NEURAL NETWORK SOLUTION TO THE WEAPON-TO-TARGET ASSIGNMENT PROBLEM

PRELIMINARY RESEARCH HAS SHOWN THAT A NEURAL NETWORK CAN SOLVE THE WEAPON-TO-TARGET ASSIGNMENT (WTA) PROBLEM -A COMPUTATIONALLY INTRACTABLE PROBLEM. THE NETWORK, SIMULATED ON A LOW COST PARALLEL PROCESSOR IN A PC HOST, HAS PRODUCED THE BEST QUALITY AND FASTEST SOLUTIONS TO A REALISTIC THREAT SCENARIO PUBLISHED TO DATE. THIS APPLICATION, EMPLOYING MORE THAN 46,000 NEURONS HAVING MORE THAN 49 MILLION INTERCONNECTIONS, IS AMONG THE LARGEST NEURAL NETWORKS EVER STUDIED AND REQUIRES ONLY 8 SECONDS FOR CONVERGENCE. A STUDY IS UNDERWAY THAT ON PERFORMANCE ON MUCH LARGER SCENARIOS, EXPLORING THE USE OF GENETIC ALGORITHMS TO BETTER SOLUTION QUALITY, EXTENDING WTA TO ACCOMODATE MORE COMPLICATED THREAT SCENARIOS, AND DEMONSTRATING THE PERFORMANCE OF THE NEURAL NETWORK USING A HIGH PERFORMANCE, LOW COST, NEURAL NETWORK ACCELERATOR.