Novel Cascade Spline Architectures for the Identification of Nonlinear Systems

Abstract

In this paper two novel nonlinear cascade adaptive architectures, here called sandwich models, suitable for the identification of general nonlinear systems are presented. The proposed architectures rely on the combination of structural blocks, each one implementing a linear filter or a memoryless nonlinear function. All the nonlinear functions involved in the adaptation process are based on spline functions and can be easily modified during learning using gradient-based techniques.\ud In particular, a simple form of the on-line adaptation algorithms for the two architectures is derived. In addition, we analytically obtain a bound for the selection of the learning rates involved in the learning algorithms, in order to guarantee a convergence towards a minimum of the cost function. Finally, some experimental results demonstrate the effectiveness of the proposed method

System Identification of Thermal Process using Elman Neural Networks with No Prior Knowledge of System Dynamics

“…For this, Pan W. et al cast this identification problem as a sparse linear regression problem and took a Bayesian viewpoint to solve it. The works in [8,12] have made use of spline techniques for the design of adaptive filters and used for linear and nonlinear system identification. Parametric and nonparametric methods using neural feedforward networks have been adopted for the system identification of an experimental turbojet engine in [13].…”

A study about Chebyshev nonlinear filters

“…In the third experiments, we compare the performance of the CN, Volterra, LN, and EMFN filters on two data sets available in the literature [52] and currently used for benchmarking in nonlinear system identification [55]. The first data set is recorded from coupled electric drives, i.e., two electric motors driving a pulley with flexible belt.…”

Highly efficient nonlinear regression for big data with lexicographical splitting

“…In order to show that our algorithm also works good in 4 Performance of LMW, CTW, FNF, and VF algorithms for a nonstationary data set, generated by (19), with abrupt change of parameters at beginning of the second half the higher-dimensional regressor spaces, we tested our algorithm using the well-known data sets California housing and Abalone [11], which have high-dimensional regressor space. We compared the performance of our algorithm with other well-known nonlinear regression algorithms Bezier spline adaptive filter (B-SAF) [15], Catmull-Rom spline adaptive filter (CR-SAF) [15], FNF [13], CTW [2], and VF [12]. In both "California housing" and "Abalone" experiment, we set the learning rates of the adaptive filters to 0.01.…”