Construction of decision rules in stochastic approximation problems (Q1058458)

We consider convex stochastic programming problems in which the decision depends on the realization of random parameters of the conditions. Such one-stage problems with an ''observation-decision'' information structure are sometimes called ''stochastic problems with a posteriori decision rules''. In fact, they are infinite-dimensional extremal problems whose study is complicated by considerable computational difficulties. Here, the construction of decision rules for a given realization of the random parameters of the conditions is reduced to solving a finite- dimensional problem. For some particular classes of stochastic problems of practical interest, we present the decision rules in explicit form. With the aid of this approach it is possible to construct operational control algorithms (decision rules instead of iterative procedures).