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

Performance analysis of LVQ algorithms: a statistical physics approach.

Anarta Ghosh1, Michael Biehl, Barbara Hammer

  • 1Rijksuniversiteit Groningen, Mathematics and Computing Science, P.O. Box 800, NL-9700 AV Groningen, The Netherlands. anarta@cs.rug.nl

Neural Networks : the Official Journal of the International Neural Network Society
|June 20, 2006
PubMed
Summary
This summary is machine-generated.

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

Enhanced broad-band intermuscular coherence in myoclonus: a targeted characterization study.

Clinical neurophysiology : official journal of the International Federation of Clinical Neurophysiology·2026
Same author

Machine learning based EMG analysis of intermuscular coherence and cumulant density in tremor and myoclonus.

Clinical neurophysiology : official journal of the International Federation of Clinical Neurophysiology·2026
Same author

The Impact of Early Postoperative Oral Intake on the Length of Hospital Stay: A Meta-Analysis.

Acta anaesthesiologica Scandinavica·2026
Same author

Endocrine and metabolic determinants of cardiometabolic risk in mild autonomous cortisol secretion.

EBioMedicine·2026
Same author

The reliability of remote photoplethysmography under low illumination and elevated heart rates.

NPJ digital medicine·2025
Same author

ATTNFNET: feature aware depth-to-pressure translation with cGAN training.

Frontiers in medical technology·2025

Learning Vector Quantization (LVQ) algorithms are analyzed for performance. Surprisingly, the original LVQ1 demonstrates superior stability and generalization compared to newer methods.

Area of Science:

  • Machine Learning
  • Pattern Recognition
  • Statistical Physics

Background:

  • Learning Vector Quantization (LVQ) is a heuristic-based classification method with many modifications for improved performance.
  • Existing LVQ algorithms lack thorough investigation into their dynamics and generalization capabilities.
  • Recent work provides a mathematical foundation for LVQ, leading to new variants with enhanced stability.

Purpose of the Study:

  • To develop a mathematical framework for analyzing the performance of various LVQ algorithms.
  • To investigate the dynamics, sensitivity to initial conditions, and generalization ability of different LVQ variants.
  • To compare the performance of five distinct LVQ algorithms.

Main Methods:

  • Utilizing concepts from statistical physics and on-line learning theory.

Related Experiment Videos

  • Developing a mathematical framework to analyze algorithmic dynamics and generalization.
  • Conducting a detailed comparative study of five LVQ algorithms: LVQ1, VQ, VQ-LVQ mixture, LVQ2.1, and a cost-function-based LVQ.

    • Main Results:

    • Significant differences in stability and generalization were observed among LVQ variants.
    • Kohonen's original LVQ1 exhibited robust performance, often outperforming recent proposals.
    • LVQ1 demonstrated excellent stability, asymptotic generalization ability, and robustness to initialization and model parameters.

    Conclusions:

    • The mathematical framework provides insights into LVQ algorithm performance.
    • LVQ1 remains a highly competitive algorithm despite its age and heuristic origins.
    • Further research into LVQ dynamics and generalization is warranted, with LVQ1 serving as a strong benchmark.