Transforming Clusters into a Color Map

[Tim Churches]

Lemniscate wrote:
> 
> Hi all,
> 
> I have recently been asked to write a program that will operate on
> huge arrays of numbers.  As such, I have quickly immersed myself in
> Numeric.  Things are going well (Numeric, in my experience, is
> extremely effective and relatively easy to learn).
> 
> However, I've been informed that I need to 'color map hotspots' in the
> results.  For example, my boss wants me to transform the results (of,
> for example, standard deviations run across each array) and then
> cluster them by color (I can explain more if anybody needs it).
> 
> There are a few programs that do this but they all have propriety
> formatting of the results (which precludes some of the things we want
> to do) and limits onto the types of comparisons that can be done.
> Ideally, we would want our own implementation that we could use in any
> way we wanted.
> 
> There isn't alot of support for this type of programming at my
> company, so the hope is that I can get a 'minimally' functional
> prototype so we can see what can be done.
> 
> >From the people I've talked with, color-maps like these are actually
> fairly common, so I was hoping that somebody could point me in the
> right direction (maybe a module already exists to do this that
> interfaces with Numeric, etc.)  I'm doing some searches but I'm not
> having a whole lot of luck (right now, I am going through stuff like
> "Numeric Color" etc.  Thanks a bunch.

Take a look at the R package - see http://www.r-project.org
in particular the screenshots page - there is a nice "heat map" of
a volcano at the bottom of the page. 

You can transfer your data to R via delimited text files, but if you 
are using Unix or Linux, then have a look at Walter Moreira's RPy module
(see http://rpy.sourceforge.net ) which embeds R in Python and allows
NumPy arrays and other Python data structures to be transparently
transferred into the R environment (and acted upon there).

Tim C