MATLAB: Using fminunc with blackbox

#blackboxfminuncMATLABminimizeroptimization

A blackbox code (obj_b3.p) defines a function that takes a vector of length 2 and returns some value based on that function. I want to find the global minimizer, and instead of guess and check I want to use fminunc (have never used before, have looked at doc and help and tried googling examples but they all have known functions that they make anonymous). To do this I intend on using a lot of x0 (guess) values.

how do i set up the basic code?

what I've tried:

v = [1;1] % creates a vector of length 2 (blackbox code won't accept a 1,2, only a 2,1)

x0=[0;0]; % Initial guess for fminunc

v = fminunc(obj_b3,x0)

matlab returns that there aren't enough inputs.

Can someone post a simple example of how to do this using a blackbox.m file (hope its obvious that the blackbox.p file does not allow one to actually see the code to evaluate).

Thanks!

Best Answer

fun = @obj_b3;
x0 = [0;0];
x = fminunc(fun,x0)

You might want to consult the documentation on optimizing a simulation or ODE in case your black-box function does not distinguish between nearby values very well.

Alan Weiss

MATLAB mathematical toolbox documentation

Related Solutions

MATLAB: Callback within the class

It seems like all the properties other than b and h should be dependent. This would allow you to get rid of the update method. If you really want to go the update route, you could write a set method for b and h which will call the update method.

Does the following work for you?

classdef form<handle
  properties
    type = 'square'
    b
    h
  end
  properties (Dependent)
    A
    Iy
    Iz
    Wy
    Wz
  end %properties

  methods
    function f = form(b,h)
      f.b=b;
      f.h=h;
    end
    
    function val = get.A(obj)
      val= obj.h*obj.b;
    end
    
    function val = get.Iy(obj)
      val= obj.h^3*obj.b/12;
    end
    
    function val = get.Iz(obj)
      val= obj.b^3*obj.h/12;
    end
    
    function val = get.Wy(obj)
      val= obj.h^2*obj.b/6;
    end
    
    function val = get.Wz(obj)
      val = obj.b^2*obj.h/6;
    end
  end % events

end%class

You could also go with the "ugly" update method:

classdef form<handle
  properties
    type = 'square'
    b
    h
    A
    Iy
    Iz
    Wy
    Wz
  end %properties
  methods
    function f = form(b,h)
      f.b=b;
      f.h=h;
    end
    function set.b(obj, val)
      obj.b = val;
      obj.update;
    end
    function set.h(obj, val)
      obj.h = val;
      obj.update;
    end
    function obj = update(obj)
      obj.A= obj.h*obj.b;
      obj.Iy= obj.h^3*obj.b/12;
      obj.Iz= obj.b^3*obj.h/12;
      obj.Wy= obj.h^2*obj.b/6;
      obj.Wz= obj.b^2*obj.h/6;
    end
  end % events
end%class

MATLAB: Defining a stopping criterion for fmincon

Hi,

Yes, you can use optimset. You can directly use fmincon istead of createOptimProblem(but both work the same).

Hope this helps!

Best Answer

Related Solutions

MATLAB: Callback within the class

MATLAB: Defining a stopping criterion for fmincon

Related Question