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

Iterative fast orthogonal search algorithm for MDL-based training of generalized single-layer networks.

K M Adeney1, M J Korenberg

  • 1Department of Electrical and Computer Engineering, Queen's University, Kingston, Ontario, Canada. kathryn_adeney@ieee.org

Neural Networks : the Official Journal of the International Neural Network Society
|January 11, 2001
PubMed
Summary

This study introduces a new method combining iterative fast orthogonal search (IFOS) and minimum description length (MDL) for generalized single-layer networks (GSLNs). This approach efficiently builds sparse GSLNs by automatically selecting and optimizing basis functions.

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

On the use of separable Volterra networks to model discrete-time Volterra systems.

IEEE transactions on neural networks·2008
Same author

Parallel cascade identification and its application to protein family prediction.

Journal of biotechnology·2001
Same author

3D-1D threading methods for protein fold recognition.

Pharmacogenomics·2001
Same author

Automatic classification of protein sequences into structure/function groups via parallel cascade identification: a feasibility study.

Annals of biomedical engineering·2000
Same author

Two methods for identifying Wiener cascades having noninvertible static nonlinearities.

Annals of biomedical engineering·2000
Same author

Two methods for calculating the responses of photoreceptors to moving objects.

Annals of biomedical engineering·1998

Area of Science:

  • Machine Learning
  • Artificial Neural Networks
  • Nonlinear Function Approximation

Background:

  • Generalized Single-Layer Networks (GSLNs) offer flexibility for approximating nonlinear functions.
  • A key limitation of GSLNs is the potential need for numerous weights and basis functions, increasing complexity.
  • Efficient structure selection and parameter estimation are crucial for practical GSLN applications.

Purpose of the Study:

  • To develop an automated method for constructing sparse Generalized Single-Layer Networks (GSLNs).
  • To address the challenge of determining optimal network structure and parameters in GSLNs.
  • To enhance the efficiency and applicability of GSLN architectures.

Main Methods:

  • Coupling the Iterative Fast Orthogonal Search (IFOS) algorithm with the Minimum Description Length (MDL) criterion.

Related Experiment Videos

  • Implementing an algorithm, termed IFOS-MDL, for automatic network structure selection and parameter estimation.
  • Employing both network growth and pruning mechanisms to create sparse GSLNs.

    • Main Results:

    • The IFOS-MDL algorithm effectively performs automatic structure selection for GSLNs.
    • The method enables efficient parameter estimation within the GSLN architecture.
    • Sparse GSLNs are constructed from large candidate basis function spaces, reducing complexity.

    Conclusions:

    • The IFOS-MDL approach provides an efficient solution for building sparse GSLNs.
    • This method overcomes the limitation of excessive weights and basis functions in traditional GSLNs.
    • The automated structure selection and parameter estimation enhance the practical utility of GSLNs for nonlinear function approximation.