Hybrid evolutionary programming: The tools for CAST
With the development of new computing paradigms, such as neural networks and genetic algorithms, new tools have become available in computer-aided systems theory. These tools can be used to tackle problems that are considered “hard” in traditional systems theory, like the modeling and identification of nonlinear dynamical systems.
We present a general methodology based on neural networks and genetic algorithms that can be applied to a wide range of problems. The main emphasis is on using the approximation capabilities of neural networks to model systems based on their input-output behavior.
We first show how the inverse problem of a static system can be solved by two feedforward neural networks in a feedback loop.
We then present a general methodology for modeling nonlinear systems with known rank (i.e., state space dimension) by feedforward networks with external delay units.
We further show how genetic algorithms can be employed to find neural networks to model dynamical systems of unknown rank. Two genetic algorithms are presented for this case: one that determines the best feed-forward network with external delay, and one that searches for a network with arbitrary topology and memory cells within each neuron.
Unable to display preview. Download preview PDF.
- 1.H. Braun and J. Weisbrod. Evolving neural feedforward networks. In J. D. Schaffer, editor, Proceedings of the Third International Conference on Genetic Algorithms (ICGA), pages 25–32, San Mateo (CA), 1989. Morgan Kaufmann Publishers.Google Scholar
- 2.J. Heistermann. Learning in neural nets by genetic algorithms. In R. Eckmiller et al., editor, Proceedings of Parallel Processing in Neural Systems and Computers (ICNC), pages 165–168, Elsevier, 1990.Google Scholar
- 3.K. Hornik. Approximation capabilities of multilayer feedforward networks. Neural Networks, 4(2):251–257, 1991.Google Scholar
- 4.K. Hornik, M. Stinchcombe, and H. White. Universal approximation of an unknown mapping and its derivatives using multilayer feedforward networks. Neural Networks, 3(5):551–560, 1990.Google Scholar
- 5.V.Y. Kreinovich. Arbitrary nonlinearity is sufficient to represent all functions by neural networks: A theorem. Neural Networks, 4(3):381–383, 1991.Google Scholar
- 6.A.U. Levin. Neural Networks for Dynamical Systems. PhD thesis, Yale University, 1992.Google Scholar
- 7.K. Lindgren, A. Nilsson, M. G. Nordahl, and I. Rade. Evolving recurrent neural networks. In J. D. Schaffer, editor, Proceedings of the Third International Conference on Genetic Algorithms (ICGA), pages 55–62, San Mateo (CA), 1989. Morgan Kaufmann Publishers.Google Scholar
- 8.R. Murray-Smith, D. Neumerkel, and D. Sbarbaro-Hofer. Neural networks for modelling and control of a non-linear dynamic system. In IEEE International Conference on Intelligent Control, 1992.Google Scholar
- 9.K.S. Narendra and K. Parthasarathy. Identification and control of dynamical systems using neural networks. IEEE Transactions on Neural Networks, 1(1):4–27, 1990.Google Scholar
- 10.D. Whitley and T. Hanson. Optimizing neural networks using faster, more accurate genetic search. In J. D. Schaffer, editor, Proceedings of the Third International Conference on Genetic Algorithms (ICGA), pages 391–396, San Mateo (CA), 1989. Morgan Kaufmann Publishers.Google Scholar