Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

About the Kohonen Algorithm: Strong or Weak Self-organization?

Gilles Pagès1, Jean Claude Fort

  • 1University of Paris 12, URA 224, France

Neural Networks : the Official Journal of the International Neural Network Society
|July 1, 1996
PubMed
Summary

The Kohonen algorithm, used in neural networks, does not exhibit strong self-organization. This study mathematically and computationally proves this limitation in common configurations.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

STK17A (DRAK1) at the crossroads of apoptosis, immunity, and cancer: Emerging roles and therapeutic opportunities.

Theranostics·2026
Same author

Reply to Roesler et al. Comment on "Karaulic et al. Exploring Novel Applications: Repositioning Clinically Approved Therapies for Medulloblastoma Treatment. <i>Cancers</i> 2025, <i>17</i>, 3659".

Cancers·2026
Same author

Connective tissue growth factor contributes to resistance to anti-angiogenic therapies in renal cancer.

Theranostics·2026
Same author

Improving angiogenesis ameliorates the efficacy of ASO-based exon skipping for the treatment of Duchenne muscular dystrophy.

Molecular therapy. Nucleic acids·2026
Same author

Exploring Novel Applications: Repositioning Clinically Approved Therapies for Medulloblastoma Treatment.

Cancers·2025
Same author

IF-CRIB: A 3D-printable device to facilitate immunofluorescence experiments and its application in screening and characterizing cells expressing a degradable form of ERK2.

Methods (San Diego, Calif.)·2025

Area of Science:

  • Artificial Intelligence
  • Machine Learning
  • Computational Neuroscience

Background:

  • The Kohonen algorithm, a type of self-organizing map, is widely used for dimensionality reduction and data visualization.
  • Understanding the self-organization properties of such algorithms is crucial for their reliable application.
  • Previous research has explored various aspects of Kohonen algorithm behavior, but its strong self-organization capacity remains a key question.

Purpose of the Study:

  • To rigorously investigate the self-organization properties of the Kohonen algorithm.
  • To precisely define and differentiate between weak and strong self-organization states.
  • To determine if the Kohonen algorithm possesses the strong self-organization property under specific, common conditions.

Main Methods:

Related Experiment Videos

  • Formal mathematical definitions of organized states, weak, and strong self-organization were established.
  • A combination of theoretical mathematical analysis and computational simulations was employed.
  • The study focused on two specific, well-known configurations of the Kohonen algorithm.

Main Results:

  • The Kohonen algorithm was proven not to possess the strong self-organization property.
  • This finding holds true for a stimuli space of [0, 1](2) with a linear unit set and the two-nearest neighbor function.
  • The result was also confirmed for a grid unit set with the eight-nearest neighbor function.

Conclusions:

  • The Kohonen algorithm, in its commonly used forms, lacks the strong self-organization property.
  • These findings have implications for the theoretical understanding and practical application of self-organizing maps.
  • Further research may be needed to explore conditions under which stronger forms of self-organization might emerge or alternative algorithms.