On approximate solutions and saddle point theorems for robust convex optimization (Q2228362)

Many important results have been established for convex optimization problems without data uncertainty. However, a great deal of attention has been recently focused also on uncertain convex optimization problems for which robust optimization has shown to be an effective method to deal with. An important question is the characterization of robust optimal solutions for such problems. In the last decades a much more interest has been focused on the situation where the robust optimal solution is taken to be exact instead of approximate. The aim of this paper is to study approximate optimal solutions for convex optimization problems with data uncertainty both in the objective function and the constraints. On the other hand, the study of (approximate) saddle point theorems is one of the most interesting areas of optimization since it is related to the optimal solutions of the primal and corresponding dual problems. There exist some results for optimization problems without data uncertainty and for problems with data uncertainty in the constraint functions. But there are practically no results giving any characterizations of approximate saddle point theorems for convex optimization problems with data uncertainty in both the objective function and the constraints, which contributed to another motivation of this paper. Thus, the paper provides new results for robust approximate optimal solutions for convex optimization problems with data uncertainty in both the objective function and the constraints. First, necessary and sufficient optimality conditions for robust approximate solutions are established. Based on these conditions, a Wolfe type robust approximate dual problem is introduced and robust approximate duality relations between primal and dual problems are investigated. Finally, some robust approximate saddle point theorems are obtained. It is also shown that the obtained general results include special cases of some optimization problems considered in the recent literature. Moreover, the propositions are illustrated on simple examples.