Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!bb3.andrew.cmu.edu!newsfeed.pitt.edu!gatech!newsfeed.internetmci.com!news.intersurf.net!multiverse!ragnarok.oar.net!malgudi.oar.net!utnetw.utoledo.edu!news
From: dweddin@uoft02.utoledo.edu
Subject: Backprop for RBF code trade!
Message-ID: <DoAuu9.Ms5@utnetw.utoledo.edu>
Lines: 30
Sender: news@utnetw.utoledo.edu (News Manager)
Organization: University of Toledo
Date: Thu, 14 Mar 1996 22:13:35 GMT



Hey all,

I'm trying to learn Backprop neural networks.

I understand what I am doing, but i find it very
difficult to train the network to do even the simplest
XOR problem.  I always choose my ETA value too high
or too low and my network can never train. it's almost
like trying to balance a knife on its edge.

so here's what i'm offering:

i am looking fo somebody out there has the source code for a Backprop 
network that can solve the XOR problem (i just plug in the data and it
will train to the correct solution nomatter what starting point)

i want a very simple bear bones code that is preferably written in
FORTRAN (fortran is easy to disect for learning purposes).

in return, i'll write a similar program for Radial Basis Function (RBF)
neural networks for anybody interested in learning these, since that's
what i've been doing all my research in.

Any Takers? Just Email me.

Don Wedding


