Convergence properties of a modified Hopfield-Tank model

The neural network model of Hopfield and Tank applied to the Travelling Salesman Problem, has been analyzed in order to improve its convergence properties. A simple change of the parameter sets always allows to reach states corresponding to valid tours. Besides a more interesting modification has been presented by adding a new term to force expression. This modified model has a high value of convergence and it is able to find short tours. So, more confidence can be given to these type of models, and real applications could be performed.