On duality with generalized convexity

A multi-stage stochastic decision model with general reward functional is considered. We get statements with respect to criteria of optimality and to the existence of optimal strategies which generalize well-known theorems of classical theory of dynamic programming.