A neural network designed to solve the N-Queens Problem

Abstract

In this paper we discuss the Hopfield neural network designed to solve the N-Queens Problem (NQP). Our network exhibits good performance in escaping from local minima of energy surface of the problem. Only in approximately 1% of trials it settles in a false stable state (local minimum of energy). Extenive simulations indicate that the network is efficient and less sensitive to changes of its initial energy (potentials of neurons). Two strategies employed to achieve the solution and results of computer simulation are presented. Some theoretical remarks about convergence of the network are added.