AUTOMATIC TRANSLATION OF EXPERT SYSTEMS TO A NEURAL NETWORK REPRESENTATION

THE PRIMARY OBJECTIVE OF THIS PHASE I EFFORT IS TO DESIGN AND IMPLEMENT A SOFTWARE PROTOTYPE FOR AUTOMATICALLY TRANSLATION EXPERT SYSTEMS INTO A NEURAL NETWORK REPRESENTATION TO ALLOW EXPERT SYSTEMS TO DERIVE BENEFITS OF NEURAL NETWORKS, INCLUDING ADAPTABILITY, IMPROVED REAL-TIME PERFORMANCE, FAULT TOLERANCE AND THE ABILITY TO GENERALIZE. THE NEURAL NETWORK CAN BE READILY TRANSITIONED TO A PARALLEL PROCESSING MACHINE AND MAKE FULL USE OF ALL ITS PROCESSORS. PHASE I WILL CULMINATE IN AN ARCHITECTURE DESIGN FOR A COMPLETE AUTOMATIC TRANSLATION PRODUCT WHICH WILL BE APPLICABLE TO AT LEAST ONE EXPERT SYSTEM BUILDING TOOL. NEURAL NETWORK ARCHITECTURES WITH HUNDREDS OF THOUSANDS OF SIMPLE PROCESSING ELEMENTS WILL SOON BE AVAILABLE. AN EXPERT SYSTEM TO NEURAL NETWORK TRANSLATION FACILITY WILL ALLOW EXPERT SYSTEMS TO TAKE FULL ADVANTAGE OF THIS POWER FOR THE ASAT PROGRAM.