Advertisement

Abstract

Mixture modelling is a hot area in pattern recognition. This paper focuses on the use of Bernoulli mixtures for binary data and, in particular, for binary images. More specifically, six EM initialisation techniques are described and empirically compared on a classification task of handwritten Indian digits. Somehow surprisingly, we have found that a relatively good initialisation for Bernoulli prototypes is to use slightly perturbed versions of the hypercube centre.

Keywords

Mixture Models EM Algorithm Multivariate Bernoulli Distribution Initialisation Techniques Binary Data Indian Digits 

References

  1. 1.
    Al-Ohali, Y., Cheriet, M., Suen, C.: Databases for recognition of handwritten Arabic cheques. Pattern Recognition 36, 111–121 (2003)zbMATHCrossRefGoogle Scholar
  2. 2.
    Carreira-Perpiñán, M.A., Renals, S.: Practical identifiability of finite mixtures of multivariate Bernoulli distributions. Neural Computation 12(1), 141–152 (2000)CrossRefGoogle Scholar
  3. 3.
    Dempster, A.P., Laird, N.M., Rubin, D.B.: Maximum likelihood from incomplete data via the EM algorithm (with discussion). Journal of the Royal Statistical Society B 39, 1–38 (1977)zbMATHMathSciNetGoogle Scholar
  4. 4.
    Duda, R.O., Hart, P.E.: Pattern Classification and Scene Analysis. Wiley, Chichester (1973)zbMATHGoogle Scholar
  5. 5.
    González, J., Juan, A., Dupont, P., Vidal, E., Casacuberta, F.: A Bernoulli mixture model for word categorisation. In: Proc. of the IX Spanish Symposium on Pattern Recognition and Image Analysis, Benicàssim (Spain), May 2001, vol. I, pp. 165–170 (2001)Google Scholar
  6. 6.
    Grim, J., Pudil, P., Somol, P.: Multivariate Structural Bernoulli Mixtures for Recognition of Handwritten Numerals. In: Proc. of the ICPR 2000, Barcelona (Spain), September 2000, vol. 2, pp. 585–589 (2000)Google Scholar
  7. 7.
    Jain, A.K., Duin, R.P.W., Mao, J.: Statistical Pattern Recognition: A Review. IEEE Trans. on PAMI 22(1), 4–37 (2000)Google Scholar
  8. 8.
    Juan, A., Vidal, E.: Bernoulli mixture models for binary images. In: Proc. of the ICPR 2004 (2004) (submitted)Google Scholar
  9. 9.
    Juan, A., Vidal, E.: On the use of Bernoulli mixture models for text classification. Pattern Recognition 35(12), 2705–2710 (2002)zbMATHCrossRefGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2004

Authors and Affiliations

  • Alfons Juan
    • 1
  • José García-Hernández
    • 1
  • Enrique Vidal
    • 1
  1. 1.DSICUniversitat Politècnica de ValènciaValènciaSpain

Personalised recommendations