How can i write::
trace(P*P')
where P is 2*3 matrix
P’ is Hermitian of P
also, P is on of optimization variables
square_pos(norm(P,'fro'))
Depending on how that is used, you might be able to avoid the squaring, thereby improving numerical properties of the problem. For instance,
trace(P*P') <= c
can be formulated as
norm(P,'fro') <= sqrt(c)