Invalid quadratic form:not a square

Nonconvex
1

I have always have a problem to solve the following question

cvx_begin sdp quiet
variable W(Nt,Nt,Rx) hermitian        
variables a t211 r111 r211  t212 r112 r212 t121 r221 r121 t122 r222 r122 t21 ra1 ra2 t11 rb2 rb1
sum=0;
for f=1:Rx
    sum=sum+W(:,:,f)
end
min sum
subject to
A=[W(:,:,1)-a*z*W(:,:,2)+r111*eye(2),(W(:,:,1)-a*z*W(:,:,2))*h111;h111'*(W(:,:,1)-a*z*W(:,:,2)),h111'*(W(:,:,1)-a*z*W(:,:,2)-z*a*t211-z*a*sigmma-r111)]
A1==hermitian_semidefinite(3);
cvx_end

error:

Error using ==> cvx.times at 262
Disciplined convex programming error:
    Invalid quadratic form(s): not a square.
Error in ==> Untitled3 at 64
A=[W(:,:,1)-a*z*W(:,:,2)+r111*eye(2),(W(:,:,1)-a*z*W(:,:,2))*h111;h111'*(W(:,:,1)-a*z*W(:,:,2)),h111'*(W(:,:,1)-a*z*W(:,:,2)-z*a*t211-z*a*sigmma-r111)];

please help me

2

Your problem does not satisfy the DCP ruleset. CVX cannot solve problems that do not satisfy these rules—even some convex ones. I strongly suspect your problem is not convex, in which case no amount of rewriting will allow CVX to solve it.

Incidentally, for the benefit of others, you should be saying minimize(sum), not min sum. And it would seem that sum is not a scalar, so your objective function is ill-posed. but I’m afraid that’s moot.