From: "Bruno Luong" <b.luong@fogale.findmycountry>
Newsgroups: comp.soft-sys.matlab
Subject: Re: Non-linear optimization
Date: Tue, 5 Mar 2013 07:26:08 +0000 (UTC)
Organization: FOGALE nanotech
Lines: 18
Message-ID: <kh46mg$s0t$>
References: <kh2m44$4eh$> <kh32ds$hrn$>
Reply-To: "Bruno Luong" <b.luong@fogale.findmycountry>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
X-Trace: 1362468368 28701 (5 Mar 2013 07:26:08 GMT)
NNTP-Posting-Date: Tue, 5 Mar 2013 07:26:08 +0000 (UTC)
X-Newsreader: MATLAB Central Newsreader 390839
Xref: comp.soft-sys.matlab:790321

"Matt J" wrote in message <kh32ds$hrn$>...
> "Toan Cao" <> wrote in message 
> If you know a global lower bound on F(x), say F_low, then the minimization problem is equivalent to
>  min  f(x)'.f(x) 
> where
>  f(x)=F(x)- f_low
> So, you could apply Levenberg-Marquardt and/or Gauss-Newton to the reformulated problem. 

I don't think it is a good suggestion. (1) It make the code difficult to handle since f_low needs to be known. It squares the conditioning of the original problem, and thus all kinds of numerical difficulties become more prominent.

When levenberg-Markquardt or pseudo-newton method is developed, f(x) is usually taken as the local Jacobian of F. And this approximation is generally known and studied. The approximation can be applied on any F (only assumed to be differentiable).