MATLAB: Optimal way to parse parameters (fmincon).

fminconParallel Computing Toolbox

Hi there,

My question is very straightforward: for a function that depends on multiple parameters, say f(a,b,c), what is the optimal way to parse the fixed parameters (in this example "b" and "c") when calling fmincon.

I found this link with three suggestions: http://www.mathworks.com/help/optim/ug/passing-extra-parameters.html#bripxyy

However, in my experiments using "global" actually improved performance slightly.

To give you a background, I ran several Monte Carlo experiments (10MM) where each run calls fmincon once. Thus, even a slight performance improvement might reduce the computational time substantially.

Thanks,

Best Answer

I honestly don't think we can KNOW the truly optimal way. It may easily depend on release, since TMW can change things at their whim, and they are not going to worry about how a change in perhaps the parser might impact something subtle like this.

It will SURELY depend on the exact implementation of how you pass the variables. Again, subtle differences might make subtle changes in the time.

I'd suggest you run the test, and pick the way that is fastest for you, on your problem.

Related Solutions

MATLAB: How to setup upper and lower boundaries depending on guess in lsqnonlin function

I think you misunderstand how these tools work, and the meaning of a boundary.

The bounds (lb and ub) are fixed, whatever you set them at the start. They cannot change for each iteration, based on your whim, or on the weather outside.

Next, you cannot have a constraint like c>0. You can only have a constraint of the form c>=0, and even that may be violated by a tiny amount (TolCon).

Perhaps you meant something else. If so, then you need to explain your question far more clearly.

MATLAB: Expressing non-linear conditions in an NLP optimization problem

You would need to use ga() for this, with the nonlinear constraints implemented in the nonlcon input argument

x = ga(fun,nvars,A,b,[],[],lb,ub,@(x)nonlcon(x,a0,b0,d0),1:2)

And yes, you would pass extra parameters to nonlcon in the way that you read. You would pass a vector of a0 values and similarly for the other parameters.

function [c,ceq]=nonlcon(x,a0,b0,d0)
   %y1,y2=ode45....
 c=C*sum(y1)+D*sum(y2)-100;
 ceq=[];
end

Best Answer

Related Solutions

MATLAB: How to setup upper and lower boundaries depending on guess in lsqnonlin function

MATLAB: Expressing non-linear conditions in an NLP optimization problem

Related Question