Hostname: page-component-78c5997874-g7gxr Total loading time: 0 Render date: 2024-11-09T01:21:27.315Z Has data issue: false hasContentIssue false
  • English
  • Français

Convergence of the one-dimensional Kohonen algorithm

Part of: Sequential methods

Published online by Cambridge University Press:  01 July 2016

Michel Benaïm ,
Jean-Claude Fort  and
Gilles Pagès
Michel Benaïm*
Affiliation:
Université Paul Sabatier, Toulouse
Jean-Claude Fort*
Affiliation:
Université Nancy I and SAMOS-Paris I
Gilles Pagès*
Affiliation:
Universités Paris 12 and Paris 6, URA 224
*
∗ Postal address: Laboratoire de Statistique et Probabilitiés, Université Paul Sabatier, 118 Route de Narbonne, 31062, Toulouse, France.
∗∗ Postal address: Faculté des Sciences, Université Nancy 1, F-54506 Vaudœuvre-les-Nancy, Cedex, France.
∗∗∗ Laboratoire de Probabilités, Tour 56 3e étage, Université Pierre et Marie Curie, 4 Place Jussieu, F-75252 Paris Cedex 05, France. Email address: [email protected]
Get access Rights & Permissions [Opens in a new window]

Abstract

We show in a very general framework the a.s. convergence of the one-dimensional Kohonen algorithm–after self-organization–to its unique equilibrium when the learning rate decreases to 0 in a suitable way. The main requirement is a log-concavity assumption on the stimuli distribution which includes all the usual (truncated) probability distributions (uniform, exponential, gamma distribution with parameter ≥ 1, etc.). For the constant step algorithm, the weak convergence of the invariant distributions towards equilibrium as the step goes to 0 is established too. The main ingredients of the proof are the Poincaré-Hopf Theorem and a result of Hirsch on the convergence of cooperative dynamical systems.

Keywords

Kohonen algorithmstochastic algorithmcooperative dynamical system

MSC classification

Primary: 62L20: Stochastic approximation
Type
General Applied Probability
Information
Advances in Applied Probability , Volume 30 , Issue 3 , September 1998 , pp. 850 - 869
Copyright
Copyright © Applied Probability Trust 1998 

Access options

Get access to the full version of this content by using one of the access options below. (Log in options will check for institutional or personal access. Content may require purchase if you do not have access.)

References

Benaïm, M., (1996). A dynamical system approach to stochastic approximations. SIAM J. Control Optim. 34, 437–472.Google Scholar
Benaïm, M., (1998). Recursive algorithms, urn processes and chaining number of chain recurrent sets. Ergod. Th. & Dynam. Sys. 18, 53–87.Google Scholar
Bouton, C. and Pagès, G. (1993). Self-organization and convergence of the one-dimensional Kohonen algorithm with non-uniformly distributed stimuli. Stoch. Proc. Appl. 47, 249–274.CrossRefGoogle Scholar
Cohort, P. (1998). Unicité d'un quantifieur localement optimal par la methode du col, Technical Report, Laboratoire de Probabilités, Université Paris 6, France.Google Scholar
Conley, C. (1978). Isolated invariant set and the Morse index. Conference Board of the Mathematical Science AMS Regional Conference Series in Mathematics.Google Scholar
Cottrell, M. and Fort, J. C. (1987). Étude d'un algorithme d'auto-organisation. Ann. Inst. H. Poincaré B 23, 1–20.Google Scholar
Flanagan, A. (1994). Self-organizing neural networks. Thèse 1306, Ecole Polytechnique de Lausanne.Google Scholar
Fort, J. C. and Pagès, G. (1995). On the a.s. convergence of the Kohonen algorithm with a general neighborhood function. Ann. Appl. Prob. 5, 1177–1216.Google Scholar
Fort, J. C. and Pagès, G. (1996). About the Kohonen algorithm: strong or weak self-organization? Neural Networks 9, 773–785.CrossRefGoogle Scholar
Fort, J. C. and Pagès, G. (1996). Convergence of stochastic algorithms: from the Kushner & Clark theorem to the Lyapounov functional method. Adv. Appl. Prob. 28, 1072–1094.CrossRefGoogle Scholar
Fort, J. C. and Pagès, G. (1996). Asymptotic behaviour of a Markovian stochastic algorithm with constant step. To appear in SIAM J. Control Optimization.Google Scholar
Gantmacher, F. R. (1959). The Theory of Matrices. Vols 1 and 2. Chelsea, New York.Google Scholar
Hirsch, M. (1985). Systems of differential equations which are competitive or cooperative II: convergence almost everywhere. SIAM J. of Math. Anal. 16, 423–439.Google Scholar
Hirsch, M. (1988). Stability and convergence in strongly monotone dynamical systems. J. Reine Angew. Math. 383, 1–53.Google Scholar
Kohonen, T. (1982). Analysis of a simple self-organizing process. Biol. Cybernet. 44, 135–140.Google Scholar
Kohonen, T. (1989). Self-Organization and Associative Memory, 3rd edn. Springer, Berlin.CrossRefGoogle Scholar
Kunze, H. and Siegel, D. (1994). A graph theoretic approach to monotonicity with respect to initial conditions. In Comparison Methods and Stability Theory. eds. Liu, X. and Siegel, D.. Marcel Dekker, New York.Google Scholar
Kushner, H. J. and Clark, D. S. (1978). Stochastic Approximation for Constrained and Unconstrained Systems. Appl. Math. Sci. Series, 26, Springer, Berlin.Google Scholar
Métivier, M. and Priouret, P. (1987). Théorèmes de convergence presque sûre pour une classe d'algorithmes stochastiques. Prob. Theory Rel. Fields 74, 403–428.Google Scholar
Milnor, J. W. (1969) Topology from the Differential Viewpoint, 2nd edn. University Press of Virginia.Google Scholar
Ritter, H. and Schulten, K. (1988). Convergence properties of Kohonen's topology conserving maps: fluctuation, stability and dimension selection. Biol. Cybernet. 60, 59–71.Google Scholar
Sadeghi, A. (1996). Asymptotic behaviour of self-organizing maps with non-uniform stimuli distribution. Technical report n 166, Kaiserslautern University, Germany.Google Scholar
Smith, H. L. (1995) Monotone dynamical systems. An introduction to the theory of competitive and cooperative systems. Mathematical Surveys and Monographs 41, AMS, Providence, RI.Google Scholar