Linear least sqaure combined with golden section search

Hello ,

I am following a procedure in a paper in which the observed data(q,t) is fitted with the following exponential approximation function;

q=a+b*exp(-t/c)

The method used in the paper is linear least square fit combined with "direct search" on the nonlinear parameter "C". Here the direct search is related with the optimization algorithm called golden section search.

Can anyone please explain how can I combine two methods in the Matlab to find all the three variables (a,b,c)?Can introduce me the functions for each method that can be used and steps to find the unknown variables by combining two methods ?

Thanks in advance ,

Ehsan

[John D'Errico]

This is a pretty basic problem to solve using any optimization tool. There is little need to use the specific method indicated. For example, the curvefitting toolbox will have no problems solving this.

Answer 1

[Alan Weiss]

This Optimization Toolbox example shows how to use a general nonlinear minimizer fminunc along with separate solution of the linear parameters.

You could just as easily use fminbnd instead of fminunc. Internally, fminbnd performs a modified golden section search. This can work as long as the nonlinear parameter is scalar (one-dimensional).

Good luck,

Alan Weiss

MATLAB mathematical toolbox documentation

Answer 2

[Matt J]

It sounds like the approach you describe is the one implemented here.