|
Mamarelis,
Vasilis
Nonlinear System Modeling with Laguerre-Volterra Networks Artificial neural networks have been used for modeling nonlinear dynamic systems since they are universal approximators of nonlinear input-output mappings. Volterra modeling has been also used for the same purpose but has been limited to weak nonlinearities (low-order Volterra systems). Network architectures inspired by the Volterra approach have been recently proposed for modeling high-order Volterra systems [1]. Laguerre expansions of Volterra kernels have been also used for greater efficiency of the estimation process [2]. A novel network architecture is proposed that combines Laguerre filter-bank preprocessing with Volterra-type networks (having polynomial activation functions) to achieve efficient modeling of nonlinear systems. Illustrative examples of the application of these Laguerre-Volterra networks will be presented using simulated and real data from neuronal systems. References 1.Marmarelis, V.Z. and X. Zhao: Volterra models and three-layer perceptrons. IEEE Trans. on Neural Networks, 8:1421--1433, 1997 2.Marmarelis, V.Z.: Identification of nonlinear biological systems using Laguerre expansions of kernels. Annals of Biomedical Engineering, 21:573-589, 1993
|