Re: package optim, expfit

[Stephen Montgomery-Smith]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 04/08/2013 07:10 AM, Olaf Till wrote:
> https://savannah.gnu.org/bugs/index.php?38063
> 
> Has anyone some expertise on this?
> 
> Olaf
> 

I'm not an expert.  But if you look at web pages like
http://octave.sourceforge.net/optim/function/expfit.html or
http://www.statsci.org/other/prony.html, it indicates that expfit is
trying to solve a very ill-conditioned and difficult problem.
Therefore adding large amounts of noise should render any algorithm
useless.  So I think to call expfit "buggy" is going too far.

I guess someone could follow up the references in the second website
and try to write one of those algorithms, and see if it does any
better.  And googling "Prony's method for non-linear exponential
fitting" gives some other references as well.  But it seems to me to
be more of a full fledged project than some minor bug-fix.


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/

iQEcBAEBAgAGBQJRYr0jAAoJEC3xK9GaktgHG20IAKO0jUA7QYPVgJe1teenJd28
DLr8L31ss7tlGEiVwTP0O1pY0u5Fed9mqhdJWFACpzK6wx6i8rmHnSumZ75xzqm7
YGvb4k459z9NNxoYBOa5esD+II/bYOSL+EiPrpW3YTc1XGL62Wf/GzqoKfp/mc3a
3gUw8fcrzBJKRRR7EmyJDg1WDrTOEqwVc4HnpvW7yS5LIxfqihfA/KjHF2lFL8jj
KWbBm8BfzxhLUChDS7+IK1REt+CsbXpDj2Z4JKPzD5MkqHcLq2kuLw9WMfVyD6bL
JJzdigkONR0ip2KQLj34aK+nHbaahOoYuevZyXaQK9igNlFEBo3PRu697CSZmyk=
=I9Bd
-----END PGP SIGNATURE-----