From octave-sources-request at bevo dot che dot wisc dot edu Tue Nov 27 08:27:16 2001
Subject: Re: Levenberg-Marquardt algorithm
From: Francesco Potorti` <F dot Potorti at cnuce dot cnr dot it>
To: Matthias Jueschke <matthias at physik3 dot gwdg dot de>
CC: octave-sources at bevo dot che dot wisc dot edu,        Paul Kienzle <pkienzle@users.sourceforge.net>
Date: Tue, 27 Nov 2001 15:27:07 +0100

   I tested both functions with some of the data sets Paul mentioned. My
   version was much slower and even one time the fit was not as good as from
   leasqr.m. 

Wonderful!  Now we have a GPL'ed, tested non-linear optimisation
routine for Octave!

I suggest you write some lines explaining what tests you applied and
how, and send them to Paul, so he can incorporate them in the sources
for future reference (or even in the manual).
	     
	     Just the message each time is disconcerting:
   warning: inverse: matrix singular to machine precision, rcond = 0
   
No one so far explained this error message.  Probably no one completely
understands the code.  I suggest that this fact be also mentioned in the
comments at the beginning of the source file, with an invitation to
knowledgeable people to investigate and report the cause.