Imposing strict inequality

Dear Friends,

I’m trying to solve this problem in CVX by one-dimensional search over variable t as

cvx_begin  SDP

variable W( NS , NS ) hermitian

maximize t

subject to

W           > 0;
trace(W*U2) > t*trace(W*E2);
trace(W*U1) > t*trace(W*E1);


The solvers in CVX consider the “<” as “\leq”. As a result, the trivial solution \bf{W}=0 is returned by CVX which is not favorable. Does anybody know how to impose strict inequality in CVX?

Thank you a lot.


Read Strict inequalities in the documentation.

Thank you Michael. Adding an offset to the right side of the inequality solved the problem.