采用动态阈值和随机梯度的带噪声混沌系统的识别方法
Identification of chaotic system with noise via dynamic threshold and stochastic gradient
-
摘要: 为了提高对带噪声混沌系统识别的准确性,结合小波神经网络,提出了基于动态阈值和随机梯度的识别方法.该方法将动态变化的阈值作用于小波系数,并与神经网络训练过程紧密结合,依据误差函数,采用随机梯度下降方法反向动态修改阈值,使系统误差更接近于理想输出.实验结果表明,该方法能合理去除混沌信号中的噪声,识别结果更为准确.Abstract: In order to improve identification exactness of chaotic system with noise,a new identification method based on wavelet neural network was presented,which was integrated with dynamic threshold and stochastic gradient.The method made dynamic threshold to affect wavelet coefficients,the train process was closely integrated with neural network and the threshold was dynamically and inversely modified based on error function and stochastic gradient,so the error was increasingly shrinking and approaching to the ideal output.The result showed that the method could reduce the noise of chaotic system and finally acquire better exactness.
-
Key words:
- chaotic system /
- dynamic threshold /
- stochastic gradient /
- wavelet neural network
-
-
[1]
陈付华.小波在图像分析中的若干关键技术研究[D].南京:南京理工大学,2002:1-6.
-
[2]
李逊,谢红胜.基于遗传算法的小波神经网络[J].计算机与数字工程,2007,35(8):5.
-
[3]
张加云,张德江,李新胜.遗传小波神经网络在钢铁企业能耗预测中的应用[J].冶金自动化,2009,33(S1):849.
-
[4]
赵金宪,金鸿章.基于小波包和神经网络的瓦斯传感器故障诊断[J].传感器与微系统,2010,29(5):80.
-
[5]
赵劲松,李元,邱彤.一种基于小波变换与神经网络的传感器故障诊断方法[J].清华大学学报:自然科学版,2013,53(2):205.
-
[6]
Krishna B,Rao Y R,Nayak P C.Time series modeling of river flow using wavelet neural networks[J].Journal of Water Resource and Protection,2012,3(1):50.
-
[7]
董长虹.Matlab小波分析工具箱原理与应用[M].北京:国防工业出版社,2004:108-118.
-
[8]
Minu K K,Lineesh M C,Jessy J C.Wavelet neural networks for nonlinear time series analysis[J].Applied Mathematical Sciences,2011,4(2):2485.
-
[1]
-
点击查看大图
计量
- PDF下载量: 22
- 文章访问数: 857
- 引证文献数: 0
下载:
