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

Convergent decomposition techniques for training RBF neural networks.

C Buzzi1, L Grippo, M Sciandrone

  • 1Dipartimento di Informatica e Sistemistica, Università di Roma "La Sapienza," Via Buonarroti 12 00185, Roma, Italy.

Neural Computation
|August 17, 2001
PubMed
Summary

This study introduces globally convergent decomposition algorithms for training generalized radial basis function neural networks. These novel methods enhance supervised learning by efficiently decomposing network parameters for improved convergence.

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

Convergent on-line algorithms for supervised learning in neural networks.

IEEE transactions on neural networks·2008
Same author

Efficient training of RBF neural networks for pattern recognition.

IEEE transactions on neural networks·2008
Same author

Decreased protein tyrosine phosphorylation and membrane fluidity in spermatozoa from infertile men with varicocele.

Molecular reproduction and development·2006
Same author

[ELECTROPHORETIC AND IMMUNO-ELECTROPHORETIC STUDIES ON THE CARBONIC ANHYDRASES OF THE HUMAN RED BLOOD CELLS].

Biochimica et biophysica acta·1964
Same author

[Study of the euglobulin behavior of agglutinating factors in chronic evolutive polyarthritis and other diseases].

Bulletin der Schweizerischen Akademie der Medizinischen Wissenschaften·1962
Same author

[Chromatographic fractionation on carboxymethyl-cellulose of normal serums and serums of patients with disseminated lupus erythematosus (L.E. serum). Immuno-electrophoretic study of protein fractioons].

Revue du rhumatisme et des maladies osteo-articulaires·1961

Area of Science:

  • Artificial Intelligence
  • Machine Learning
  • Computational Neuroscience

Background:

  • Generalized Radial Basis Function (GRBF) neural networks are powerful tools for complex function approximation.
  • Supervised training of GRBF networks often faces challenges with parameter optimization and convergence.
  • Existing training algorithms may struggle with computational efficiency and achieving global convergence.

Purpose of the Study:

  • To develop and analyze globally convergent decomposition algorithms for supervised training of GRBF neural networks.
  • To improve the efficiency and reliability of GRBF network training.
  • To provide a theoretical framework and practical validation for the proposed algorithms.

Main Methods:

  • Decomposition algorithms are proposed, splitting network parameters into weights and centers.

Related Experiment Videos

  • A sequential minimization approach is used for center location selection.
  • A criterion for updating centers at each step is defined to ensure convergence.

    • Main Results:

    • The global convergence of the proposed decomposition algorithms is theoretically proven.
    • Computational results demonstrate the effectiveness of the algorithms on various test problems.
    • The methods show promise for efficient and robust GRBF network training.

    Conclusions:

    • The developed decomposition algorithms offer a reliable approach for supervised GRBF network training.
    • These algorithms achieve global convergence, addressing a key limitation in existing methods.
    • The findings contribute to the advancement of neural network training methodologies.