Optimality conditions and constraint qualifications for quasiconvex programming (Q2278888)

The minimization of upper semicontinuous quasiconvex functions \(f\) in a closed convex set is considered subject to nonpositivity of some lower semicontinuous quasiconvex functions, each described by way of a generator, i.e. as the supremum of a family of quasiaffine functions. For essentially quasiconvex \(f\) (i.e. having no local non-global minima) a first necessary and sufficient contraint qualification condition is presented for a duality theorem of Lagrange type, i.e. an exact optimality condition stating that a Greenberg-Pierskalla subgradient lies in a certain convex cone involving the generators. For general quasiconvex \(f\) another necessary constraint qualification condition is obtained for a similar duality result involving a Martinez-Legaz subgradient of \(f\) and another convex cone, which is sufficient when the constraint set is compact. It is finally shown that these results generalize previously known results for convex \(f\).