Necessary and sufficient global optimality conditions for convex maximization revisited (Q1378644): Difference between revisions

The paper is concerned with global optimization conditions characterizing the solutions of some types of convex maximization problems of the type \[ f(\overline x)\geq f(x)\quad\text{for every }x\in D,\tag{1} \] where \(D\) is a closed convex subset of \(\mathbb{R}^n\) satisfying \(D^0\neq\emptyset\), \(\overline x\in D\) and \(f:\mathbb{R}^n\to \mathbb{R}\) is convex. Four of such criteria are revisited: the condition (HU) of Hiriart-Urruty, the condition (S) of Strekalovski (that is reproved, generalized and modified), the specialization of the Singer-Toland duality (ST) to problem (1) and the reformulation of an optimality condition (CDC) for the so-called canonical d.c.-problems. A discussion on the interconnections between the criteria is carried out and it is proved that each of them follows from each of the remaining ones. In doing this, some new proofs, generalizations and modifications result. The cases studied deal with problems in Euclidean spaces, but most results may be extended to more general spaces. New formulations of optimality conditions are provided for problem (1) with differentiable objective functions along with a specialization to the case where it is convex quadratic. Finally, some new optimality conditions, holding for more general problem classes than (1), are presented.