Graph coloring using neural networks

A novel neural network solution is presented for the graph coloring problem. The algorithm is an extension of the algorithm proposed by D.H. Ballard et al. (1987). By adding Potts neurons, it was possible to reduce the search space without excluding feasible solutions. The combination of the Hopfield type of neurons and the Potts neurons for minimizing the number of colors needed is used. The neural network has been programmed in C⁺⁺ and implemented on an Alliant FX-8 computer. The results obtained for a large number of graphs show the effectiveness of the algorithm, and the applicability of the proposed technique to one out of n programming problems