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

An analytical framework for local feedforward networks.

S Weaver1, L Baird, M Polycarpou

  • 1Department of Electrical and Computer Engineering, University of Cincinnati, Cincinnati, OH 45221-0030, USA.

IEEE Transactions on Neural Networks
|February 7, 2008
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

Achievement of Target Gain Larger than Unity in an Inertial Fusion Experiment.

Physical review letters·2024
Same author

Lawson Criterion for Ignition Exceeded in an Inertial Fusion Experiment.

Physical review letters·2022
Same author

Unchecked immunity: a unique case of sequential immune-related adverse events with Pembrolizumab.

Journal for immunotherapy of cancer·2019
Same author

The Use of Nisin as a Preservative in Crumpets.

Journal of food protection·2019
Same author

Assembly of high-areal-density deuterium-tritium fuel from indirectly driven cryogenic implosions.

Physical review letters·2012
Same author

Antibodies to metabotropic glutamate receptor 5 in the Ophelia syndrome.

Neurology·2011
Same journal

Universal perceptron and DNA-like learning algorithm for binary neural networks: LSBF and PBF implementations.

IEEE transactions on neural networks·2013
Same journal

Guest editorial: special section on white box nonlinear prediction models.

IEEE transactions on neural networks·2011
Same journal

Data-based fault-tolerant control of high-speed trains with traction/braking notch nonlinearities and actuator failures.

IEEE transactions on neural networks·2011
Same journal

Guest editorial: special section on data-based control, modeling, and optimization.

IEEE transactions on neural networks·2011
Same journal

Neural network-based multiple robot simultaneous localization and mapping.

IEEE transactions on neural networks·2011
Same journal

Data-driven model-free adaptive control for a class of MIMO nonlinear discrete-time systems.

IEEE transactions on neural networks·2011
See all related articles

Neural networks can suffer from interference, where learning new information disrupts old knowledge. This study introduces a framework to measure interference and network localization, showing single-hidden-layer networks can be highly local.

Area of Science:

  • Computational Neuroscience
  • Machine Learning
  • Artificial Intelligence

Background:

  • Interference in neural networks is a critical issue where learning in one input space region degrades performance in another.
  • Spatially local networks exhibit reduced susceptibility to such interference, making them more robust for complex learning tasks.

Purpose of the Study:

  • To develop a theoretical framework for quantifying interference and network localization in neural networks.
  • To analyze the properties of sigmoidal multilayer perceptron (MLP) networks using the backpropagation algorithm.

Main Methods:

  • Development of a novel theoretical framework incorporating measures for interference and network localization.
  • Analysis of network weights, architecture, and learning algorithms, specifically backpropagation with a quadratic cost function.

Related Experiment Videos

  • Examination of single-hidden-layer sigmoidal MLP networks.

    • Main Results:

    • The developed framework allows for the precise measurement of interference and localization properties.
    • Demonstrated that single-hidden-layer sigmoidal MLPs are not inherently nonlocal as commonly believed.
    • Proved that with a sufficient number of adjustable weights, these networks can achieve arbitrary levels of localization while maintaining universal approximation capabilities.

    Conclusions:

    • The theoretical framework provides new insights into the interference properties of neural networks.
    • Single-hidden-layer sigmoidal MLPs can be designed to be highly local, challenging existing misconceptions.
    • These findings have implications for designing more stable and efficient neural network architectures.