PROJECT 4

Article: The Picture of Uncertainty

Class: CS 6825 Section 1

Name: Stephen Marth 8105

 

 

OUTLINE

 

 

I . Visualizing Complex Data

  1. Importance of Visualization to understanding data
  2. Human Processing of Image data surpasses that of merely words of numbers
  3. Human understanding and interpretation depends on what is displayed
    1. Visualization should be careful not to misrepresent data
    2. Visualization may be misleading if it omits data
  1. Uncertainty in Data
  1. Visualization is an abstraction – representing numbers with a picture
  2. Complex data processes introduce uncertainties
    1. Data collection
    1. Transformation
    1. Data Display
    2. Uncertainty does not negate the value of the visualization
    3. Awareness that uncertainty exists is important
  1. Applications Affected by Uncertainty
  1. Geospatial applications are especially vulnerable
    1. The picture produced is not absolute reality
    2. Every point of a region is not represented
  1. Representations of complex scientific and numerical data
    1. Molecular and computational-fluid dynamics
    2. Medical Imaging
    3. Bioinformatics
    4. Multi-dimensional financial information
  1. Data Types that introduce uncertainty

    2. A. MRI Images – used to diagnose tumors

    B. Weather information (air-pressure samples)

    C. DNA and Gene

    D. Molecular studies with regard to Drug Therapy

  3. Role of Uncertainty
  1. Uncertainty visualization has been given little attention
  2. People tend to believe what is visually represented
  3. Visualization does not convey areas that are uncertain
    1. Doing so may add confusion to the viewer
    2. Viewer may become suspicious
    3. Uncertainty causes less impact - decision-makers often avoid dealing with uncertainty
    4. People want clear pictures
    1. Uncertainty should often be represented with fuzzy edges or muted colors
  1. Indication of uncertain data is needed to making accurate decisions based on the display
  2. Uncertainty must first be identified, then quantified
    1. Statistical techniques used to represent uncertain data
    2. Potential distribution around a mean value is created
    3. The harder part is to determine how much uncertainty is acceptable for an application
  1. Methods for representing uncertainty
  1. Broad range of visualization methods in use
    1. Vector glyphs used with flow simulations
    2. Distortion in map projections
  1. Application independent method called RDT (Reconfigurable Disk Tree)
    1. Setup of links and nodes that can be rearranged
    2. Displays hierarchical spatial, non-spatial data and uncertainty information
  1. Enhance user’s perception of uncertainty
    1. Side-by-side viewing
    2. Pseudocoloring
    3. Transparency
    4. Texture
    5. Animation
  1. Depends on how important conveying uncertainty is to the application
    1. Visually obscure areas where uncertainty exists so it can not be used
    2. May be displayed in background color and accessed if needed
    3. Clearly separate data from uncertainty using animation
  1. Consideration of human intuition
    1. color variations /dim colors
    2. fade out
    3. fuzzy areas
    4. opacity/transparency
    5. morphing between various versions of reality
  1. Advances
    1. Modeling isn’t new but there is more emphasis on it
    2. Much of the programming is custom or developed using shareware or commercial toolkits
    3. Sun’s Java 3-D provides platform independence
    4. Internet ready
    5. Flexible enough for many applications
    6. Formal models needed for uncertainty to speed up detection and analysis
    7. Challenge is to present information to decision makers so they will incorporate uncertainty in the decision-making process