Shadow price of information in discrete time stochastic optimization (Q2413091)

Given a filtration \(({\mathcal{F}}_t)_{0 \leq t \leq T}\) in a probability space \((\Omega,\mathcal{F},P)\), consider the multistage stochastic optimization problem: Minimize the expectation \(Eh(x(\omega),\omega)\) of a stochastic convex function \(h=h(x,\omega)\) subject to the random decision strategies \(x=(x_t(\omega))_{0 \leq t \leq T}\) with random input vectors \(x_t\), \(0 \leq t \leq T\), adapted to the filtration. Starting with the function \(\phi(z(\cdot)) := \inf_{x(\cdot)} Eh(x(\omega)+z(\omega))\) related to the objective function \(Eh(x(\omega),\omega)\), the problem is analyzed by means of the conjugate of \(\phi\) and the related dual problem. Conditions for the existence of optimal solutions are given. Moreover, a dynamic programming approach for solving the dual problem is provided.