Re: Change objetive function in Non-Linear Fit
- To: mathgroup at smc.vnet.net
- Subject: [mg25726] Re: [mg25673] Change objetive function in Non-Linear Fit
- From: "Mark Harder" <harderm at ucs.orst.edu>
- Date: Thu, 19 Oct 2000 04:35:56 -0400 (EDT)
- Sender: owner-wri-mathgroup at wolfram.com
Miguel,
I don't think you can change the objective function in NLFit, which was
written specifically to minimize the Least-squares error between the model
and the data.
For your purpose, use the FindMinimum function. Write a Mathematica
function that takes symbolic parameters of your f & g functions & the list
of x values, calculates the matrix derived from these function values, and
returns the value of the determinant of that matrix. By default,
FindMinimum will iterate to a minimum of the determinant over parameter
space by a method that requires calculation of derivatives of the objective
function. If it can't do this, there may be problems, so you can direct it
to use a derivative-free method by setting the Options: Method->Gradient,
Compiled->False. Or you can supply your own gradient function using the
Gradient-> option. See the Mathematica documentation for the details.
-mark harder
-----Original Message-----
From: Miguel Mora <morafonz at yahoo.com.mx>
To: mathgroup at smc.vnet.net
Subject: [mg25726] [mg25673] Change objetive function in Non-Linear Fit
>Hello:
>
>I want to change the objetive function in the
>Non-Linear Fit Function.Can I do this?
>
>I have the function defined is the determinant of a
>Matrix that involves two non linear functions.
>
>If I can't change the objetive function. How can I
>made a non linear fit simultaneosly on two non linear
>functions? I have the functions, f(x), g(x), both
>non-linear, and a set of data (f,g,x). How can I made
>a non linear fit?
>
>Thanks, Miguel Mora
>UASLP
>
>_________________________________________________________
>
>