H=(randn(kr,ks)+1j*randn(kr,ks))/sqrt(2); cvx_begin sdp variable F(kr,kr) complex %variable F(kr,kr) variable W(ks,M) variable Z2(kr,kr) variable Z3(kr,kr) symmetric

X=FHW Error using * (line 126) Disciplined convex programming error: Only scalar quadratic forms can be specified in CVX .

This is non-convex. To make it simple, assume H is the identity matrix. Then you are multiplying two matrix variables, which is noon-convex. In one dimension, itâ€™s F*W, with F and W both optimization (CVX) variables, which is non-convex.

F*W

Read Why isn't CVX accepting my model? READ THIS FIRST! .