KNOWLEDGE-BASED ENVIRONMENT FOR DEVELOPING AND TESTING EMBEDDED NEURAL NETWORKS

NUMEROUS GOVERNMENT AND INDUSTRY RESEARCH AND DEVELOPMENT PROJECTS ARE EXPLORING THE APPLICATION OF NEURAL NETWORKS WITHIN EMBEDDED SOFTWARE SYSTEMS. ALTHOUGH MODELS, IMPLEMENTATIONS, AND APPLICATIONS OF NEURAL NET TECHNOLOGY ARE MULTIPLYING, LITTLE ATTENTION HAS BEEN PAID TO TESTING, BENCHMARKING, AND COMPARING THESE COMPUTATIONAL SYSTEMS. IN FACT, FOR MANY SPECIFIC NEURAL NETWORK ARCHITECTURES LITTLE IS KNOWN ABOUT THEIR CLASSIFICATION AND INFORMATION STORAGE CAPABILITIES. THERE DOES NOT EXIST A SIMPLE, WIDELY ACCEPTED, SET OF TEST CASES WHICH CAN BE UTILIZED TO COMPARE CANDIDATE NEURAL NETWORK ARCHITECTURE, AND EVALUATE THEIR APPLICABILITY TO SPECIFIC EMBEDDED APPLICATIONS. THE PURPOSE OF THIS PROJECT IS THE DEVELOPMENT OF A NEURAL NETWORK KNOWLEDGE-BASED (N2KB). THE N2KB WILL BE DESIGNED TO PROVIDE INFORMATION ON NEURAL NETWORK MODELS, RECOMMENDATIONS ON THE APPROPRIATE NEURAL NETWORK PARADIGMS TO USE IN SPECIFIC APPLICATIONS, FUNCTIONAL PROTOTYPES OF NEURAL NETWORKS, AND NETWORK DEBUGGING AND TESTING FACILITIES.