[GRASS5] Evidential reasoning classifier

[benducke at compuserve.de]

That's interesting.
I have been working on an implementation of Dempster
Shafer Theory as well (albeit for purposes of spatial
prediction). I am a big fan of it and pleased to see
that others find it useful, as well.

It seems to me that uncertainty management and decision
support is something where GRASS 6 currently lacks in comparison
with IDRISI. Things needed, that have been mentioned on this
list before include neural networks, fuzzy logics and 
Bayes models.

As you may know, if you have been reading this mailing
list lately, I am also currently developing a program
that allows to install extensions into a running version
of GRASS 6 including menu entries, HTML docs and the
modules themselves. That program is about 95% finished now.
My own Dempster-Shafer modules are about 80% done.

So maybe it would be about time to join forces and create
a "Decision Support" extension that includes, DST, Bayes,
fuzzy logics, NNs in one neat package along with good
documentation on the mathematical background and practical
uses?

Benjamin

----- Originalnachricht -----
Von: Sajith VK <sajithvk at gmail.com>
Datum: Dienstag, 19. Juli 2005 6:47 am
Betreff: [GRASS5] Evidential reasoning classifier

> Hello all,
>    I am working on evidential reasoning (Dempster Shafer theory) 
> for image
> classification. I am implementing it as a grass module.
> 
> Till now, I have developed the training module, which generate 
> signaturesin the for need for evidential reasoning. (It select the 
> frequency of occurance
> of each pixel in the class instead of the mean and covariance).
> 
> I can upload the module, if any one else need this kind of 
> signature for their
> work, so that duplication of works can be avoided........
> 
> Note: My code works, But I cant say anything on the "Quality" of 
> the code.
> -- 
> Change the rules, or the rules will change you
>     ---------Kumaranasan
> 
> _______________________________________________
> grass5 mailing list
> grass5 at grass.itc.it
> http://grass.itc.it/mailman/listinfo/grass5
>