Neural learning of chaotic dynamics: the error propagation algorithm

An algorithm is introduced that trains a neural network to identify chaotic dynamics from a single measured time-series. The algorithm has four special features: the state of the system is extracted from the time-series using delays, followed by weighted principal component analysis data reduction; the prediction model consists of both a linear model and a multi-layer-perceptron; the effective prediction horizon during training is user-adjustable, due to 'error propagation': prediction errors are partially propagated to the next time step; and to decide when to stop training, a criterion is monitored during training to select the model that has a chaotic attractor most similar to the real system's attractor. The algorithm is applied to laser data from the Santa Fe time-series competition (set A). The resulting model is not only useful for short-term predictions but it also generates time-series with similar chaotic characteristics as the measured data.

[1]  C. M. van den Bleek,et al.  Neural networks for prediction and control of chaotic fluidized bed hydrodynamics: a first step , 1997 .

[2]  J.C. Principe,et al.  Reconstructed dynamics and chaotic signal modeling , 1994, Proceedings of 1994 IEEE International Conference on Neural Networks (ICNN'94).

[3]  Jose C. Principe,et al.  Prediction of Chaotic Time Series with Neural Networks , 1992 .

[4]  Takens,et al.  Neural network model to control an experimental chaotic pendulum. , 1996, Physical review. E, Statistical physics, plasmas, fluids, and related interdisciplinary topics.

[5]  Andreas S. Weigend,et al.  Time Series Prediction: Forecasting the Future and Understanding the Past , 1994 .

[6]  L. A. Aguirre,et al.  Validating Identified Nonlinear Models with Chaotic Dynamics , 1994 .

[7]  Hermann Haken,et al.  At least one Lyapunov exponent vanishes if the trajectory of an attractor does not contain a fixed point , 1983 .

[8]  F. Takens.  Detecting strange attractors in turbulence , 1981 .

[9]  Sun-Yuan Kung,et al.  A delay damage model selection algorithm for NARX neural networks , 1997, IEEE Trans. Signal Process..

[10]  Ioannis G. Kevrekidis,et al.  DISCRETE- vs. CONTINUOUS-TIME NONLINEAR SIGNAL PROCESSING OF Cu ELECTRODISSOLUTION DATA , 1992 .

[11]  G. P. King,et al.  Extracting qualitative dynamics from experimental data , 1986 .

[12]  Firdaus E. Udwadia,et al.  An efficient QR based method for the computation of Lyapunov exponents , 1997 .

[13]  J. E. Jackson.  A User's Guide to Principal Components , 1991 .

[14]  C. Diks,et al.  Detecting differences between delay vector distributions. , 1996, Physical review. E, Statistical physics, plasmas, fluids, and related interdisciplinary topics.

[15]  Gustavo Deco,et al.  Neural Learning of Chaotic System Behavior (Special Section on Nonlinear Theory and Its Applications) , 1994 .

[16]  J. Yorke,et al.  Chaotic behavior of multidimensional difference equations , 1979 .