A mathematical theory of randomized computation. II (Q1114406)

In previous work [ibid. 64, No.4, 115-118 (1988; Zbl 0657.68061)] a randomized program was considered as a linear operator mapping an input probability measure to the output subprobability measure. In the present article the identity between the order topology on a randomized domain and the Scott topology is shown. Some respective properties of Banach lattices are deduced. Further, the author states that the set of regular operators T: \(V\to W\), where V, W are Banach lattices, coincides with the set of order continuous operators. The product and exponent of randomized domains are defined. The first is defined coordinatewise and is a KB- space if both factors are KB-spaces. One of the main results states that a randomized domain is the positive unit hemisphere of a KB-space and mappings between randomized domains are positive operators. Modification of Daniell's integral for Banach lattices is considered, too. If \({\mathcal R}\) is any randomized domain with positive order continuous operator, then the fixed point theorem states that T: \({\mathcal R}\to {\mathcal R}\) has a fixed point, and there exists such Fix: [\({\mathcal R}\to {\mathcal R}]\to {\mathcal R}\) that Fix(T) is the least fixed point of T.