Komlós theorem for unbounded random sets (Q5926374): Difference between revisions

Let \(F_n: T\to X\) be a sequence of multifunctions defined on a finite measure space \(T\), and whose values are closed, convex subsets of a separable Banach space \(X\). There are formulated the conditions under which there exists an integrable multifunction \(F\) and a sequence \(n_i\) such that for any subsequence \(n_{i_j}\) the mean \[ {1\over m} \sum^m_{j=1} F_{n_{i_j}} \] Mosco-converges a.e. to \(F\). The results may be regarded as a variant of the strong law of large numbers for multivalued random variables that do not have to be identically distributed nor independent, and are allowed to be unbounded. This leads to some applications to normal integrands, by considering multifunctions with epigraphs of lsc functions as values.