遗传优化神经网络的激光散斑数据建模与分析  

Modeling and analysis of laser speckle data based on genetic optimization neural network

作　　者：庞亮 孟雪井[2] PANG Liang;MENG Xuejing(Wuhan Institute of Design and Sciences,Wuhan 430025,China;School of Mathematics and Statistics,Zhongnan University of Economics and Law,Wuhan 430025,China)

机构地区：[1]武汉设计工程学院,武汉430025 [2]中南财经政法大学数学与统计学院,武汉430025

出　　处：《激光杂志》2018年第9期148-151,共4页Laser Journal

基　　金：2016年湖北省教育厅科学研究计划指导性项目(No.B2016427)

摘　　要：激光散斑时间序列存在高度非线性和不确定性等特点,导致传统模型存在预测精度不高、耗时长等问题。为了提高预测精度,将遗传算法与神经网络相融合,提出基于遗传优化神经网络的激光散斑时间序列预测模型。将时间序列不同的预测值作为网络初始输入值,对原始输入值使用正交三角函数进行扩展,在满足模型拟合充要条件的基础上,计算神经网络的权重,构建激光散斑时间序列预测模型。对于神经网络连接权值和阈值选取上的随机性缺陷,采用遗传算法对其进行优化,以得到的最优适应度值训练神经网络预测模型。实验结果表明,优化后的权值和阈值能使神经网络具有更好的预测精度,缩短了预测时间。The time series of laser speckle has the characteristics of high nonlinearity and uncertainty,which leads to the problems of low prediction precision and long time consuming in traditional model. In order to improve the prediction accuracy,the genetic algorithm is fused with the neural network,and a time series prediction model of laser speckle based on neural network is proposed. Different prediction values of time series are taken as initial input values of the network. The original input values are expanded by orthogonal trigonometric functions. On the basis of satisfying the necessary and sufficient conditions of model fitting,the weights of neural network are calculated,and a time series prediction model of laser speckle is constructed. Considering the randomness of neural network connection weights and threshold selection,we use genetic algorithm to optimize it,and get the best fitness value to train the neural network prediction model. The experimental results show that the optimized weights and thresholds can make the neural network better prediction precision and shorter prediction time.

关 键 词：神经网络 激光散斑 时间序列 预测模型 

分 类 号：TN919.81[电子电信—通信与信息系统]