Correction of efficient solutions in a multistage stochastic optimization problem

A dynamical linear optimization problem with a random objective vector, which arises, in particular, in financial analysis, is considered. Optimal solutions are sought in accordance with two criteria: the maximality of the expectation anti the minimality of the valiance of an objective function. It is assumed that additional information about the distribution of the objective vector is obtained during the control process. The question about the change of efficient solutions depending on the realized observation is studied. The concept of adjoint stochastic optimization problems is introduced. The relation between the problems of efficient observation anti evaluation is studied. The research employs methods of the theory of control under incomplete information [1-3].