passing extra params to fmincon: anonymous function vs old method

Hi,

I was trying to pass extra parameters to fmincon function. I did it in two different ways: anonymous functions and old method, and I got 2 different results.

When I use anonymous functions, Matlab said that the initial point is already local minima. But when I use the old method, the optimization proceeds and run until the objective function x reaches f(x) = 0.0468.

Why is it like that? With this situation I prefer to use the old method to pass the extra parameters. However, the old method does not allow me add Output function to fmincon.

Thanks in advance. Amad

----------------

Anonymous function

Code:

myfun = @(k)funObj(kSearch,x0,ti,tP);
[kSearch,resnorm] = fmincon(myfun,kSearch,[],[],[],[],lb,ub,[],optionLSQ);

Output:

                                Max     Line search  Directional  First-order 
 Iter F-count        f(x)   constraint   steplength   derivative   optimality Procedure 
    0     51    0.0502957            0

Local minimum found that satisfies the constraints.

--------------------------

Old method:

Code:

[kSearch,resnorm] = fmincon(@funObj,kSearch,[],[],[],[],lb,ub,[],optionLSQ,x0,ti,tP);

Output:

                                Max     Line search  Directional  First-order 
 Iter F-count        f(x)   constraint   steplength   derivative   optimality Procedure 
    0     51    0.0502957            0                                         
    1    114    0.0502503            0     0.000244       -0.752         17.2   
    2    172    0.0495527            0      0.00781       -0.551         34.1   
    3    232    0.0494097            0      0.00195       -0.408         10.8

Answer 1

[Alan Weiss]

The error is in this line of code:

 myfun = @(k)funObj(kSearch,x0,ti,tP);

funObj does not have a variable k, so no matter what value of k is chosen, funobj always returns the same value. You probably meant to write

 myfun = @(kSearch)funObj(kSearch,x0,ti,tP);

Alan Weiss

MATLAB mathematical toolbox documentation

[Amad]

Hi Alan,

Thank you so much, your solution is really helpfull.