Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!news.mathworks.com!tank.news.pipex.net!pipex!lade.news.pipex.net!pipex!datcon!dmd
From: dmd@datcon.co.uk (David Drysdale)
Subject: Re: What is conjugate?
Message-ID: <1996Feb1.123917.14677@datcon.co.uk>
Organization: Data Connection Limited
X-Newsreader: TIN [version 1.2 PL2]
References: <DM3E9E.EE7@uns.bris.ac.uk>
Date: Thu, 1 Feb 1996 12:39:17 GMT
Lines: 24

M.J.Ratcliffe@bris.ac.uk wrote:
: In a similar vein, does anybody have a really good explanation of 
: Levenberg-Marquardt. Once again most references seem to be written in 
: Ancient Greek; as far as I understand (which is not very far!) LM is a 
: sort of blend between gradient descent and conjugate gradients, depending 
: on some sort of confidence criterion.

For Levenberg-Marquardt, I found that the description in
Numerical Recipes was fine -- good enough for me to implement
and use for my thesis.  As I recall, you have to read stuff from
the preceeding sections -- you can't just read the L-M section
on its own.

In the unlikely event that you haven't heard of it, the book is:
"Numerical Recipes in {C, Pascal, Fortran}: The art of scientific computing"
Press, Teukolsky, Vetterling, Flannery  
Cambridge University Press 1992 ISBN 0-521-43108-5 (C version, hardback)

: Help!

: Thanks 
: Max

David Drysdale
