Non negativity constraint in fmincon

Hello everyone..

I have a question concerning fmincon. I am trying to optimize a portfolio but i cannot have short sales, so i want to add a non-negativity constraint in fmincon. I tried to do it using lb=0 but for some reasons, and i do not find which it does not accept my constraint and still gives me as output some negative numbers...

Here are the part of the code where i put this and that does not work:

AnnualReturn=((ones(1,10)+ExpectedReturnUD).^12)-ones(1,10);
Cov=CovUD;
W0=zeros(10,1);
lb=zeros(10,1);
ub=ones(10,1);
Aeq=[ones(1,10);AnnualReturn];
beq=[1;0.12]
WEP=fmincon(@(W) SigmaFct(W,Cov),W0,[],[],Aeq,beq,lb,ub);

or

for i=1:100
 [W(i,:)]=fmincon(@(W) SigmaFct(W,Cov),W0,[],[],[ExpectedReturnSample;ones(1,10)],[rbar2(1,i),1],zeros(10,1),ones(10,1));
end

If anyone has an answer for me, that would really help me!!! thanks a lot!

Answer 1

[Alan Weiss]

You did not pass an options structure, so you are using the fmincon active-set algorithm. This algorithm does not strictly satisfy bound constraints; see Iterations can Violate Constraints.

I suggest you use the interior-point or sqp algorithms, each of which strictly satisfies bound constraints.

Alan Weiss

MATLAB mathematical toolbox documentation