一种径向基混沌神经网络分段退火策略及应用  

PARTITIONED SIMULATED ANNEALING STRATEGY ON CHAOTIC NEURAL NETWORK WITH RADIAL BASIS FUNCTION AND ITS APPLICATION

作　　者：许楠[1] 宁常鑫[1] 徐耀群[2] 

机构地区：[1]黑龙江八一农垦大学信息技术学院,黑龙江大庆163319 [2]哈尔滨商业大学系统工程研究所,黑龙江哈尔滨150028

出　　处：《计算机应用与软件》2014年第6期158-161,194,共5页Computer Applications and Software

基　　金：2013年黑龙江省教育厅科学技术研究项目(面上)(12531456)

摘　　要：在Chen L等人提出的暂态混沌神经网络模型基础上,采用径向基函数构成的非线性自反馈连接项,分析该网络的动力学特性,模拟退火策略采用分段式结构,合理控制网络的混沌搜索过程及收敛速度。将其应用于组合优化问题中,通过仿真实验,说明若合适调整一级、二级模拟退火参数,则该新型网络能够较好克服陷入局部极小点并较快收敛到最优解,从而验证该方案的有效性。Based on the transiently chaotic neural network model proposed by Chen Let al, we use non-linear self-feedback connection item composed with radial basis function to analyse the dynamic characters of the network, simulate annealing strategy by adopting its partitioned structure, and reasonably control the chaotic search process and convergence speed of the network. We apply this new network to combinatorial optimisation problem. It is illustrated by the simulation experiment that the new network can well escape from falling into local minimum and converges to optimised solution quicker if the first and second-order annealing parameters are suitably adjusted. The effectiveness of this scheme is therefore verified.

关 键 词：径向基函数 非线性自反馈 组合优化 模拟退火 最优解 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]