# How to solve the DCP error. Disciplined convex programming error: Invalid quadratic form(s): not a square

> Disciplined convex programming error: Invalid quadratic form(s): not a square.

                    cvx_begin sdp quiet
variable alpha1 nonnegative
variable A(1,U) complex
variable B nonnegative
minimize rho/2*pow_pos(abs(A(1,u)-H(:,u)'*w1(:,u)+g(1,u)),2)+rho/2*pow_pos(abs(alpha1-delta1+f),2)
subject to
for u=1:U
A(1,u)==0;
B>=sum(pow_pos(abs(A),2));
gamma*(B+sigma)-2*real(w1(:,u)'*H(:,u).*A(1,u))+pow_pos(abs(H(:,u)'*w1(:,u)),2)<=0;
alpha1*(B+sigma)-2*real(w1(:,u)'*H(:,u).*A(1,u))+pow_pos(abs(H(:,u)'*w1(:,u)),2)<=0;
end
cvx_end

alpha1*(B+sigma) has a non-convex multiplication of variables alpha and B`. Have you proven this is a convex constraint? I will assume it is not, unless shown otherwise.

This picture is the constraint condition of objective function.
The eigenvalue of the Hesse matrix of the objective function is greater than or equal to zero.
Therefore, I think it is a convex constraint.
I’m not familiar enough with CVX to get it to work correctly.
How do I express this constraint condition?
Thank U so much.

You mention eigenvalue of Hessian of objective. Then you say convex constraint. You need to prove the constraint is convex. Perhaps \alpha is not supposed to be an optimization (CVX) variable?