Most work in numerical optimization starts from the convention that the problem to be solved is given in the form (P) minimize f 0 (x) over all x ∈ X, with X ⊂ lR n. But this notion of what optimization is all about may be unnecessarily limiting, both in the kind of modeling it promotes and the computational approaches it suggests. While all optimization eventually boils down to minimizing some function over some set, the formulation (P) says nothing about the mathematical structure of the… CONTINUE READING