Neural Learning of Chaotic Dynamics The Error Propagation Algorithm.pdf

Published in: IEEE World Conference on Computational Intelligence, p. 2483, 1998. Copyright IEEE. Neural Learning of Chaotic Dynamics: The Error Propagation Algorithm Rembrandt Bakker, Jaap C. Schouten, Cor M. van den Bleek C. Lee Giles Delft University of Technology, NEC Research Institute Depa rtment of Chemical Process Technology 4 Indep endence Way Julianalaan 136, 2628 BL Delft, The Netherlands Princeton, NJ 08540 r.bakker@stm.tudelft.nl giles@research.nj . Abstract equations that will produce a time-series with identical chaotic characteristics, having the same chaotic attractor. An algorithm is introduced that trains a neural network to The model could be based on first principles if the system is identify chaotic dy namics f rom a single measured time- well understood, but here we assume knowledge of j ust the series. The algorithm has f our sp ecial features: time-series and use a neural network based, black-box, model. In concise neural network j argon we formulate our 1. The state of t he system is extracted f rom the time- goal: train a network to learn the chaotic attractor. A number series using delays, f ollowed by weighted Principa l