A time-delay neural network model for unconstrained nonconvex optimization (Q1627490)

This authors present some convergence and bound results on the trajectory and energy function of general neural network models. A new first-order time-delay neural network model is developed to solve unconstrained nonconvex optimization problems. It can be regarded an useful extension of the Barzilai-Borwein gradient method to neural network models. Their analysis for general neural network implies that the new neural network model is globally convergent to an equilibrium point or stationary point. Simulation results show that the new model is quite efficient. For the entire collection see [Zbl 1402.65005].