一种基于改进暂态混沌神经网络的信道分配算法  

A New Algorithm Based on the Improved Transient Chaotic Neural Network for Cellular Channel Assignment

作　　者：朱晓锦[1] 陈艳春 马世伟[1] 秦霆镐[1] 

机构地区：[1]上海大学机电工程与自动化学院上海市电站自动化技术重点实验室,上海200072

出　　处：《电子与信息学报》2007年第9期2230-2234,共5页Journal of Electronics & Information Technology

基　　金：上海市教委"曙光计划"项目(04SG41);上海市重点学科建设项目(T0103);教育部留学回国人员科研启动基金资助课题

摘　　要：该文针对暂态混沌神经网络(TCNN)求解信道分配问题(CAP),分析混沌神经网络模型及其混沌性态,依据其按自反馈连接权值的减小,由混沌态通过逆分岔而收敛到稳定状态的特性,提出了一种对暂态混沌神经网络进行分段退火的策略,即依据混沌神经网络运行过程中,对应Lyaponov指数的变化特性而确定分段点,使网络能有效地利用混沌态进行全局搜索和加快收敛:在7小区的信道分配中,网络收敛速度提升了30%左右,在25小区的Kunz基准测试程序的仿真中,收敛速度也提升了近15%;仿真结果表明其有效减少了网络运算的迭代步数,提高了网络的搜索效率:通过相应理论和仿真结果的分析,对网络的搜索性能、参数的选择与设置进行了进一步的讨论。In this paper, the Transient Chaotic Neural Network（TCNN） is used to solve the Channel Assignment Problem（CAP）, and a new method named two-stage annealing method in TCNN is proposed. The neural network gradually convergences, through the transient chaos, to a stable equilibrium point according to the damping of the self-feedback connection weight, and the dividing point in the new model is chosen according to the change of the corresponding Lyapunov exponent. The two- stage annealing method can make sure the network take good advantage of the chaos to search the global minimum and enhance the convergence rate. In the 7-cell cellular network, the convergence rate is 30% higher than the TCNN model, and is also upgraded 15% in the Kunz＇s benchmark test. Simulated results show that the new model has a higher searching ability and lower computing time in searching the global minimum. The searching ability and the choosing of the parameters are also discussed based on the simulated results.

关 键 词：混沌神经网络 HOPFIELD神经网络 模拟退火 混沌噪声 信道分配问题 

分 类 号：TN916.9[电子电信—通信与信息系统]