Abstract
A new conceptual framework for solving design optimization problems based on a neural computing paradigm is examined. It is seen that highly interconnected networks of nonlinear analog neurons are extremely effective in computing. These neurons represent an approximation to the biological neurons. The general behavior of an artificial neural network can be readily adapted to solve an optimization problem by appropriately selecting synaptic connections. Results of the computer simulation of a neural network designed to solve four optimization problems are presented to illustrate the computational power of these networks. The optimum solutions obtained using neural networks compare favorably with the optimum solutions obtained using gradient-based search techniques. The results indicate that the efficiency and power of this biological information processing scheme can be effectively utilized to solve other optimization problems with similar efficiency,