A generalized model of statistical Hopfield neural network to solve TSP
Yoo-Shin Kim, Seong-Geun Cha, Ja-Myeong Kim · 2002
The travelling salesman problem (TSP) is the well-known NP-complete problem and many researchers have tried to solve it, but no method can solve this problem completely. One of the most dominant method is the statistical Hopfield (1982, 1984) neural network. We propose a generalized model which overcomes the weak points of Van Den Bout (1989) and Park's (1991) method. We improve the energy function and consider all kinds of perturbation effects and the ratio of these effects. Through a simulation for a randomly generated distribution of 10 city, we found that our proposed model shows that 90 out of 100 cases reach the optimum and a near optimum solution within a 5% error.