A signal processing framework based on dynamic neural networks with application to problems in adaptation, filtering, and classification

被引:84
作者
Feldkamp, LA [1 ]
Puskorius, GV [1 ]
机构
[1] Ford Motor Co, Ford Res Lab, Dearborn, MI 48121 USA
关键词
automotive diagnostics; backpropagation through time; Kalman filtering; multistream training; recurrent multilayer perceptrons; recurrent networks; stability; system identification; time series prediction;
D O I
10.1109/5.726790
中图分类号
TM [电工技术]; TN [电子技术、通信技术];
学科分类号
0808 ; 0809 ;
摘要
We present in this paper a coherent neural network-based framework for solving a variety of difficult signal processing problems. The framework relies on the assertion that time-lagged recurrent networks possess the necessary representational capabilities to act as universal approximators of nonlinear dynamical systems. This property applies to modeling problems posed as system identification, time-series prediction, nonlinear filtering adaptive filtering, and temporal pattern classification. We address the development of models of nonlinear dynamical systems, in the form of time-lagged recurrent neural networks, which can be used without further training (i.e., as fixed-weight networks). We employ a weight update procedure based on the extended Kalman filter (EKF); as a solution to the recency effect, which is the tendency for a network to forget earlier learning as it processes new examples, we have developed a technique called multistream training. We demonstrate our training framework by applying it to four problems. First, we show, that a single time-lagged recurrent neural network can be trained not only to produce excellent one-time-step predictions for two different time;series, but also to be robust to severe errors in the provided input sequence. The second problem involves the modeling of a complex system containing significant process noise, which was shown in [1] to lead to unstable trained models. We illustrate how multistream training may be used to enhance the stability of such models. The remaining two problems are drawn from real-world automotive applications. The first of these involves input-output modeling of the dynamic behavior of a catalyst-sensor system which is exposed to art operating engine's exhaust stream. Finally we consider real-time and continuous detection of engine misfire, which is cast as a dynamic pattern classification problem.
引用
收藏
页码:2259 / 2277
页数:19
相关论文
共 28 条
[1]  
Anderson B. D. O., 1979, OPTIMAL FILTERING
[2]   LEARNING LONG-TERM DEPENDENCIES WITH GRADIENT DESCENT IS DIFFICULT [J].
BENGIO, Y ;
SIMARD, P ;
FRASCONI, P .
IEEE TRANSACTIONS ON NEURAL NETWORKS, 1994, 5 (02) :157-166
[3]  
Bishop C. M., 1995, NEURAL NETWORKS PATT
[4]  
COTTER NE, 1991, P INT JOINT C NEUR N, V1, P799
[5]  
COTTER NE, 1990, P INT JOINT C NEUR N, V3, P553
[6]  
Feldkamp LA, 1996, IEEE IJCNN, P155, DOI 10.1109/ICNN.1996.548883
[7]  
FELDKAMP LA, 1994, IEEE DECIS CONTR P, P2754, DOI 10.1109/CDC.1994.411384
[8]   Adaptive behavior from fixed weight networks [J].
Feldkamp, LA ;
Puskorius, GV ;
Moore, PC .
INFORMATION SCIENCES, 1997, 98 (1-4) :217-235
[9]  
Haykin, 1996, ADAPTIVE FILTER THEO, V3rd
[10]  
Haykin S., 1994, NEURAL NETWORKS COMP