Running rule-based expert systems on parallel processors

摘要

Some issues in executing rule-based systems on parallel processor systems are addressed here. A MIMD shared memory multiprocessor model is first considered for running rule-based expert systems. Rule-based expert systems are modelled by state space and AND/OR graphs. The interdependences among rules are analyzed to guide rule-base partitioning and assignment as well as parameter allocation to memory banks. Also, methods for eliminating the dependences and for avoiding indeterminacy are proposed. A novel architecture is also proposed for the parallel execution of expert systems. This architecture has a regular mesh structure. It assembles a neural network and is thus named the generalized neural network. Execution and task decomposition of expert systems on this architecture are also discussed in this paper.