Bootstrap—自适应混沌克隆网络与陀螺漂移预测  

Bootstrap—Adaptive Chaos Clone Network and Drift Forecasting of Gyro

作　　者：张伟[1,2] 胡昌华[1] 焦李成[2] 

机构地区：[1]第二炮兵工程学院,陕西西安710025 [2]西安电子科技大学智能信息处理研究所,陕西西安710071

出　　处：《电子学报》2009年第9期2035-2040,共6页Acta Electronica Sinica

基　　金：国家自然基金重点课题(No.60736026);教育部新世纪优秀人才支持计划

摘　　要：提出一种Bootstrap—自适应混沌克隆网络建立陀螺漂移预测模型.基于统计理论的Bootstrap方法,设计了自适应混沌克隆网络训练策略,以获得统计意义下陀螺漂移时间序列小样本预测模型的最优估计;运用字符串技术生成复合函数抗体,使预测模型具有在复杂函数子空间逼近强非线性函数的能力;构造了一系列自适应混沌克隆网络进化算子,使复合函数抗体在具有混沌特征的进化中自适应性更强;预测建模实验表明:Bootstrap—自适应混沌克隆网络建立的预测模型训练和测试精度稳定,对比几种典型预测算法,其建模方式更灵活,自适应性更强,适于工程应用.A Bootstrap—adaptive chaos clone network is put forward for drift forecasting modeling of gyro.Drift times series of gyro has little samples,in order to obtain statistic optimization estimate of the forecasting model, training strategy of adaptive chaos clone network is designed based on Bootstrap method.String techniques are used to produce compound function antibodies,ability of the forecasting model which approach strong nonlinear function in complex function subspace is high.A series evolve operator in adaptive chaos clone network is structured, adaptability of antibodies improve in evolution with chaos character. Forecasting modeling experiments show that forecasting model built by Bootslrap——adaptive chaos clone network is of smooth training and test precision, compared with several typical forecasting algorithm, the modeling approach is flexible, forecasting model has well adaptability and can meet the requirements of engineering application.

关 键 词：BOOTSTRAP 自适应混沌克隆网络 陀螺漂移预测 复合函数抗体 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]