Decreasing the number of variables in optimization process (SQP - fmincon)

Dear community,

I was wondering if there is anyway to decrease the number of variables during an optimization process going on using MATLAB built-in function (fmincon - SQP)?!

For example assume you have specified 10 variables to start the optimization problem and you want to decrease it to 9 variable (i.e. emit one of the variables based on a cetrain condition) after 20 iterations.

Best regarsd,

Sajjad

Answer 1

[Matt J]

You can use the MaxIterations option to tell the solver to run no more than 20 iterations. Then, you can reformulate the optimization however you wish based on that result and launch a second optimization.

[Sajjad Asefi]

Thanks for the answer, but I don't want to restart the optimization process

I'm just asking about the possibility of changing variable number during the process

[Matt J]

No, you can't. It isn't even clear how that could still be called an optimization process. If the space of unknowns that the function depends on is changing in the course of the iterations, it means that the definition of the function and the location of its minimum are changing, too - the thing you're searching for isn't well-defined in advance.