Stochastic programming duality: \({\mathcal L}^\infty\) multipliers for unbounded constraints with an application to mathematical finance (Q1434072)

Suppose that for a closed convex subset \(C\subset \mathbb R^ n\) and a complete probability space \((\Xi,\mathcal F, P)\) there are given the convex functions \(f_{1i}:\mathbb R^ n\to \mathbb R,\; i=0,\dots ,m_ 1,\) and the functions \(f_{2i}:\Xi \times \mathbb R^ n \to \mathbb R,\; i=0,\dots ,m_ 2,\) which are convex in \(x\in \mathbb R^ n\) for fixed \(\xi\in \Xi\), and \(f_{2i}(\cdot,x)\in \mathcal L^ 1(\Xi,\mathcal F,P;\mathbb R)\) for \(i=0,\dots ,m_ 2.\) The author considers the following stochastic programming problem: \((\mathcal P)\) Minimize \(f_{10}(x) + E\{f_{20}(\xi,x)\}\) subject to \(f_{1i}(x) \leq 0,\, x\in C,\, i=1,\dots ,m_ 1,\,\) and \(f_{2i}(\xi,x)\leq 0\; P\)-a.s., \(i=1,\dots ,m_ 2.\) For an appropriately defined Lagrangean \(L:\mathbb R^ n\times \mathbb R^{m_ 1}\times \mathcal L^ \infty(\Xi,\mathcal F,P;\mathbb R)\to \overline{\mathbb R}\) the dual problem is: \((\mathcal D)\) Maximize \(g(y,z)\) subject to \(y\geq 0, z\in \mathcal L^ \infty(\Xi,\mathcal F,P;\mathbb R),\,\) where \(g(y,z) = \inf \{L(x,y,z) : x\in C\}\). The inequality \(\inf \mathcal P \geq \sup \mathcal D \) is always true, and a duality results means to look for conditions ensuring the equality. If the existence of dual multipliers is also guaranteed, then we have a strong duality result, expressed as \(\inf \mathcal P =\max \mathcal D.\) The author proves some duality and strong duality results for problems of this kind without assuming the problem bounded or satisfying a Slater condition of strict feasibility. Also, the complete recourse requirement is replaced by a condition called direction-free feasibility, meaning that a recession cone attached to the problem is a subspace of \(\mathbb R^ n\). A calmness condition is also supposed to hold. The obtained results extend some results of \textit{R. T. Rockafellar} and \textit{R. J.-B. Wets} [Pac. J. Math. 62, 173--195 (1976; Zbl 0346.90057) and SIAM J. Control Optim 16, 574--589 (1976; Zbl 0346.90058)]. They are related also to some results on asset pricing of \textit{W. Schachermayer} [Math. Finance 4, 25--55 (1994; Zbl 0893.90017)].