If the quadprog problem is infeasible, how i can solve it with a certain tolerance?

15 views (last 30 days)
F1=[0.2700; 0.2300; 0.2500; 0.2400; 0.2400; 0.2400; 0.2500; 0.2500; 0.2400; 0.2500; 0.2400; 0.2600; 0.2600; 0.2300; 0.2600;
0.2700; 0.3100; 0.3200; 0.3500; 0.3700; 0.3700; 0.4000; 0.4100; 0.4600; 0.4800; 0.4400; 0.5000; 0.4800; 0.5000; 0.4800;
0.4800; 0.4700; 0.4500; 0.4700; 0.4500; 0.4700];
F2=[1.8900; 1.8600; 0.3700; 0.7400; 0.1400; 1.8000; 0.6100; 1.6900; 0.4000; 0.3000; -0.9700; 0.4800; 1.3900; -1.7800; -2.3000;
-0.8300; 0; -0.1800; 1.8400; 0.1400; -1.4400; -0.0700; -0.1800; 0.8000; 1.9000; 2.1700; 0.6800; 1.3100; 3.9500; 0.6600; -0.2500;
1.1700; 0.9500; 1.4700; 1.4400; 1.4400];
F3=[0.8100; 1.3600; 2.1100; -2.4200; 2.6500; 0.2700; -0.4200; 3.7900; -0.7900; 2.0800; -1; 1.2200; 3.3800; -2.7100; -4.3900; 1.2900;
1.6300; -2.4700; 3.2800; 4.0800; -2.4500; 2.2500; -3.6500; 1.4600; 2.5800; 3.8800; 2.9500; 2.9400; 3.9500; 2.3100; 3.3000; 0.2500;
4.2300; -0.3600; 4.3800; 1.9300];
F4=[2.0700; -1.4100; 3.3800; -3.3800; 4.0700; 0.7700; 0.4400; 4.3600; 2.0900; 1.2800; -3.0300; 3.3500; 2.9300; -1.4500; -4.5100;
0.8600; -0.4200; -3.0900; 2.6900; 4.8700; -1.0200; 0.5200; -3.9500; 1.2500; 0.9200; 4.0700; 2.2600; 1.8100; 2.2700; 4.6800; 5.8600;
2.5700; 2.7200; -2.7300; 4.1200; 1.1800];
F5=[0; 0.8600; 5.1000; 2.1200; 1.5800; -1.3600; 0.4000; 8.8600; -0.2700; 3.7200; -2.1900; 7.6600; 4.7100; -3.6200; -3.0100; 4.9100;
-4.1100; -1.1100; 5.2900; 2.9600; -4.0800; 4.0800; -3.8400; 0.5200; -0.6000; 1.9700; 3.9400; 4.2000; 2.5600; 1.2100; 5.2200; -3.9700;
3.0300; -0.5700; 0.5800; 3.0800];
F6=[-0.2600; 4.5100; 12.3100; 15.8200; 4.1000; -2.7200; 5.7800; 3.1000; -3.4500; 2.2900; -14.3100; 6.9000; 11.3500; 2.3000; -5.3700;
4.7600; 2.5300; 2.9800; -1.9900; 1.6900; -2.4900; 2.3900; -5.6600; 0.9800; -6.1000; -2.8200; 6.9800; 5.0800; -3.8800; -4.4000;
7.2300; -3.8500; 1; -4.8700; 4.9300; 4.8700];
R=[2.197; 2.707; 3.189; 3.191; 1.62; 0.07845; 1.318; 3.396; -0.9583; 0.6649; -1.762; 2.268; 2.688; -1.746; -4.104; -0.8574; 0.1571;
-1.031; 1.917; 0.9404; -0.7029; -1.181; -1.434; 0.08219; 0; 2.197; 1.682; 1.027; 1.878; 1.091; 1.461; 0.2271; 0.9026;
-0.9091; 2.829; 1.644];
numberOfAssetClass=6;
T=length(F1);
Fmatrix=[F1,F2,F3,F4,F5,F6];
F1mean=mean(F1);
F2mean=mean(F2);
F3mean=mean(F3);
F4mean=mean(F4);
F5mean=mean(F5);
F6mean=mean(F6);
FmeanVector=[F1mean; F2mean; F3mean; F4mean; F5mean; F6mean];
Rmean=mean(R);
H=NaN(numberOfAssetClass,numberOfAssetClass);
for i=1:numberOfAssetClass
for j=1:numberOfAssetClass
H(i,j) = 0;
for t=1:T
H(i,j) = H(i,j) + (Fmatrix(t,i)-FmeanVector(i))*(Fmatrix(t,j)-FmeanVector(j));
end
H(i,j) = H(i,j)/T;
end
end
H=H*2;
f=NaN(numberOfAssetClass,1);
for i=1:numberOfAssetClass
f(i) = 0;
for t=1:T
f(i) = f(i) + 2*(-R(t) + Rmean)*(Fmatrix(t,i)-FmeanVector(i));
end
f(i) = f(i)/T;
end
Aeq=Fmatrix;
beq=R;
lb = zeros(6,1);
ub = ones(size(lb));
b=quadprog(H,f,[],[],Aeq,beq,lb,ub);

Accepted Answer

Matt J
Matt J on 24 Jul 2021
Edited: Matt J on 24 Jul 2021
It never really makes sense to have more equality constraints (here 36) than you do unknowns ( here 6). In order for the problem to be feasible, some of the constraints would have to be redundant.

More Answers (0)

Categories

Find more on Quadratic Programming and Cone Programming in Help Center and File Exchange

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!