[comp.lang.icon] NeuralNets: help requested

eric@ai.leeds.ac.uk (E S Atwell) (10/03/90)

I mailed icon-project@edu.arizona.cs asking for any Icon implemetations of
Neural Networks; the reply came that the Icon project team didnt have any,
but other subscribers to icon-group might.  Is there anyone else out there
interested in neural nets, who can help us?    cf the following:
 
From eric Mon Oct  1 12:22:38 1990
Subject: Neural Networks
 
Are there any Icon implementations of neural networks that you know of?
I have an eural network simulator package for Sun workstations (NeuralWare)
which requires input and outputs to be specified as numerical values in a file;
I am experimenting with language processing (specifically, parsing), so need
to convert my training data (a Treebank or Corpus of syntax-trees for parsed
English text samples) into numerical form.  I have started with varoius Icon
reformatting programs to transform the trees into required input/output training
format; but it struck me that it would be much neater to do the whole thing
in a single icon program, rather than Icon programs interfacing to a 'black
box' package.  Any suggestions?  I couldnt find anything suitable in the
library programs availale by FTP from edu.arizona.cs.
Thanks for any help you can offer.  The AI Division here at Leeds has only
recently been established (over the past year we have swelled from 5 to 10
teaching staff, and we're just setting up new courses including a BSc in
Cognitive Science), so this may be a good time to impress my new colleagues
with the versatility and elegance of Icon (opinion is split between Pop-11,
Prolog, Lisp, and C at the moment).
 
 Eric Steven Atwell
                Centre for Computer Analysis of Language And Speech (CCALAS)
                Artificial Intelligence Division, School of Computer Studies
 phone: +44 532 335761                                      Leeds University
 FAX:   +44 532 335468                                         Leeds LS2 9JT
 JANET: eric@uk.ac.leeds.ai                                          England
 EARN/BITNET/ARPA: eric%leeds.ai@ac.uk
 

shafto@EOS.ARC.NASA.GOV (Michael Shafto) (10/04/90)

I played around with neural nets in Icon a few
years ago.  Bottom line:  the flexibility of
Icon in manipulating lists and other data structures
was a big plus at development time, but you would
want to be in C for serious work due to the hard
requirement for raw speed (plus the complexity of
your program is not a real driving factor in this
kind of work).

Mike