Lyapunov指数的算法改进与加权预测  

Improved calculation of Lyapunov exponent and weighted prediction

作　　者：梁勇[1] 孟桥[2] 陆佶人[2] 

机构地区：[1]南京财经大学工业工程系,南京210003 [2]东南大学无线电工程系,南京210096

出　　处：《声学技术》2006年第5期463-467,共5页Technical Acoustics

摘　　要：对在计算最大Lyapunov指数方面获得广泛应用的Rosenstein方法作出了改进。Rosenstein方法在计算时,只选取一个最近邻点,容易受到噪声的影响。Kantz提出用固定邻域内多点平均的方法来减少噪声的影响,但固定邻域在实际计算时存在困难。针对这个问题,提出了实用的可变邻域的取法。并且提出了将最大Lyapunov指数用于对预测模型进行加权,仿真表明预测效果得到了明显改善。Rosenstein algorithm, widely used in calculating the maximal Lyapunov exponent, is improved. The Rosenstein algorithm selects only one neighboring point, therefore is susceptible to noise. Kantz selects several points of a fixed neighborhood to reduce the influence, resulting in some problems in practical use. A weighted method to select a variable neighborhood is given in this paper. The maximal Lyapunov exponent is used to be weights in the prediction model. Simulation indicates that the weighted prediction is more precise than the normal prediction model.

关 键 词：混沌 李雅普诺夫指数计算 算法改进 加权预测 

分 类 号：N93[自然科学总论] TN911.7[电子电信—通信与信息系统]