摘要
论文首先给出了用均场退火算法(AFA)求解四色问题的神经网络结构和能量函数,为了避免网络陷入局部极小的缺陷,在均场的基础上增加了“爬山”项,使网络最终能收敛到一个全局最优或近似全局最优解。仿真结果表明,该方法较文献[4]中的离散的二元Hopfield-型神经网络和文献[7]中的瞬态混沌神经网络在收敛速度方面有明显的提高,效果较好。
In this paper,a mean field annealing-based algorithm(AFA)is given for solving four-coloring map problem.First,it gives its neural network and computational energy.To avoid that the network will be stuck into local minimum,it adds'hill-climbing'term to the mean field,so that it can be expected to force the network to converge to the globally optimal or near-optimal solutions.Numerical simulations of four-coloring map problem show that AFA has an apparent advantage in the convergence speed by comparing AFA with the discrete binary Hopfield neural networks in conference(4)and the neural networks with transient chaos in.The experimental results show that it is an effective method.
出处
《计算机工程与应用》
CSCD
北大核心
2005年第3期67-69,共3页
Computer Engineering and Applications
基金
湖南省教育厅科研基金资助项目(编号:02C133)
关键词
四色问题
均场退火算法
神经网络
four-coloring map problem,mean field annealing approach,neural network
