Choose samples from statistical distribution for fmincon optimization

Hi everyone,

I have a problem with several variables. Each of these variables (both opimized and non optimized) has statistical distribution. How could I set fmincon to select optimization samples randomly from their statistical distribution?

Answer 1

[Jeff Miller]

It is probably a better idea to select the samples randomly and then have fmincon optimize for the given sample.

If you must take samples randomly during the minimization process, you can do that inside the user-defined 'fun' that fmincon tries to minimize. But this almost always causes problems: fmincon gets very confused when the values returned by 'fun' vary somewhat non-systematically due to the random sampling in fun.

Another option is to sample systematically rather than randomly inside fun. For example, sample each variable in evenly-spaced probability steps within its statistical distribution.

[soheil radfar]

Dear Jeff,

Thanks for your answer.

But, I do not know how to implement this within my codes.

I attached the codes.

Please take a look at them if you have time.

I think there are also errors in defining of nonlinear constraints.

Thanks for your consideraion.

[Jeff Miller]

Sorry I cannot be more helpful, but I really have no idea what problem you are trying to solve. For example, what function are you trying to minimize? And why do you need to take random samples during the minimization process?

You should probably ask a new question and provide a lot more explanation of what you are trying to do. Maybe someone will recognize the problem and be able to help.

[soheil radfar]

Thanks Jeff,

As you said, I asked my question in more details in a separate question:

https://nl.mathworks.com/matlabcentral/answers/486207-problems-with-fmincon-for-design-optimization