Geometric Algebra Based Neural Networks

  • Mirek Kárný
  • Kevin Warwick
  • Vera Kůrková
Part of the Perspectives in Neural Computing book series (PERSPECT.NEURAL)

Abstract

In the last period there were introduced some new types of neural networks based on a more general type of algebra. They obey the multidimensional neural activities compared to the one-dimensional neural activity within the standard neural network framework. Instead of using basically the product of two scalar values they utilise some special algebraic product of two multidimensional quantities. Most of them can be considered as a special type of geometric (Clifford) algebra [1], [2] based neural networks (GANNs).

Keywords

Arena 

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. [1]
    Hestenes D. and Sobczyk G. Clifford Algebra to Geometric Calculus. D. Reidel Publishing Company, 1984.CrossRefMATHGoogle Scholar
  2. [2]
    Hestenes D. New Foundations for Classical Mechanics. D. Reidel Publishing Company, 1986.CrossRefMATHGoogle Scholar
  3. [3]
    Arena P., Fortuna L., Re R., and Xibilia MG. On the capability of neural networks with complex neurons in complex-valued function approximation. In Proceedings IEEE International Conference on Circuit and Systems, pages 2168–2171, Chicago, 1993.Google Scholar
  4. [4]
    Benvenuto N., Marchesi M., Piazza F., and Uncini A. A comparison between real and complex-valued neural networks in communication applications. In Kohonen et al., editor, Artificial Neural Networks, pages 1177–1180. Elsevier Science Publishers B.V. ( North-Holland ), 1991.Google Scholar
  5. [5]
    Birx DL. and Pipenberg SJ. A complex mapping network for phase sensitive classification. IEEE Transactions on Neural Networks, 4 (1): 127–135, 1993.CrossRefGoogle Scholar
  6. [6]
    Hirose A. Dynamics of fully complex-valued neural networks. Electronic Letters, 28 (16): 1492–1494, 1992.CrossRefGoogle Scholar
  7. [7]
    Hirose A. Continuous complex-valued back-propagation learning. Electronic Letters, 28 (20): 1854–1855, 1992.CrossRefGoogle Scholar
  8. [8]
    Leung H. and Haikyn S. The complex backpropagation algorithm. IEEE Transactions on Signal Processing, 39 (9): 2101–2104, 1991.CrossRefGoogle Scholar
  9. [9]
    Arena P., Fortuna L., Muscato G., and Xibilia MG. Multilayer perceptrons to approximate quaternion-valued functions. Neural Networks, 9(6): 1–8, 6 1996.Google Scholar
  10. [10]
    Masters T. Signal and Image Processing with Neural Networks: a C++ Sourcebook. John Wiley Sons, 1994.Google Scholar
  11. [11]
    Blow Th. and Sommer G. Real multidimensional fourier transform. In SCIA ‘87, 1997. submitted.Google Scholar
  12. [12]
    Chudÿ L. and Chudÿ V. Why complex-valued neural networks? (on the possible role of geometric algebra framework for the neural network computation). In Proceedings of IEEE Workshop on Computer-Intensive Methods in Control and Dignal Processing, pages 109–114, Prague, 1996.Google Scholar
  13. [13]
    Kalman BL. and Kwasny SC. Why tanh? choosing a sigmoidal function. In Proceedings of Int. Joint Conference on Neural Networks, Baltimore, MD, 1991.Google Scholar
  14. [14]
    Friedman JH. An overwiev of predictive learning and function approximation. In V. Cherkassky, J. Friedman, and H. Wechsler, editors, From statistics to neural networks., NATO ASI Series, pages 1–62. Springer-Verlag, 1994.Google Scholar
  15. [15]
    Hertz J. and Krogh A. Palmer RG. Introduction to the theory of neural computation. Addison-Wesley, 1991.Google Scholar

Copyright information

© Springer-Verlag London Limited 1998

Authors and Affiliations

  • Mirek Kárný
    • 1
  • Kevin Warwick
    • 2
  • Vera Kůrková
    • 3
  1. 1.Institute of Information Theory & AutomationPrague 8Czech Republic
  2. 2.Department of CyberneticsUniversity of ReadingWhiteknights, ReadingUK
  3. 3.Institute of Computer ScienceAcademy of Sciences of the Czech RepublicPrague 8Czech Republic

Personalised recommendations