Discussion Questions from CS 7390

March 6, 1996

Presented by Jacques Haus

Topic: Information Visualization

• How might the Information Mural system be adapted to deal with extremely wide but sparse data sets?
• The SAGE paper claims that current graphics packages are limited because they have time consuming and complex interfaces. How is Sage Brush better than traditional widget tools for building displays, in regard to this limitation.
• The Information Mural : The authors demonstrated applications of the technique in program visualization. What other areas would it be useful in? For example, would it be useful in a window manager when representing windows miniaturized (instead of icons, show the content of the window)? What about when someone's viewing an image from far away (ie, in a paint program shrinks part of the screen in order to blow up another part of it) or in general to help with any focus + context sort of display? What other uses might there be?
• SAGE uses automatic mechanisms to support users (not replace them). This sounds to me like a good approach to any design task (programming, CAD, etc) not just data visualization. Where else has it been applied?
• Could Sage or a SAGE like system be used in algorithm animation? What about program visualization? WHat could SAGE do to support this? How hard is it to make time a property which can be associated just like any other type of data?
• Information Mural uses anti-aliasing techniques to carry more information in a single display, but there is always a limit of how much it can represent, besides the user may not like the level of detail selected by the algorithm. Will it be better if the user has the control of how many lines a pixel represents in the information mural, and provide a scroll bar if the shrinked picture is still too big for the screen?
• SAGE helps the user to design graphics by providing a knowledge base of design choices. Will it be more powerful if a learning mechanism is provided to adapt to the user's choice for more advanced users?
• In the "Napoleon 1812" exapmle, the paper says that SAGE infers temperature, end latitude and end longitude so the user does not need to specify them. I'm skeptical that it can do this so automatically. How did it know that the temp. was a color attribute and ending lat/long were positions? What if it paired end lat/temperature for position, and end longitude for color? Seems too contrived, to me.
• How does simply specifying a bitmap in SageBook allow you to generalize that bitmap to any situation or data set? What else do you need besides just the raw pixels? How much does the user need to specify?
• How does SAGE guide users to design intelligently - does it just use the samples in SAGE Book, or is there some other method?
• In order for SAGE to make intelligent assumptions when the user does not provide all the requirements, SAGE must be intelligent? How is it driven? Does it use AI methods? How big is the code?
• Information Mural
  • What is the purpose of this paper? The "anti-aliasing" techniques presented certainly are not new to the computer graphics community, and neither is the application domain. A quantitative study of why the information mural is superior to alternative approaches would have been more interesting. Also, I would've liked a discussion on in what situations the IM is a viable solution, and when it is only of limited use (i.e. not specific examples, but in more general terms).
  • It would seem that computing the IM image is expensive. How well is animation supported? Of course, the mipmapped texture capabilities of today's workstations give you this for free.
• SAGE Either I misunderstood their discussion on prototypes, or they just gave bad examples. They say "[The network prototype] eliminates the need for users to construct networks from primitives each time."
  • You still have to create the network and lay out the nodes and arcs.
  • Aren't the circles and lines that the network prototype provides the primitives? Or is it the case that the network prototype is a tool that, e.g., can be applied to a list of 2-tuples that indicate the nodes' positions? I'm unclear on what the prototypes buy you. BTW, we really seem to have departed from Software Visualization. These papers need to be put in context.
• The paper claims that SAGE can automatically choose "good" method to represent your data, even if you leave all of the choices to it. In the example of Napoleon's 1812 campaign, the system inferred the mappings for end latitude, end longitude, and temperature. How can Sage make appropriate choices without user direction? I guess it is discussed in another of Roth's papers (Automating the Presentation of Information) buty they don't talk enough about it here.