Fusion:

1
Fusion
Interactive Coordination of Diverse Data,
Visualizations, and Mining Algorithms

• Chris NorthNathan Conklin, Kiran Indukuri, Varun
  Saini, Qiang Yu
• Lab for Information Visualization and Evaluation
• Department of Computer Science
• Virginia Tech

2
Problem

• Need visual interfaces for databases
• Hard to build Unique data, tasks requires
  unique UI
• Hard to maintain Brittle under data schema
  changes
• Need coordinated approach for
• Diverse data sources, multiple tables
• Diverse visualization components
• Data mining

3
Dynamaps
4
Dynamaps in Fusion
5
Dynamaps in Fusion
6
Proposed Solution Fusion

• Data fusion
• Visualization fusion
• User data owners
• Provide user interface for data
• For themselves or for dissemination

7
Fusion UI Layers
8
Goals

• Flexibility
• Custom visual workspaces
• Diverse databases
• Usability
• Empower users, data owners, non-programmers
• Rapid construction
• Extensibility
• User extensible, 3rd party components
• Diverse tools, data
• Accessibility
• Web-based, browser, no installation
• Website builder

9
Bioinformatics in Fusion
10
Bioinformatics in Fusion
11
Bioinformatics in Fusion
12
DataFaces

• Data schemas and visualization schemas
• Familiar data structure
• Exended to visualization structure
• Distinct layers, yet integrated
• Helps understand multiple-view visualizations

13
Re-description

• Re-describe a subset from one context to another
• Want generalization
• Brushing is (too) specific
• Seek out interesting re-descriptions
• Data mining visualization

Experiment 1
Experiment 2
14
Data Mining and Visualization

• ILP Inductive Logic Programming
• Derive rules that relate subsets between contexts
• E.g If flavonoid then positively expressed
• Visualizations define the contexts
• Bias selection tools
• Evaluation criteria

15
Mining in Fusion
16
Mining in Fusion
17
Mining in Fusion
18
Mining in Fusion
19
Demo Summary

• Data fusion
• Visualization fusion
• Generalized coordinations for data mining
• http//infovis.cs.vt.edu/snap/
• Thanks to Agilent, Census Bureau