Are there any other methods to pass value into a predefined function of matlab . (Fmincon problem- passing extra parameters)

I am trying out an optimization problem which involves usage of fmincon.

I am using the command in following way ------> [optival,fval] = fmincon(@objfun,x0,A,b,Aeq,Beq,Lb,Ub,@confun)

My question is how to pass additional parameter into confun()

As per matlab syntax for confun() we can only pass one parameter inside it , which is z. But for my problem I need values of L, Lb, Ub to be passed into confun() for my problem to work. Is there any other way I can pass the values inside the confun()

This is how my confun() looks but it can only take Z . Z has decesion variables like a,b so it cannot store other values like Lb , Ub & L . Besides Lb & Ub are passed in fmincon and cannot be used as a parameter for Z.

function [c,ceq] = confun(z)
% z = [a,b]
% assign local variable name for readability
a = z(1);
b = z(2);


% we needto enter L
xe = xa + L*30;
ye = ya + L*20;

% constraints c <=  0
% lower and upper bounds on xa an xb {20<xa<50 ; 30<xb<70} [Thus ub and lb also needs to be added in this function]
c(1) = lb(1) - xa;
c(2) = xa - ub(1);
c(3) = lb(2) - xb;
c(4) = xb - ub(2);

% constraints ceq = 0, you don't have any equality constraints, but
ceq = []; 
end

Can the values L, Lb, Ub be passed into it by means of some other function inside it?? Is there any way to approach this??

[Stephen23]

The method used by Jan is explained here:

https://www.mathworks.com/help/optim/ug/passing-extra-parameters.html

That page is also linked from the fmincon documentation near the top of the page in the info block "Passing Extra Parameters". See also:

https://www.mathworks.com/help/matlab/math/parameterizing-functions.html

Answer 1

[Jan]

function [c,ceq] = confun(z,L,Lb,Ub)
...
end

Now create an anonymous function to provide the parameters:

L = ...
Lb = ...
Ub = ...
myCofun = @(z) confun(z,L,Lb,Ub);

fmincon(@objfun,x0,A,b,Aeq,Beq,Lb,Ub, myCofun)

Now fmincon calls the cofun with the single argument z and the other arguments are taken from the definition of the anonymous function.

An alternative but less stable approach is storing the parameters in a persisent variable:

function [c,ceq] = confun(z, inL, inLb, inUb)
persistent L Lb inUb
if nargin > 1
    L  = inL;
    Lb = inLb;
    Ub = inUb;
    return;
end
... your code
end

Then you define the values and call confun with 4 inputs, while the first is a dummy. The values of the paramters are stored persistently in the function and you can run fmincon with the standard call.

This has the drawback, that any clear command removes the values and you have to care to set the variables reliably before running the optimization. I prefer the anonymous function.

[Matt J]

See also Passing Extra Parameters - MATLAB & Simulink.

[Ron Herman]

Thanks Matt. I reviewed the link as well saw the answer above. I got the above concept.

[Ron Herman]

One last thing ....Can I apply the same concept for return of more parameters as well.

For example

x1 = ...
x2 = ...
x3 = ...
myobjfun = @(xo) objfun(xo,x1,x2,x3);
fmincon(myobjfun,x0,A,b,Aeq,Beq,Lb,Ub, myCofun)

but I want my some other parameters inside objfun () to also be returned. Is it possible???

[f,ab]=objfun(x0,x1,x2,x3)  % is this allowed???
a1=x1*2+5;
a2=x2*3+7;
a3=x3*4+9;
ab=[a1 a2 a3];
f=a1*a2*a3;   % value of objective function which is used in optimization

The optimization questions usually use only the value of objective function in fmincon which is " f " in this case.

Can I also return more than one parameter back like "ab" in this case????

Usually we do

[f]=objfun(x0,x1,x2,x3)
a1=x1*2+5;
a2=x2*3+7;
a3=x3*4+9;
ab=[a1 a2 a3];
f=a1*a2*a3; 

When fmincon() is used we just return one parameter in objective function. But my reuirement is also to get some more datas inside objective function like " f " as well as "ab ". Can this be done when I am doing a optimisation problem using fmincon()?????

The question in this comment is asked in refrence with fmincon

[Walter Roberson]

No.

[Walter Roberson]

However you can define multiple outputs, and after you have the best x, call the objective function again with the best x and record the additional results

[Ron Herman]

Thank you sir....

Can you give a small example if possible to visualise this better.....??? I dint exactly get the above point