Fseminf can't be used if the constraints' dimsion is higher than 2?

Hello

What can you do if the constraints have higher dimension than 2?

Here

https://www.mathworks.com/help/optim/ug/two-dimensional-semi-infinite-constraint.html

is the 2D example, but in my problem I have products like ... . Not exactly this but the point is that I have n number of and I have to multiply all of them(not only 2 of them).

[Richárd Tóth]

maybe it's possible to create a 1D or 2D constraint matrix from the N-D constraint matrix? I think the order of the elements doesn't even matter, right?

[Torsten]

It won't be a problem if you use "ndgrid" on the w1,...,wn.

[Richárd Tóth]

It is a problem, becouse by using ndgrid I got N-dimensional matrices, so the 's are all N-dimensional, but fseminf is coded to handle only 1 or 2D constraints. At least that's what I see in the findmax.m file. My code runs fine if my constraints are 1D or 2D, but when I move to 3D i get this error:

Array dimensions must match for binary array op.
Error in findmax>findmax2 (line 104)
crmax = (data>[data(1,:)-1;data(1:m-1,:)]) & (data>=[data(2:m,:);data(m,:)-1]) & ...
Error in findmax (line 64)
    [gt, d] = findmax2(data, flag);
Error in semicon (line 47)
    [gt,d]=feval(@findmax, T{i}, flag<0);
Error in fseminf (line 477)
        ctmp = feval(confcn{3},xnew,LAMBDA,NEWLAMBDA,OLDLAMBDA,...

The data passed to findmax has a size of 32x32x63.

[Richárd Tóth]

Seems like transforming the constraint matrix into a lower dimension form doesn't work, I tried adding this line after K1 is assigned in the two dimensional example

K1=K1(:);

and the solution I get from fseminf is not a valid solution even though the matrix elements are the same.

[c,ceq,K1] = mycon(x,[0.5,0.5]);  % Sampling interval 0.5
max(max(K1))

returns a positive number

[Lucas Schaare]

Hello,

were you able to solve this problem with 3D constraint matrix?

Thank you in advance.

Lucas

Answer 1

[Bruno Luong]

If you have one inequality (doesn't matter how the lhs is computed, with finite or infinite series or products), then you have ONE constraint.

[Richárd Tóth]

I have more than 1 inequality, all of them looks similar to the others, like this

phi_phi2,phi3 correspond to 's