Finite convergence in stochastic programming (Q4026249)

[For the entire collection see Zbl 0744.00037.] A multistage stochastic programming problem or, more precisely, a non- anticipative and generally nonlinear multistage stochastic programming problem is considered. The aim of the paper is to suggest a solution method for the above mentioned optimization problem. The method is based on the boundary of Lagrangian function and the corresponding theory of saddle point. To this end, first, two time vector functions are defined. They correspond to a primal and dual solution. Furthermore, the assumptions are given under which the feasible set of primal solutions is achieved in the final time. At last, the assumptions are presented under which the similar result also hold for the optimal solution. The presented results as well as the corresponding proof are based on previous work of the author.