> One more quick question-- which I asked before but never got working to
> my satisfaction. The display is a 3D histogram of the distribution of
> documents based on strong concepts returned by the Kohonen SOM algorithm.
> You can see the histogram at
>
> http://www.wizards.dupont.com/Equinox/Equinox.html
>
> Hopefully that link is still up. In usability studies at the AI group
> at Arizona they found that users liked color graphs so to please these types
> I'd like to color code the histogram. Since I really don't have 4D data
> the best I could do is to code by elevation (# documents associated with
> a particular concept). A quick attempt of this yields:
>
> ScalarMap documents=new ScalarMap((RealType) range,Display.ZAxis);
> display.addMap(documents);
> ScalarMap colormap = new ScalarMap((RealType) range,Display.RGB);
> display.addMap(colormap);
>
> So we're just mapping the number of documents and the color to the
> elevation/altitude. The results are interesting but not very pleasing
> (I can post an image if you want). Do you have any other suggestions
> where we can vary the color so that peaks stand out not just by height
> but by color (my suggestion a few years ago was to make the peaks white
> so it looked like snow).
My first thought is that the complex process of generating a
spatial mapping of semantic information must create numerous
intermediate numerical variables that can be mapped to color,
or to parameters of complex Shape glyphs, in order to better
understand the semantic information.
Alternatively, mapping range to ZAxis and to RGB, you might
use texture maps (to get nice bright colors), pick color tables
with bright colors, and resample to higher resolution to have
higher resolution texture maps.
You might dispense with the topography surface, and instead
also map range to Shape, to have a surface approximated by an
array of colored glyphs. You can play tricks with mappings to
Shape and ZAxis, to make a sort of 3-D bar chart (a couple
years ago Ugo did something like this for 2-D bar charts, as
I recall).
Good luck,
Bill
----------------------------------------------------------
Bill Hibbard, SSEC, 1225 W. Dayton St., Madison, WI 53706
hibbard@xxxxxxxxxxxxxxxxx 608-263-4427 fax: 608-263-6738
http://www.ssec.wisc.edu/~billh/vis.html
