Title :
The modeling of non-linear systems using fast radial basis functoin
Author :
Fathalla, Joma M.
Author_Institution :
Gen. Public Comm. of Electr., Water & Gas
Abstract :
The radial basis function (RBF) is designed and implemented for modelling a nonlinear system. The combination of k-means clustering algorithm, p-nearest neighbour, least mean square together with the Gaussian function are used for determining the optimal network parameters adaptively instead of trial and error. In this paper an algorithm is presented to design (RBF) with smallest possible number of centres, proper width, and an optimal network output. An example of modelling a nonlinear system is simulated and the satisfactory results are obtained and illustrated.
Keywords :
Gaussian processes; least mean squares methods; nonlinear systems; parameter estimation; pattern clustering; radial basis function networks; Gaussian function; fast radial basis function; k-means clustering algorithm; least mean square; nonlinear system; optimal network parameter; p-nearest neighbour; Adaptive algorithm; Algorithm design and analysis; Artificial neural networks; Clustering algorithms; Multi-layer neural network; Neural networks; Nonlinear systems; Partitioning algorithms; Radial basis function networks; Signal design; K-means clustering; Non-linear system; RBFnetwork; adaptive algorithm; proper parameters;
Conference_Titel :
Systems, Signals and Devices, 2009. SSD '09. 6th International Multi-Conference on
Conference_Location :
Djerba
Print_ISBN :
978-1-4244-4345-1
Electronic_ISBN :
978-1-4244-4346-8
DOI :
10.1109/SSD.2009.4956724