Cylindrical measures on tensor products of Banach spaces and random linear operators (Q1105283)

The author generalizes results of Skorokhod on random linear operators on Hilbert spaces to the general Banach space case. Let E, F denote real Banach spaces and suppose that the tensor product \(E\otimes F'\) (where F' is the dual of F) is provided with a cross-norm. A linear random function \(A: E\otimes F'\to {\mathcal L}^ 0(\Omega)\) (where \(\Omega\) denotes a certain probability space) is called a weak random linear operator (w- operator), if A is stochastically continuous. Similarly, a linear random function \(A: E\to {\mathcal L}^ 0(\Omega,F)\) is called a strong random linear operator (s-operator), if A is stochastically continuous. If F has the metric approximation property, it is proved that every weakly stochastically continuous, linear random function \(A: E\to {\mathcal L}^ 0(\Omega,F)\) is an s-operator. The main results deal with the problem, under which conditions random linear operators of the above type define genuine random operators, i.e. elements of \({\mathcal L}^ 0(\Omega,L(E,F))\). The results are of the following type. Let A be a w-operator and let S, T denote deterministic bounded linear operators such that T takes values in E, and S is defined on F. Then SAT is a genuine random operator, if e.g. A is of type \(p>1\) and if S and \({}^ tT\) are p-absolutely summing.