抄録
In this paper, we propose a saturation computation method of neural networks for efficiently solving combinatorial optimization problems. In this computation method, once the neuron is in excitatory state, then its input potential is considered to be in positive saturation where the input potential can only be reduced but cannot be increased, and once the neuron is in inhibitory state, then its input potential is considered to be in negative saturation where the input potential can only be increased but cannot be reduced. The proposed method is applied to N-Queens problem. The performance is evaluated through simulations where the results show that the saturation method improves the searching capability of neural networks and shortens the computation time. Particularly, the simulation results show that the performance of the proposed method surpasses the exiting methods for N-queens problem in synchronous parallel model.
