基于迭代容积卡尔曼滤波的神经网络训练算法

doi:10.11896/j.issn.1002-137X.2016.10.048

计算机科学 ›› 2016, Vol. 43 ›› Issue (10): 256-261.doi: 10.11896/j.issn.1002-137X.2016.10.048

基于迭代容积卡尔曼滤波的神经网络训练算法

袁光耀,胡振涛,张谨,赵新强,付春玲

河南大学图像处理与模式识别研究所开封475004,河南大学图像处理与模式识别研究所开封475004,河南大学图像处理与模式识别研究所开封475004,河南大学图像处理与模式识别研究所开封475004,河南大学物理与电子学院开封475004

出版日期:2018-12-01 发布日期:2018-12-01
基金资助:
本文受国家自然科学基金(61300214),中国博士后基金(2014M551999),河南省高校科技创新团队支持计划(13IRTSTHN021),河南省高校青年骨干教师资助

Novel Neural Network Training Algorithm Based on Iterated Cubature Kalman Filter

YUAN Guang-yao, HU Zhen-tao, ZHANG Jin, ZHAO Xin-qiang and FU Chun-ling

Online:2018-12-01 Published:2018-12-01

摘要/Abstract

摘要： 针对现有应用非线性滤波算法对神经网络进行训练时存在精度不足的问题,提出了一种基于迭代容积卡尔曼滤波的神经网络训练算法。首先,将前馈神经网络各个节点的连接权值和偏置作为状态向量,建立前馈神经网络的状态空间模型。其次,利用Spherical-Radial准则生成容积点,并依据Gauss-Newton迭代策略来优化量测更新过程中获取的状态估计值和状态估计误差协方差,通过容积卡尔曼滤波估计精度的改善,提升神经网络节点的连接权值和偏置的训练效果。理论分析和仿真实验结果验证了所提算法的可行性和有效性。

关键词: 前馈神经网络,状态空间模型,容积卡尔曼滤波,Gauss-Newton迭代

Abstract: In view of insufficient accuracy in the existing application of nonlinear filtering algorithm for neural network training,a novel neural network training algorithm based on iterated cubature Kalman filter was proposed.Firstly,the connection weights and bias of feedforward neural network are used as the state vector to establish the state space mo-del.Secondly, using the Spherical-Radial standard to generate cubature points,the state estimation and covariance acquired during the measurement update process are optimized based on Gauss-Newton iteration strategy.The training effect of neural network connecting weights and bias is enhanced through the improvement of the estimation precision of cubature Kalman filter.The theoretical analysis and simulation results show the feasibility and effectiveness of the algorithm.

Key words: Feedforward neural network,State-space model,Cubature kalman filter,Gauss-Newton iterate

袁光耀,胡振涛,张谨,赵新强,付春玲. 基于迭代容积卡尔曼滤波的神经网络训练算法[J]. 计算机科学, 2016, 43(10): 256-261. https://doi.org/10.11896/j.issn.1002-137X.2016.10.048

YUAN Guang-yao, HU Zhen-tao, ZHANG Jin, ZHAO Xin-qiang and FU Chun-ling. Novel Neural Network Training Algorithm Based on Iterated Cubature Kalman Filter[J]. Computer Science, 2016, 43(10): 256-261. https://doi.org/10.11896/j.issn.1002-137X.2016.10.048

参考文献

[1] Li Yuan,Deng Hui-wen.Application of speech recognition system based on algebra algorithm [J].Computer Science,2008,35(8):122-124(in Chinese) 李原,邓辉文.新型前馈网络学习算法在语音识别中的应用[J].计算机科学,2008,35(8):122-124
[2] Arsiriy E A,Antoshchuk S G.Improving the efficiency of MLP back propogation learning at the classification of quasi-stationary signals [C]∥Intelligent Data Acquisition and Advanced Computing Systems (IDAACS).2011:365-368
[3] Singhal S,Wu L.Training multilayer perceptrons with the extended Kalman algorithm[C]∥Advance in Neural Information Processing Systems,California:Morgan Kaufman Publishers Inc.1989:133-140
[4] Wu X D,Wang Y N.Extended and Unscented Kalman filtering based feedforward neural networks for time series prediction[J].Applied Mathematical Modelling,2012,36(3):1123-1131
[5] Zarei J,Shokri E.Robust fault detection via unknown input observer based on cubature Kalman filter[J].21st Iranian Confe-rence on Electrical Engineering (ICEE),2013,28(2):303-308
[6] Xi Y H,Peng H.MLP training in a self-organizing state space model using unscented Kalman particle filter[J].IEEE Transactions on Systems Engineering and Electronics,2013,24(1):141-146
[7] de Freitas J F G,Niranjan M,Gee A H,et al.Sequential monte catlo methods to train neural network models[R].England:Cambridge University,1998
[8] Arasaratnam I,Haykin S,Hurd T R.Cubature Kalman Filtering for continuous-discrete system:theory and simulations[J].IEEE Transactions on Signal Processing,2010,58(10):4977-4993
[9] Lefebvre T,Bruyninckx H,Schutter J D.Kalman filters for nonlinear systems:a comparison of performance [J].International Journal of Control,2004,77(7):639-653
[10] Mou Z K,Sui L F.Improvement of UKF Algorithm and Robustness Study[C]∥International Workshop on Intelligent Systems and Applications.2009:1-4
[11] Zhang Yi-peng,Chen Liang,Hao Huan.LM based training algorithm for quantum neural networks[J].Computer Science,2013,40(9):221-224(in Chinese) 张翼鹏,陈亮,郝欢.一种基于LM的量子神经网络训练算法[J].计算机科学,2013,40(9):221-224
[12] Zha F,Xu J N,Li J S,et al.IUKF neural network modeling for FOG temperature drift[J].Journal of Systems Engineering and Electronics,2013,24(5):828-844
[13] Mu J,Cai Y L.Iterated cubature Kalman filter and its application[C]∥IEEE International Conference on Cyber Technology in Automation,Control,and Intelligent Systems (CYBER).Kunming,China,2011:33-37
[14] Wu X D,Chang Y C,Mao J X,et al.Predicting reliability and failures of engine systems by single multiplicative neuron model with iterated nonlinear filters[J].Reliability Engineering and System Safety,2013,119:244-250
[15] Zhang Shao-you,Hao Yan-ling,Wu Xu.Complexity analysis of three deterministic sampling nonlinear filtering algorithms[J].Journal of HARBIN Institute of Technology,2013,45(12):111-115(in Chinese) 张召友,好烟令,吴旭.3种确定性采样非线性滤波算法的复杂度分析[J].哈尔滨工程大学学报,2013,5(12):111-115

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

基于迭代容积卡尔曼滤波的神经网络训练算法

Novel Neural Network Training Algorithm Based on Iterated Cubature Kalman Filter

PDF (PC)

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 0

Metrics

本文评价

推荐阅读 0