Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!swrinde!zaphod.mps.ohio-state.edu!uakari.primate.wisc.edu!aplcen!haven!uvaarpa!mcnc!rti!dbn
From: dbn@rti.rti.org (Daniel B. Nissman)
Newsgroups: comp.ai.neural-nets
Subject: Conjugate Gradient methods for feedforward nets
Keywords: backprop,quickprop,training algorithms
Message-ID: <4028@rtifs1.UUCP>
Date: 22 Aug 90 19:06:17 GMT
Distribution: comp
Organization: Research Triangle Institute, RTP, NC
Lines: 11

I would like a pointer to the application of the conjugate gradient method
to training feedforward networks.  A symbolic description of how to actually
implement this method in such a network would be greatly appreciated.  Also,
how does this differ from Fahlman's Quickprop algorithm?  Pros and cons of 
using one method over another would be desireable as well.

A related question is:  which of these methods works best in scale-up problems
and why?

Many thanks,
Daniel Nissman