神经网络处理平面图着四色问题的研究

A STUDY OF NEURAL-BASED FOUR-COLORING MAP PROBLEM NETWORK

以Hopfield神经网络为基础,提出一种网络模型,可以正确处理平面图着四色问题.它具有结构简单,连接规则,容错性强等优点,并可用电子电路实现.在各个神经元外加激励稍有不同或网络连接权值有偏差的情况下可以得到正确输出.用两种途径进行仿真计算,其结果与理论分析相吻合.

A novel network based on Hopfield neural network to handle the four-coloring map problem is presented in this paper. The structure of this network is simple, regular, and fault-tolerant. The network can be implemented with electronic cirouits. The correct outputs can be obtained when the external excittaions to neurons or the inter-conneeting weights are slightly different. Two different approaches are used to simulate the performances of the network. The results are in accord with theoretical analysis.