APPLICATION OF NEURAL NETWORK TO CONSTRAINED OPTIMIZATION PROBLEMS USING PENALTY FUNCTION APPROACH