Duality theorems for separable convex programming without qualifications (Q2359787)

It is known that with respect to Lagrange duality for separable convex programming problems strong duality in sense of ``sup=inf'' holds. That means that there is no duality gap but one cannot guarantee that optimal Lagrange multipliers exist. For this additional constraint qualifications are necessary. In the main part of the paper it is shown that separable convex inequality systems always satisfy a weak constraint qualification (the so-called closed cone constraint qualification Q-CCCQ) which is necessary and sufficient for a generalized Lagrange-type duality in quasiconvex programming. Hence, a strong duality assertion in sense of ``inf=max'' without constraint qualification can be derived for this duality concept for separable convex optimization problems. A further duality theorem and a necessary and sufficient optimality condition for a separable convex programming problem where the Slater condition fails are provided.