基于天牛须优化Spiking神经网络的导线串扰预测  

Prediction of Cable Crosstalk Based on Spiking Neural Network Optimized by Beetle Antennae Search

作　　者：任乾华[1] 姜鸿晔 李晓露 张和茂 REN Qian-hua;JIANG Hong-ye;LI Xiao-lu;ZHANG He-mao(College of Electronics and Information,Changchun Guanghua University,Changchun Jilin 130033,China;College of Electronic information and Automation,Civil Aviation University of China,Tianjin 300000,China)

机构地区：[1]长春光华学院电气信息学院,吉林长春130033 [2]中国民航大学电子信息与自动化学院,天津300000

出　　处：《计算机仿真》2020年第11期34-38,共5页Computer Simulation

摘　　要：串扰的分析预测是实现电气线路互联系统兼容性的重要工作,其中线缆间的相互串扰是典型的电磁兼容问题,为了更加精确地预测线缆间的串扰,提出了一种基于天牛须优化Spiking神经网络的预测模型。首先将具有强大的非线性计算能力的Spiking神经网络作为基础模型,再利用搜索性能更优的天牛须算法(BAS)初始化Spiking神经网络(SNN)中突触延时和连接权值,最终预测结果表明在相同数据的基础上,上述算法迭代次数更少,且误差减小1-2个数量级,提高了模型寻优速度和稳定性,证明了该方法的有效性与实用性。The analysis and prediction of crosstalk is a key work to ensure the compatibility of Electrical Wiring Interconnection System(EWIS),and the crosstalk between cables is a typical Electromagnetic Compatibility(EMC)problem.In order to predict the crosstalk between cables more accurately,this paper proposed a prediction model based on Spiking Neural Network Optimized by Beetle Antennae Search(BAS-SNN).Firstly,the authors used Spiking neural network(SNN)with strong non-linear computing ability as the basic model,and then applied Beetle Antennae Search(BAS)with better search performance initialized the synaptic delay and connection weights in SNN.The final prediction results show that for same data,this algorithm has fewer iterations,and its error is decreased by 1-2 orders of magnitude,which improves the optimization speed and stability of model,and proves the effectiveness and practicability of the method.

关 键 词：线缆串扰 预测 天牛须搜索 神经网络 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]