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

Sequential RAM-based neural networks: learnability, generalisation, knowledge extraction, and grammatical inference.

M C de Souto1, P J Adeodato, T B Ludermir

  • 1Dept. of Electrical Eng.-Imperial College, London, UK. M.DESOUTO@IC.AC.UK

International Journal of Neural Systems
|November 24, 1999
PubMed
Summary

This study analyzes the learnability of sequential Random Access Machine (RAM)-based neural networks using Automata Theory. It determines conditions under which these networks can learn and generalize, aiding in knowledge extraction and integration.

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

Hybrid Training Method for MLP: Optimization of Architecture and Training.

IEEE transactions on systems, man, and cybernetics. Part B, Cybernetics : a publication of the IEEE Systems, Man, and Cybernetics Societyยท2011
See all related articles

Area of Science:

  • Artificial Intelligence
  • Computational Neuroscience
  • Machine Learning Theory

Background:

  • Understanding the computational capabilities (computability) and learning abilities (learnability) of artificial neural networks is crucial.
  • Sequential Random Access Machine (RAM)-based neural networks represent a specific class of models with unique learning characteristics.
  • Existing learning rules for these networks require rigorous analysis to define their scope and limitations.

Purpose of the Study:

  • To address the learnability problem for sequential RAM-based neural networks.
  • To establish the conditions and problem classes for which these networks can effectively learn and generalize.
  • To develop methods for knowledge extraction from and insertion into these networks.

Main Methods:

Related Experiment Videos

  • Application of Automata Theory as the primary analytical framework.
  • Formal analysis of the learning rules associated with sequential RAM-based neural networks.
  • Investigation of the temporal behavior and generalization capabilities.

    • Main Results:

    • Identification of specific problem classes that sequential RAM-based neural networks can learn under given training procedures.
    • Characterization of the conditions necessary for successful learning and generalization.
    • Development of novel techniques for knowledge manipulation within these networks.

    Conclusions:

    • The study provides a theoretical foundation for understanding the learnability of sequential RAM-based neural networks.
    • The findings enhance comprehension of the temporal dynamics and generalization capacity of these models.
    • Insights gained can inform the integration of symbolic and connectionist artificial intelligence paradigms.