Finding the local maximum in a parameter space

Hi,

I have code which generates a signal that is a function of time, and three parameters.

What I want to do is find the parameter values such that the 100KHz component of the signal is maximised.

Currently I'm doing it the manual way, where I limit the number of values to 4 for each parameter, pick a set of parameter values, take the FT of the resulting signal, and measure how large the 100KHz component is. Repeat 4*4*4 times!

Could you please advise as to how I could implement this effectively. I'm looking to automate the process and be able to scan the parameter space more comprehensively (rather than limit to 4 values per parameter).

Thanks for your input.

Answer 1

[Matt J]

You could use FMINSEARCH or one of the functions in the Optimization Toolbox, if you've got it. These algorithms are iterative and require an initial guess. Your sweep of the 4x4x4 discretization of the parameter space is still a good way to find one.

[Matt J]

A B Commented

Thanks.

In that case I'll just implement the sweep method.

Do you have any suggestions on how I could effectively present the 4D data (3 parameters + the signal value)?

[Matt J]

In that case I'll just implement the sweep method.

Why?? What made you give up?

Do you have any suggestions on how I could effectively present the 4D data

Depends what "effective" means here.

[A B]

Why?? What made you give up? I want a surface map covering the parameter space, and FMINSEARCH only provides the optimal set of values. So on reflection it seems to me that I need to scan over the space myself.

Depends what "effective" means here. The main reason I'm plotting the surface map is to visualise how each parameter effects the value of the maximum. I think that I'll use a 3D plot, and will only be able to present two parameter axes per plot.

Answer 2

[A B]

Relocated to Comment by Matt J