» Download Article

Authors
Pizzileo, Barbara; Li, Kang; Irwin, George W.

Identifier
10.3182/20090706-3-FR-2004.00250

Index Terms
Nonlinear System Identification; Neural Networks; Machine Learning and Data Mining

Abstract
In identification of complex dynamic systems using fuzzy neural networks, one of the main issues is the curse-of-dimensionality, which makes it difficult to quickly compute all the parameters associated with the network with all possible inputs and rules being included. In the literature this issue has been addressed by the selection of either the inputs or the rules. This is due to the fact that not all possible inputs or rules have to be necessarily included because of the correlations between them. Adding unnecessary inputs or rules simply increases the model complexity and worsens the network generalization performance. Selecting the best set of inputs or rules is a combinational problem and can be computationally too expensive. In this paper, the problem is solved by first proposing a refinement procedure for rule selection. The algorithm is then adapted and integrated with prior input selection to further improve the model accuracy. Simulation results confirm the efficacy of the method.

References

REFERENCES

J. Yen, and L. Wang. Simplifying fuzzy rule-based models
using orthogonal transformation methods. IEEE Trans.
Systems, Man and Cybern-Pt B: Cybern, volume 29,
no.1, pages 13-23, 1999.

X. Hong, C. J. Harris, and S. Chen. Robust neurofuzzy
rule base knowledge extraction and estimation using
subspace decomposition combined with regularization
and D-optimality. IEEE Trans Syst Man Cybern-Pt B:
Cybern, volume 34, pages 598-608, 2004.

X. Hong and C. J. Harris. A neurofuzzy network knowl-
edge extraction and extended Gram-Schmidt algorithm
for model subspace decomposition. IEEE Trans Fuzzy
Syst, volume 11, pages 528-541, 2003.

K. Li, J. Peng and G. W. Irwin. A fast nonlinear model
identi¯cation method. IEEE Trans on Automat. Contr.,
volume 50, number 8, pages 1211-1216, 2005.

B. Pizzileo, K. Li and G. W. Irwin. A fast fuzzy neural
modeling method for nonlinear dynamic system. LNCS,
Springer Berlin, vol. 4491, pp. 496-504, July, 2007.

A. Sherstinsky and R. W. Picard. On the e±ciency
of the orthogonal least squares training method for
Radial Basis Function networks. IEEE Trans. on Neural
Networks, volume 1, pages 195-200, 1996.

K.Z Mao and S.A. Billings. Algorithms for minimal model
structure detection in non linear dynamical system
identi¯cation. Int. J. of Contr., volume 68, number 2,
pages 311-330, 1997.

K. Li, J. Peng and E. W. Bai. A two-stage algorithm for
identi¯cation of nonlinear dynamic systems. Automat-
ica, volume 42, number 7, pages 1189-1197, 2006.

B. Pizzileo, K. Li, and G. W. Irwin. A fast method for
neural network modelling and re¯nement. Accepted for
a Special Issue of IJMIC, 2008.

B. Pizzileo, and K. Li. A new fast algorithm for fuzzy rule
selection. IEEE Int Conf on Fuzzy Systems, pp. 1-6,
Imperial College, London, 23-26 July 2007.

M. Setnes, and R. Babuska. Rule Base Reduction: Some
Comments on the Use of Orthogonal Transforms. IEEE
Trans. Systems, Man and Cybern-Pt C: applications and
reviews, volume 31, no. 2, pages 199-206, May 2001.

X. Hong and C. J. Harris . Variable selection algorithm for
the construction of MIMO operating point dependent neurofuzzy networks. IEEE Trans Fuzzy Syst, volume 9, pages 88-101, 2001.