La. Feldkamp et Gv. Puskorius, A SIGNAL-PROCESSING FRAMEWORK BASED ON DYNAMIC NEURAL NETWORKS WITH APPLICATION TO PROBLEMS IN ADAPTATION, FILTERING, AND CLASSIFICATION, Proceedings of the IEEE, 86(11), 1998, pp. 2259-2277
We present in this paper a coherent neural network-based framework for
solving a variety of difficult signal processing problems. The framew
ork relies on the assertion that time-lagged recurrent networks posses
s the necessary representational capabilities to act as universal appr
oximators of nonlinear dynamical systems. This property applies to mod
eling problems posed as system identification, time-series prediction,
nonlinear filtering adaptive filtering, and temporal pattern classifi
cation. We address the development of models of nonlinear dynamical sy
stems, 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 sho
w, that a single time-lagged recurrent neural network can be trained n
ot only to produce excellent one-time-step predictions for two differe
nt 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] t
o lead to unstable trained models. We illustrate how multistream train
ing may be used to enhance the stability of such models. The remaining
two problems are drawn from real-world automotive applications. The f
irst of these involves input-output modeling of the dynamic behavior o
f 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.