Neural Network based Modeling, Characterization and Identification of Chaotic Systems in Nature
Year: 2015
Type: dissertation
Abstract: Modeling of chaotic systems, based on the output time series, is quite\npromising since the output often represents the characteristic behaviour of\nthe total system. It has been an interesting topic for researchers over the past\nfew years. So far, some methods are developed for the identification of\nchaotic systems. Because of the intense complexity of chaotic systems, the\nperformance of existing algorithms is not always satisfactory. Application of\nchaotic system theory to socially relevant problems like environmental\nstudies is the need of the hour Neural networks have the required self-learning capability to tune the\nnetwork parameters (i.e. weights) for identifying highly non-linear and\nchaotic systems. In the present work, effectiveness of modeling a chaotic\nsystem using dynamic neural networks has been demonstrated. From the rich\nliterature available for non-linear modeling with neural networks, the\nRecurrent Neural Network (RNN) structure is selected. The Extended\nKalman Filter (EKF) algorithm is used to train the RNN. Further, the\nExpectation Maximization algorithm is used to effectively arrive at the initial\nstates and the state covariance. Particle filter algorithm with its two important\nvariants namely Sampling Importance Resampling (SIR) and Rao\nBlackwellised algorithms are also used for training the given RNN. Four\nstandard chaotic systems, Lorenz, Rossler, Chua and Chen, are modelled\nwith the three algorithms. The best algorithm is found to be EKF-EM based\non the least mean square error criterion. Validation of RNN model with EKFEM\nalgorithm is done in time domain by Estimation of embedding\ndimension, Phase plots, Lyapunov Exponents, Kaplan -Yorke dimension and\nBifurcation diagrams. Analysis of the chaotic systems is also performed in\nthe transform domain using Fourier, Wavelet and Mapped Real Transforms.\nviii\nNatural chaotic systems are analyzed based on the selected model structure\nand training algorithm, taken for analysis. Sunspot, Venice Lagoon and\nNorth Atlantic oscillations are the three of the natural chaotic systems\nmodelled with the selected RNN model structure and EKF-EM algorithm.
Authors: R Archana, R. Gopikakumari, A. Unnikrishnan
Institution: Cochin University of Science and Technology
Language: English
Subfield: Statistical and Nonlinear Physics
Field: Physics and Astronomy
Domain: Physical Sciences
Keywords: chaotic systems, recurrent neural networks, extended Kalman filter, expectation-maximization algorithm, particle filter, sampling importance resampling, Rao-Blackwellized particle filter, Lorenz system, Rössler system, Chua's circuit, Chen system, embedding dimension, Lyapunov exponents, Kaplan-Yorke dimension, bifurcation diagrams, solar cycle, Venice Lagoon, North Atlantic Oscillation, wavelet analysis, Fourier transform
Open Access status: green
OpenAlex ID: W2741030913
