques

• Grid-based space-filling display
• It can not show more information than a simple
  scatterplot (maybe worse)
• Color legend needs training
• Rare visualization tools for more than 2
  variables

14
Demo

• Basic Usage
• Conditioning
• Bi-variable you see its hard to select colors
• PCP a good alternation for multi variable
  scatter plot
• Bad example who can tell me whats going on
  here?

15
THANKS!

• You may find the demo at
• http//unc.dl.sourceforge.net/sourceforge/geovista
  studio/studiodemo1.zip
• Enjoy!