摘要
文章针对随机混沌系统辨识引入贝叶斯正则化方法的BPNN模型,使神经网络具有自适应性和推广能力;并交替使用贝叶斯正则化算法和混沌退火算法对网络参数进行优化,使系统具有最佳参数．以Logistic系统为例进行仿真分析,结果表明辨识模型不仅能够拟合原混沌系统,而且训练后的网络对含噪声的随机混沌系统有很好的辨识能力,精度良好．为下一步的设计控制器对混沌系统进行控制消除混沌,奠定了良好的基础．
To identify stochastic chaotic system, Bayesian regularization and the hybrid algorithms of Chaos-Annealing strategy are used to train BPNN and to optimize the model＇s parameters by the original data, and a wanted neural network model having best parameters is found. Then this trained model is applied to identify stochastic chaotic system. The simulation applied to identify chaotic systems of Logistic system show that the identified models can approach the original system and can identify stochastic chaotic system having advantage of good precision. This is the basic of controlling chaotic systems using controllers.
出处
《襄樊学院学报》
2006年第2期74-76,106,共4页
Journal of Xiangfan University
基金
湖北省教育厅科研项目(2001D69001)
关键词
随机混沌系统
贝叶斯正则化
混沌退火算法
Stochastic chaotic system
Bayesian regularization
Algorithms of Chaos-Annealing strategy
