ICANN ’93 pp 609-612 | Cite as

Analyzing Kohonen Maps with Geometry

  • Stéphane Zrehen


An organization measure based on geometrical criteria has already been proposed for Kohonen maps in the two-dimensional case [3]. This measure is shown to be generalizable to all network topologies and input dimensions. It can be used for demonstrations on the convergence of the learning algorithm.


Learning Algorithm Weight Vector Input Space Voronoi Tessellation Input Distribution 
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.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. [1]
    Kohonen T., (1982): Self-Organization of Topologically Correct Feature Maps, Biol. Cyb.43, 59–69MathSciNetCrossRefMATHGoogle Scholar
  2. [2]
    Kohonen T., (1984): Self-Organization and Associative Memory (2nd ed.) Berlin, Springer Verlag.MATHGoogle Scholar
  3. [3]
    Zrehen S., Blayo F. (1992): A Geometric Organization Measure for Self-Organizing Kohonen Maps, Proceedings of Neuro-Nîmes Conference 1992,p. 603–610, Paris: EC2.Google Scholar
  4. [4]
    Demartines P. (1992): Mesures d’Organisation du Réseau de Kohonen. Congrès Satellite du Congrès Européen de Mathématiques, Paris.Google Scholar
  5. [5]
    Bauer H-U, Pawelczik K (1992), Quantifying the Neighborhood Preservation of Self-Organizing Maps, IEEE Transactions on Neural Networks, 1992.Google Scholar
  6. [6]
    Cottrell M. Fort J.C, (1987): Etude d’un processus d’auto-organisation. Annales de l’Institut Henri Poincaré 23, n° 1, pp. 1–20.MathSciNetMATHGoogle Scholar
  7. [7]
    Cottrell M., Fort J.C, (1986): A Stochastic Model of Retinotopoy: A Self-Organizing Process, Biol. Cyb. 53, 405–411.MathSciNetCrossRefMATHGoogle Scholar

Copyright information

© Springer-Verlag London Limited 1993

Authors and Affiliations

  • Stéphane Zrehen
    • 1
  1. 1.Laboratoire de MicroinformatiqueEPFL-DILausanneSwitzerland

Personalised recommendations