A derivative of the Hopfield-Tank neural network model that reliably solves the traveling salesman problem

Summary form only given. The authors investigated the suitability of the Hopfield-Tank neural network model for solving several constraint satisfaction problems. They started by attempting to reproduce the results of Hopfield and Tank, who claim to have built a neural network that finds good solutions to the traveling salesman problem (TSP). They found this very difficult. They describe the Hopfield-Tank neural network model (HTD) and how it is used to solve the TSP. They describe their derivation of the HTD. The authors give a number of parameters in the HTD whose values are critical to the successful operation of the network, and discuss the heuristics used to find the proper values for these parameters. The authors then present test results for the HTD that indicate that it reliably yields good solutions for the TSP.<>