Advertisement

Symbolic Representation of a Multi-Layer Perceptron

  • Fériel Mouria-Beji
Conference paper

Abstract

We propose a Top-Down Inferring algorithm Tdinfer for artificial neural network rule extraction. These rules formalize the decision process of a standard multi-layer network and make its prediction explicit and understandable. They do not involve any weight values and no restrictions are made on the activation values. The algorithm is applied to a speech and character recognition problems.

Keywords

Output Layer Input Layer Hide Unit Antecedent Clause Syntactic Pattern Recognition 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. [1]
    Mouria-Beji, F.: CODEPHON-NN: A context-dependent phonemic model based on neural networks. In Proc. IEEE Int. multiconference on Computational Engineering in Systems Applications. IEEE-SMC, April (1998).Google Scholar
  2. [2]
    Mouria-Beji, F.: Neural network use in a non-linear vectorial interpolation technique for speaker recognition. In IEEE World Congress on Computational Intelligence, IEEE/IJCNN. vol. 2, pp. 1200–1205, Anchorage, Alaska. May (1998).Google Scholar
  3. [3]
    Peifer, H., Gutknecht, M. and Stolze, M.: Cooperative hybrid systems. In Proceedings of 11th IJCAI, pages 824–829, (1991).Google Scholar
  4. [4]
    Choi, E. C. Y. and Gedeon, T. D.: Comparison of extracted rules from multiple networks. IEEE Transactions on Neural Networks, 6, January (1995).Google Scholar
  5. [5]
    Mitrad, S.: Puzzy multi-Iayer perceptron, inferencing and rule generation. IEEE Transactions on Neural Networks, 6(1), January (1995).Google Scholar
  6. [6]
    Terrace, H. M. K. and Ridge, K.: Integating rules and neural computation. IEEE Transactions on Neural Networks, 6, January (1995).Google Scholar
  7. [7]
    Quinlan, J. R.: Comparing connectionist and symbolic learning methods. In Drastall, G.A., Hanson, S.J. and Rivest, R.I., editors, Computational Learning Theory and Natural Learning Systems. MIT Press, Cambrige, Mass., (1994).Google Scholar
  8. [8]
    Mooney, R.J., Shavilk, J.W. and Towell, G.G.: Symbolic and neural learning algorithms: An experimental comparison. Machine Learning, 6(2):111–143, Mar. (1991).Google Scholar
  9. [9]
    Yamashita, K., Hirose, Y. and Hijiya, S.: Back propagation algorithm wich varies the number of hidden unitsintegating rules and neural computation. Transactions on Neural Networks, 4:61–66, January (1991).Google Scholar
  10. [10]
    Towell, G. G. and Shavlik, J. W., Extracting refined rules from knowledge-based neural networks. Machine Learning, 13(1):71–101, Oct. (1993).Google Scholar
  11. [11]
    Pu, K. S.: Error-correcting parsing for syntactic pattern recognition. In Klinger, A., Pu, K. S. and Kunii, T. L., editors, Data Structures, Computer Graphics, and Pattern Recognition, pages 449–492. Academic Press, Inc., (1977).Google Scholar
  12. [12]
    Mouria-Beji, F. and Boulahia, J.: ANNREX: An algorithm for neural network rule extraction. In Proc. IEEE Int. multiconference on Computational Engineering in Systems Applications. IEEE-SMC, April (1998).Google Scholar
  13. [13]
    Sima, J.: Neural expert systems. Transactions on Neural Networks, 8(2):261–271, January (1995).CrossRefGoogle Scholar
  14. [14]
    Hayes, S., Ciesielsk, U. and Kelly, B.: Comparaison of an expert system and a hybrid neural network. In AAAI-92 Workshop on Integrating Neural and Symbolic Process, the Cognitive Dimension, San Jose, California, (1992).Google Scholar
  15. [15]
    Pellegrini, C., Hilario, M. and Alexandre, F.: Modular integration of connexioniste and symbolic processing in knowledge based systems. In Proceedings of Int. Symposium on Integrating Knowledge and Neural Heuristics, pages 824–829, Pensacola Beach, Florida, (1994).Google Scholar
  16. [16]
    Hinton, G. E., Rummelhart, D. E. and Williams, R. J.: Learning representations by error propagation. In Mc Clelland, J. L., Rummelhart, D. E. and the PDP Research Group, editors, Parallel Distributed Processing. Mc Graw-Hill Book Company, Cambrige, MA, MIT Press.Google Scholar
  17. [17]
    Mouria-Beji, F. and Boulahia, J.: Extraction and insertion rules during the training process of a neural network. In the International Conference on Artificial and Computational Intelligence for Decision, Control and Automation in Engineering and Industrial Applications, March (2000).Google Scholar

Copyright information

© Springer-Verlag Wien 2001

Authors and Affiliations

  • Fériel Mouria-Beji
    • 1
    • 2
  1. 1.Artificial Intelligence GroupENSI/LIATunisTunisia
  2. 2.INRIA/LORIAVillers-lès-NancyFrance

Personalised recommendations