Decision variable and function to be minimised are not directly related

Hi,

Is there anyone to help me on this,

I have Five variables(Bounds well known). These five variables after plugging into my model will generate a curve, and upon processing obtained curve,another curve is obtained which is to be matched with experimental curve i.e. I have to minimise error between experimental and analytical curve. So here Decision variables are not directly effecting fitness function.

As per my understanding upon going through matlab help, decision variable should be present in fitness function which doesn't suit to my problem.

Is there any way to optimise using solvers available in matlab?

[Brendan Hamm]

Well if you are trying to minimized the error between the two curves it would appear that you objective function would be something of the form:

min_x { |f(x) - g(x)|||_2 }, i.e. to minimize the norm of the 2 curves. Therefore both curves play a role. If this is not the case, then it is not clear from your question what you mean.

Answer 1

[Matt J]

Seems like a straightforward application of lsqcurvefit() to me. It doesn't matter whether your final curve has a closed form formula in terms of the decision variables. The requirement of lsqcurvefit is that you provide a handle of the model function F(theta,x) and that F is differentiable in the decision variables, theta. It doesn't care how F() is implemented internally.