sci.op-research

Johan,

Thanks for the link. The problem has tens of thousands of variables;
so, I don't think it's small enough. Unfortunately, the right hand
side of the y_w_l <= Sum_j (F_m_j_l * x_m_j) constraint can easily
have 10 or 20 terms. I take it that the linear relaxation is not a GP
because moving everything to one side of the inequality results in
negative coefficients? I have a bound for BIGM that I am planning to
use. Do you think it's worth it for me to learn more about convex
programming to solve this problem? I've heard that even non-convex
problems can be manipulated into convex formulations through
appropriate approximations and transformations.