A Mothed of Improving Identification Accuracy via Deep Learning Algorithm under Condition of Deficient Labeled Data

详细信息    查看官网全文

• 英文论文题名：A Mothed of Improving Identification Accuracy via Deep Learning Algorithm under Condition of Deficient Labeled Data
• 论文作者：Jian-Guo Wang ; Zhi-Duo Cao ; Bang-Hua Yang ; Shi-Wei Ma ; Min-Rui Fei ; Hao Wang ; Yuan Yao ; Tao Chen ; Xiao-Fei Wang
• 英文论文作者：Jian-Guo Wang ; Zhi-Duo Cao ; Bang-Hua Yang ; Shi-Wei Ma ; Min-Rui Fei ; Hao Wang ; Yuan Yao ; Tao Chen ; Xiao-Fei Wang ; School of Mechatronical Engineering and Automation ; Shanghai University ; Shanghai Key Lab of Power Station Automation Technology ; Department of Chemical Engineering ; "Tsing-Hua University" ; Department of Chemical and Process Engineering ; University of Surrey ; Shanghai Zhenhua Heavy Industries Company Limited
• 年：2017
• 作者机构：School of Mechatronical Engineering and Automation,Shanghai University,Shanghai Key Lab of Power Station Automation Technology;Department of Chemical Engineering,"Tsing-Hua University";Department of Chemical and Process Engineering,University of Surrey;Shanghai Zhenhua Heavy Industries Company Limited;
• 英文论文关键词：System Identification ; Deep Learning ; Controlled Auto Regressive ; Finite Impulse Response
• 会议召开时间：2017-07-26
• 会议录名称：第36届中国控制会议论文集（B）
• 英文会议录名称：Proceedings of the 36th Chinese Control Conference（B）
• 语种：英文
• 分类号：TP18
• 学会代码：KZLL
• 会议名称：第36届中国控制会议
• 会议地点：中国辽宁大连
• 主办单位：中国自动化学会控制理论专业委员会
• 学会名称：中国自动化学会控制理论专业委员会
• 页数：6
• 文件大小：261k
• 原文格式：D
• 会议级别：全国

摘要

In industrial process, some important variables such as quality index, efficiency index and concentration of product components are difficult or even impossible to be measured directly due to the limitation of technology. This phenomenon leads to few labeled data and plenty of unlabeled data. Traditional identification method for controlled auto regressive(CAR) model usually cannot deal with unlabeled training data. As a result, these traditional identification methods may receive poor identification precision or even cannot work entirely. To solve the problems above, this paper proposes a new identification method based on deep learning(DL). Firstly, the CAR model is transformed into finite impulse response(FIR) model solve the problem of lack of autoregressive part; Secondly, autoencoder of deep learning make full use of unlabeled data to pretrain the model; Thirdly, small amount of label data is used for fine-tuning. As a semi-supervised learning method, deep learning can be able to extract more information from unlabeled data than traditional supervised learning method. The results show that the proposed method can acquire higher identification accuracy than BP neural network.
In industrial process, some important variables such as quality index, efficiency index and concentration of product components are difficult or even impossible to be measured directly due to the limitation of technology. This phenomenon leads to few labeled data and plenty of unlabeled data. Traditional identification method for controlled auto regressive(CAR) model usually cannot deal with unlabeled training data. As a result, these traditional identification methods may receive poor identification precision or even cannot work entirely. To solve the problems above, this paper proposes a new identification method based on deep learning(DL). Firstly, the CAR model is transformed into finite impulse response(FIR) model solve the problem of lack of autoregressive part; Secondly, autoencoder of deep learning make full use of unlabeled data to pretrain the model; Thirdly, small amount of label data is used for fine-tuning. As a semi-supervised learning method, deep learning can be able to extract more information from unlabeled data than traditional supervised learning method. The results show that the proposed method can acquire higher identification accuracy than BP neural network.

引文

[1]Narendra K S,Parthasarathy K.Identification and control of dynamical systems using neural networks.IEEE trans on neural networks,1990.vol(1).4-27.
    [2]Zadeh L A.From Circuit Theory to System Theory[J].Proc,IRE,1962,50(5):856-865.
    [3]Chongzhi Fang,Deyun Xiao.Process identification[M].Beijing:Tsinghua University press,2001
    [4]Lideng Pan,Yangdong Pan.System identification and modeling[M].Beijing:Chemical Industry Press,2004
    [5]Sorenson H W.Least squares Estimation:from Gauss to Kalman[J].IEEE Spectrum,1970,7:63-68.
    [6]Narendra,K.S.,Parthasarathy,K.Identification and control of dynamical systems using neural networks[J].IEEE Transactions on Neural Networks,1990,1(1):4-27.
    [7]Yu W,Li X.O.Fuzzy identification using fuzzy neural networks with stable learning algorithms[J].IEEE Transactions on Fuzzy Systems,2004,12(3):411 420.
    [8]Dayal B.S,Macgregor J.F.Identification of finite impulse response models:methods and roubustness issues[J].Ind.Eng.Chem.Res.,1996,35:4078 4090.
    [9]G.E.Hinton,R.R.Salakhutdinov,Reducing the dimensionality of data with neural networks.Science,313(2006)504–507
    [10]G.E.Hinton,S.Osindero,Y.-W.The.A fast learning algorithm for deep belief nets.Neural Comput.18(2006)1527–1554
    [11]Qinghua Hu,Rujia Zhang,Yucan Zhou.Transfer learning for short-term wind speed prediction with deep neural networks.Renewable Energy,85(2016)83-95.
    [12]Pierre Baldi.Autoencoder,unsupervised learning,and Deep Architectures.Journal of machine learning research 50(2012)27-37
    [13]Hoo-Chang Shin,Matthew R.Orton,et.al.Stacked Autoencoders for Unsupervised Feature Learning and Multiple Organ Detection in a Pilot Study Using 4D Patient Data.IEEE transactions on pattern analysis and machine intrlligence,vol.35,NO.8,2013
    [14]Chaoqun Hong,Jun Yu,et.al.Multimodal deep autoencoder for human pose recovery.IEEE transactions on image processing,vol.24,NO.12,2015
    [15]Deep Networks Overview.http://uflfl.stanford.edu/wiki/index.php/Deep_Networks:Overview.
    [16]Stanford deep learning tutorial.http://deeplearning.stanford.edu/wiki/index.php/U FLDL_Tutorial.
    [17]Y.Bengio,P.Lamblin,D.Popovici,H.Larochelle,Greedy layer-wise training of deep networks.Conference on Advances in Neural Information Processing Systems,2007(19):153-160.
    [18]Shaohua Tan,Jianbin Hao,Joos Vandewalle.Efficient identification of RBF neural net models for nonlinear discrete-time multivariable dynamical systems.Neurocomputing,1995,9(1):11-26.
    [19]Li Jia,Min-Sen Chiu,Shuzhi Sam Ge.A noniterative neuro-fuzzy based identification method for Hammerstein processes.Journal of Process Control,2005,15:749-761.