Statistical verification of optimality conditions for stochastic programs with recourse (Q1178441)

The papers deals with a solution algorithm for stochastic linear programming problems with recourse and continues an earlier paper of the authors [Math. Oper. Res. 16, No. 3, 650-669 (1991; Zbl 0746.90045)] in which a numerical (stochastic decomposition) algorithm was introduced. The aim of this paper is to suggest a termination rule for this algorithm. The presented termination rules are based on Kuhn-Tucker optimality conditions and on Lagrangian duality. This subdifferential properties of the optimized functions and corresponding statistical estimates are employed as well. An illustration example is presented at the end of the paper.