An Improved Algorithm for TSP Problem Solving with Hopfield Neural Networks

Hopfield and Tank have shown that neural networks can be used to solve certain computationally hard problems, in particular they studied the Traveling Salesman Problem (TSP). In this paper,on the base of the analysis of tradiontial methord, introduced an improved algorithm for TSP Problem Solving with Hopfield Neural Networks. We found the accuracy of the results depend on the initial parameters to a large extent, discussed how to set initial parameters properly; analysed the internal relationship between the terms in energy function, and improved the energy function. Used a fixed starting point to eliminate the equivalent solution problem,and the number of neurons is reduced from the N2 to (N-1)2. The improved algorithm reduced the unnecessary equivalent solution in calculate process, enhanced the computational efficiency. Experiment results showed that the algorithm improved the speed and the convergence.