Determining the structures and parameters of radial basis function neural networks using improved genetic algorithms

Abstract

The method of determining the structures and parameters of radial basis function neural networks (RBFNNs) using improved genetic algorithms is proposed. Akaike’s information criterion (AIC) with generalization error term is used as the best criterion of optimizing the structures and parameters of networks. It is shown from the simulation results that the method not only improves the approximation and generalization capability of RBFNNs, but also obtain the optimal or suboptimal structures of networks.

Keywords

radial basis function neural network / genetic algorithms / Akaike’s information criterion / overfitting

References

[1] PowellM J DMasonJ C, CoxM G. Radial basis functions for multivariable interpolation: A review. Algorithms for Approximation, 1985, Oxford, Clarendon Press: 143-167

[2] Broomhead D S, Lowe D. Multivariable functional interpolation and adaptive networks. Complex Systems, 1988(2):321–355

[3] LightW A. Some aspects of radial basis function approximation. Approximation Theory, Spline Functions and Applications, 1992, 356: 163-190

[4] PowellM J DLightW A. The theory of radial basis function approximation in 1990. Advances in Numerical Analysis, 1992, Oxford, Oxford University Press: 105-210

[5] BishopC M. Improving the generalization properties of radial basis function neural network. Neural Computation, 1991, 3(4): 579-588

[6] LiuJianqinArtificial life theory and application, 1997, Beijing, Metallurgy Industry Press: 136-142(in Chinese)

[7] ChenS, BillingsS A, CowanC F N, et al.. Practical identification of NARMAX models using radial basis functions. Int J Control, 1990, 52(6): 1327-1350

[8] LucasiusC B, KatemanGMannerR, ManderickB. Towards solving subset selection problems with the aid of the genetic algorithm. Parallel Problem Solving from Nature (Vol. 2), 1992, Amsterdam, Elsevier Science Publishers

[9] FonsecaC M, MendesE M, FlemingP J, et al.. Non-linear model term selection with genetic algorithms. Proceedings of IEE/IEEE Workshop on Natural Algorithms in Signal Processing, 1993, 2: 271-278