On a Carathéodory-type selection theorem (Q1122678): Difference between revisions

Let (T,\({\mathcal T},\mu)\) be a complete finite measure space, Y be a separable Banach space and Z be a complete separable metric space. Let \(\Phi: T\times Z\to 2^ Y\) be a convex valued correspondence and \(U:=\{(t,x)\in T\times Z: \Phi (t,x)\neq \emptyset \}.\) Under measurability assumptions on \(\Phi\) and U the existence of a Carathéodory selection for \(\Phi\) is shown, i.e., there is a function f: \(U\to Y\) such that \(f(t,x)\in \Phi (t,x)\) for all (t,x)\(\in U\) and f(\(\cdot,x)\) is measurable for all x and f(t,\(\cdot)\) is continuous for each t. This result is used by the authors in J. Approximation Theory 56, 256-266 (1989)] to extend the theory of Nash equilibria to the setting of an arbitrary measure space of agents and an infinite dimensional strategy space. For further literature on Carathéodory selection theorems see authors' paper in J. Math. Anal. Appl. 122, 393-407 (1987; Zbl 0629.28007).