The International Arab Journal of Information Technology (IAJIT)


Approximating I/O Data Using Wavelet Neural

 In this paper, we deal with the problem of function approximation from a given set of input/output dat a. This problem consists of analyzing training examples, so that we can predict the output of the model given new inputs. We present a new method for function approximation of the I/O da ta using Wavelet Neural Networks (WNN). This method is based on a new efficient method of optimizing the position of a si ngle function called mother wavelet of the WNN; it uses the objective output of WNN to move the position of wavelet single funct ion. This method calculates the error committed in every mother wavelet area using the real output of the WNN trying to con centrate more mother wavelets in those input region s where the error is bigger, thus attempting to homogenize the contribut ion to the error of every mother wavelet, this meth od improves the performance of the approximation system obtained, c ompared with other models derived from traditional algorithms.

[1] Awad M., Pomares H., Herrera. L., Gonz lez J., Guill n A., and Rojas F., Approximating I/O Data Using Radial Basis Functions: A New Clustering-Based Approach, in Proceedings of 8 th International Workshop on Artificial Neural Networks, Spain, pp. 289-296, 2005.

[2] Bugmann G., Normalized Gaussian Radial Basis Function Networks, Technical Report, Centre for Neural and Adaptive Systems, School of Computing, University of Plymouth, United Kingdom, 1998.

[3] Cao Z., Guo F., and Wang J., Research on Flux Observer Based on Wavelet Neural Network Adjusted by Antcolony Optimization, in Procceedings of International Conference on Machine Learning and Cybernetics , Beijing, pp. 862-866, 2007.

[4] Chen Y., Yang B., and Dong J., Time-Series Prediction Using A Local Linear Wavelet Neural Network, Computer Journal of Neurocomputing , vol. 69, no. 4-6, pp. 449-465, 2006.

[5] Delyon B., Juditsky A., and Benveniste A., Accuracy Analysis for Wavelet Approximations, IEEE Transactions Neural Networs , vol. 6, no. 2, pp. 332-348, 1995.

[6] Gonzalez J., Rojas I., and Pomares H., A New Clustering Technique for Function Approximation, IEEE Transactions Neural Networks , vol. 13, no. 1, pp. 132-142, 2002.

[7] Higgins C., Classifications and Approximation with Rule-Based Networks, Dissertation PhD, California Institute of Technology, pp. 33-36, 1993.

[8] Kanjilal P. and Banerjee N., On the Application of Orthogonal Transformation for the Design and Analysis of Feedforward Networks, IEEE Transactions Neural Network s, vol. 6, no. 5, pp. 1061-1070, 1995.

[9] Krishnaprasad. S. and Pati C., Analysis and Synthesis of Feed Forward Neural Networks Using Discrete Affine Wavelet Transformations, IEEE Transactions on Neural Networks , vol. 4, no. 1, pp. 73-85, 1993.

[10] Li X. and Liu D., Modeling and Optimal for Vacuum Annealing Furnace Based on Wavelet Neural Networks with Adaptive Immune Genetic Algorithm, in Procceedings of International Conference on Advances in Natural Computation , China, pp. 922-930, 2005.

[11] Marquardt W., An Algorithm for Least-Squares Estimation of Nonlinear Inequalities, Computer of Journal on Applied Mathematics , vol. 11, no. 2, pp. 431-441, 1963.

[12] Moody E. and Darken C., Fast Learning in Networks of Locally Tuned Processing Units, Neural Computation , vol. 2, no. 1, pp. 281-294, 1989.

[13] Pomares H., Rojas I., Ortega J., Gonz lez J., and Prieto A., A Systematic Approach to Self- Generating Fuzzy Rule-Table for Function Approximation, in Procceedings of IEEE Transactions System , vol. 30, no. 3, pp. 431-447, 2000.

[14] Sheng L. and Shu-Ching C., Function Approximation Using Robust Wavelet Neural Networks, in Procceedings of 14 th IEEE International Conference on Tools with Artificial Intelligence , Taiwan, pp. 483-488, 2002.

[15] Vachtsevanos G. and Wang P., A Wavelet Network Framework for Diagnostics of Complex Engineered Systems, in Proceedings of MARCON Maintenance and Reliability Conference , Atlanta, pp. 623-637, 1998.

[16] Wallich P., Wavelet Theory: An Analysis Technique that s Creating Ripples, Technical Document , Scientific American, pp. 34-35, 1991. Approximating I/O Data Using Wavelet Neural Networks: Control the Position of Mother Wavelet 29

[17] Zhang J., Walter G., Miao Y., and Lee W., Wavelet Neural Networks for Function Learning, IEEE Transactions Signal Processing , vol. 43, no. 6, pp. 1485-1497, 1995.

[18] Zhang Q. and Benveniste A., Wavelet Networks, IEEE Transactions on Neural Networks , vol. 3, no. 6, pp. 889-898, 1992. Mohammed Awad received his BSc degree in industrial automation engineering in 2000, from the Palestine Polytechnic University, and his PhD degree in 2005 from University of Granada, Spain. From February, 2006 till now he is an assistant professor in the Faculty of Engineering a nd Information Technology at the Arab American University, Palestine. From February, 2006 till now he is the chair of the Department of Computer Information Technology (CIT) at the Arab American University, Palestine. His current areas of research interest include artificial neural networks and evolutionary computation, function approximation using radial basis function neural networks, input variable selection and fuzzy and neuro-fuzzy system .