On Parallelizing the Dempster-Shafer Method Using Transputer Network

Abstract Manipulating uncertain information is a necessary capability for any intelligent system. Several approaches, such as the Bayesian theory, Certainty Factors, and the Dempster-Shafer method, have been proposed to handle uncertainty. Among them the Dempster-Shafer method is the most theoretical sound and consistent with human behavior; however, it is argued on its computational complexity. This article presents a parallel reasoning algorithm based on the Dempster-Shafer method and implements it on the transputer network. We first analyze the best topologies of the transputer network with various numbers of processors; then the performance of the parallel program, such a speedup and efficiency, is measured on these best topologies.