Abstract

Conventional electronic expert systems are usually based on serial search from the knowledge base for the appropriate facts and rules to conclude a query. The application of those expert systems to solve symbolic logic problems has been limited due to the inefficiency of massive searches. We describe a new method to encode the knowledge base of facts and rules so that Al problems are more suitable to optical solutions. Facts with two arguments are encoded as binary matrices. Each matrix element defines whether the fact relating the arguments is true (1) or false (0). All rules are manipulated in matrix–algebraic formulation. Conclusions to a query are derived from the knowledge base by performing on a query vector a sequence of vector–matrix multiplication operations.

© 1987 Optical Society of America