Geometric program with posynomial > 1 I have an optimization setup where I can represent all objectives/constraints as posynomials. Unfortunately one constraint has the form $g(x) > k$ where $g$ is a posynomial: $$ \operatorname{minimize}_{R,G,a,C_1,C_2} R + \frac{1}{G} \left( k_1 + \frac{C_1}{a C_2}(1+G R) \right) k_2 \\\ \mathrm{subject\,\,to} \\\ a \left(1 + \frac{C_1}{C_2}(1+ G R)\right) > k_2 \\\ \vdots $$ For a valid geometric program, the constraints must have the form $g < k$. Is there any hope to transform this into a valid geometric program?

I've not enough reputation to join the comment thread above, but there's a matlab script which implements Maranas and Floudas's 1997 algorithm to provide a (slowish) global optimum for an SP which you may find useful.