Abstract
A local robustness approach for the selection of the architecture in multilayered feedforward artificial neural networks (MLFANN) is studied in terms of probability density function (PDF) in this work. The method is used in a non-linear autoregressive (NAR) model with innovative outliers. The procedure is proposed for the selection of the locally most robust (around a particular sample) MLFANN architecture candidate for exact learning of a finite set of the real sample. The proposed selection method is based on the output PDF of the MLFANN. As each MLFANN architecture leads to a specific output PDF when its input is a distribution with heavy tails, a distance between probability densities is used as a measure of local robustness. A Monte Carlo study is presented to illustrate the selection method.
Work leading to this paper has been partially supported by the Ministry of Education and Research, Germany, under grant BMBF-CH-99/023 and the Technical University Federico Santa Maráa, Chile, under grant 240022-DGIP.
This is a preview of subscription content, log in via an institution.
Buying options
Tax calculation will be finalised at checkout
Purchases are for personal use only
Learn about institutional subscriptionsPreview
Unable to display preview. Download preview PDF.
References
Allende H., Moraga C.: Time Series Forecasting with Neural Networks. Forschungsbericht No 727/2000. Universität Dortmund Fachbereich Informatik. (2000)
Allende H., Heiler S.: Recursive Generalized M-Estimates for Autoregressive Moving Average Models. Journal of Time Series Analysis. 13, 1–18 (1992)
Castro J.L., Trillas E.: The logic of neural networks. Mathware and Softcomputing V(1), 23–27, (1998)
Choueiki M.H. and Mount-Campbell C.A.: Training data development with the optimality criterion. IEEE Transactions on Neural Networks, 10, 56–63 (1999)
Connor J. T., Martin R. D.: Recurrent Neural Networks and Robust Time Series Prediction. IEEE Transactions of Neural Networks 5(2), 240–253, (1994)
Connor J. T.: A robust Neural Networks Filter for Electricity Demand Prediction. J. of Forecasting 15, 437–458 (1996)
Faraway J., Chatfield C.: Time series with Neural Networks. A comparative study using the airline data. Appl. Statist. 47(2), 231–250 (1998)
Gabr M.M.: Robust estimation of bilinear time series models. Comm. Statist. Theory and Meth 27(1), 41–53 (1998)
Hansen J., Nelson R.: Neural Networks and traditional time series methods. IEEE transaction on Neural Networks 8(4), 863–873. (1997)
Heider R., Moraga C.; Evolutionary synthesis of neural networks based on graph-grammars. Proc. Int. Conf. On Soft and Intelligent Computing. 119–128, Budapest. ISBN 9634205100, (1996)
Hisham M. and Mount-Campell C.A.: Training data development with the D-optimality criterion. IEEE Transaction on Neural Networks Vol 10, 56–63 (1999)
Lachtermacher G. and Fuller J.D.: Backpropagation in Time Series. J. of Forecasting 14, 381–393 (1995)
Moraga C., Salinas L.: Interpreting neural networks in the frame of the logic of Lukasiewicz. Proceedings Int’l Workconference on artificial and natural neural networks IWANN’2001. Granada, Spain. Springer, (2001)
Moraga C., Temme K.-H.: S-neural networks are fuzzy models. Proceedings Int’l Conference on Information Processing and Management of Uncertainty in Knowledge based Systems, 1518–1523. Technical University of Madrid, Spain, (2000)
Subba Rao, T.: On the theory of bilinear time series models. Comm. Statist. Soc. B 43(2), 244–255 (1981)
Tierney L., Kass R. and Kadane JB.: Approximative marginal densities of non-linear functions. Biometrika 76, 425–433. Correction in Biometrika 78, 233–234.
White H.: Learning in Neural Networks: A statistical perspective. Neural Computation 1 425–464 (1989)
White H.: Artificial Neural Networks: Approximation and Learning Theory, Basil Blackwell, Oxford (1992)
White H.: Estimate Inference and Specification Analysis, New York: Cambridge University Press, (1994)
Referes A. P. N., Zapranis A. D.: Neural Model Identification, Variable Selection and Model adequacy. J. Forecasting 18, 299–322, (1999).
Yao X.; Evolution of Artificial Neural Networks, Proc. IEEE 87(9), 47–49. (1999)
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2001 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Allende, H., Moraga, C., Salas, R. (2001). Neural Model Identification Using Local Robustness Analysis. In: Reusch, B. (eds) Computational Intelligence. Theory and Applications. Fuzzy Days 2001. Lecture Notes in Computer Science, vol 2206. Springer, Berlin, Heidelberg. https://doi.org/10.1007/3-540-45493-4_21
Download citation
DOI: https://doi.org/10.1007/3-540-45493-4_21
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-42732-2
Online ISBN: 978-3-540-45493-9
eBook Packages: Springer Book Archive