Abstract:
Recently researchers have suggested several computational models in which, one programs by specifying large networks of simple devices. Such models are interesting because they go to the roots of concurrency - the circuit level. A problem with the models is that it is unclear how to program large systems and expensive to implement many features that are taken for granted m symbolic programming languages. This paper describes the Concurrent Inference System (CIS), and its implementation on a massively concurrent network model of computation. It shows how much of the functionality of current rule-based systems can be implemented in a straightforward manner within such models. Unlike conventional implementations of rule-based systems in which the inference engine and rule sets are clearly divided at run time, CIS compiles the rules into a large static concurrent network of very simple devices. In this network the rules and inference engine are no longer distinct. The Thinking Machines Corporation, Connection Machine - a 65,536 processor SIMD computer - is then used to run the network. On the current implementation, real time user system interaction is possible with up to 100,000 rules.