Path: utzoo!utgpu!watserv1!watmath!att!att!linac!pacific.mps.ohio-state.edu!zaphod.mps.ohio-state.edu!uakari.primate.wisc.edu!aplcen!jhunix!ins_atge
From: ins_atge@jhunix.HCF.JHU.EDU (Thomas G Edwards)
Newsgroups: comp.ai.neural-nets
Subject: Re: Using Newton's Method to speed up backpropagation.
Summary: Hmmm
Message-ID: <6925@jhunix.HCF.JHU.EDU>
Date: 19 Nov 90 03:09:11 GMT
References: <86170@lll-winken.LLNL.GOV> <6905@jhunix.HCF.JHU.EDU> <86351@lll-winken.LLNL.GOV>
Organization: The Johns Hopkins University - HCF
Lines: 23

In article <86351@lll-winken.LLNL.GOV> loren@dweasel.llnl.gov (Loren Petrich) writes:
>	I looked in pt.cs.cmu.edu and I was not allowed access to the
>contents of connect/code.

Hmm...there are two possibilities.  1)  That isn't the right address
2)  It isn't there any more.  I'm going to check with R. Scott
Crowder who wrote the code, but until then, I'll send a copy of
cascor1.c to the first five people who request it (Loren gets copy
#1).

>	I had used Quickprop -- and multiple candidate units. Without
>much success. The NN almost always gets stuck with some weights VERY
>big and others small. And when I put in a big weight decay rate, the
>NN usually gets very lousy results.

I won't profess to be an expert at Quickprop.  You might just want to
try good old conjugate-gradient for weight updates.

Otherwise, it is not apparent what is going wrong.  It should be able
to rather quickly solve the two-spirals problem.  Oh well, I'll send
you cascor1.c.

-Thomas Edwards