How to add user defined function as constraint for optimization?

How to add user defined function as constraint for optimization? Example is given below. The problem that I know is a problem is that I am parsing a constraint to a function that sees that constraint as a variable, which it is not.

x = optimvar("x",3,3);
y = const1;
z = const2;

in loop:

return_val = my_fun(x(i,j),y)
prob.Constraints.Con1 = return_val >= z;

end loop

[Torsten]

This should help:

https://de.mathworks.com/help/optim/ug/fcn2optimexpr.html

[Faruk Šehić]

Thank you very much. I use R2018b version and fcn2optimexpr() has been introduced in R2019b. Is there any alternative for that?

[Torsten]

Yes. Use the solver-based instead of the problem-based approach.

[Faruk Šehić]

Thank you.

Answer 1

[Matt J]

Ultimately, you must use the solver-based approach, like Torsten says. However, if you download this,

https://www.mathworks.com/matlabcentral/fileexchange/74481-prob2matrices-a-selective-version-of-prob2struct

you can still use problem-based tools to set-up the linear portion of the problem, e.g.,

x = optimvar("x",3,3,'Lower',0,'Upper',1);

Constraints.rowsum=sum(x,1)<0.5;
Constraints.colsum=sum(x,2)==1;

p=prob2matrices('Constraints',Constraints);

x=fmincon(@objective,x0,p.A,p.b,p.Aeq,p.beq,p.lb,p.ub, @(x) nonlcon(x,y,z))

function [c,ceq] = nonlcon(x,y,z)

 c=z-my_fun(x,y); %my_fun must be differentiable in x

 ceq=[];

end

[Faruk Šehić]

Thank you.