A study of the transiently chaotic neural network for combinatorial optimization

We consider the dynamic behavior of the transiently chaotic neural network (TCNN). Although its dynamical behavior is of interest in investigating neural dynamics, we observed that the chaotic phase in a TCNN is not a necessary condition for the network to reach the global solution for a combinatorial optimization problem. In fact, the global solution is a result of the time varying terms in a TCNN. We, therefore, generalized the TCNN to a nonautonomous Hopfield neural network (NAHNN). Simulation shows that the NAHNN is very effective in solving the traveling salesman problem (TSP).