Jove
Visualize
Contact Us

Related Experiment Videos

Training Dynamics and Neural Network Performance.

Omid M. Omidvar1, James L. Blue, Charles L. Wilson

  • 1University of the District of Columbia, USA

Neural Networks : the Official Journal of the International Neural Network Society
|July 1, 1997
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 journal

Q-learning based asynchronous Boolean control networks stabilization with data loss.

Neural networks : the official journal of the International Neural Network Society·2026
Same journal

New results on prescribed-time synchronization of complex networks via intermittent control.

Neural networks : the official journal of the International Neural Network Society·2026
Same journal

Variance-constrained multi-view ensemble broad network for imbalanced data.

Neural networks : the official journal of the International Neural Network Society·2026
Same journal

Dynamic analysis and reliable mechanical optimization application of ring HNN effected with a memristive neuron.

Neural networks : the official journal of the International Neural Network Society·2026
Same journal

DAFF-Net: A detection and search method for small-scale low surface brightness galaxies.

Neural networks : the official journal of the International Neural Network Society·2026
Same journal

Quasi-synchronization for complex networks with hybrid pinning intermittent control.

Neural networks : the official journal of the International Neural Network Society·2026
See all related articles
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

This study introduces novel training methods for multilayer perceptrons (MLPs) that significantly improve classification performance. These optimized MLPs achieve results comparable to or better than probabilistic neural networks (PNNs), with reduced network size.

Area of Science:

  • Machine Learning
  • Artificial Neural Networks
  • Pattern Recognition

Background:

  • Multilayer perceptrons (MLPs) are widely used for classification tasks.
  • Existing MLP training methods can be suboptimal, leading to performance limitations.
  • Probabilistic neural networks (PNNs) have shown superior performance in specific classification problems.

Purpose of the Study:

  • To investigate the training dynamics of feedforward MLPs using insights from recurrent network models.
  • To develop modified training strategies for MLPs to enhance classification performance.
  • To achieve performance comparable to or exceeding PNNs using a standard MLP architecture.

Main Methods:

  • Analysis of recurrent network dynamics to inform MLP training.

Related Experiment Videos

  • Implementation of modified neuron activation functions to reduce singular Jacobians.
  • Application of successive regularization and Boltzmann pruning for weight space constraint.
  • Integration of prior class probabilities for normalized error calculations.

    • Main Results:

    • Modified MLPs demonstrated improved error-reject performance by factors of 2-4 on digit and fingerprint classification.
    • Network size was reduced by 40-60% through the proposed optimization techniques.
    • Performance equal to or better than PNNs was achieved with a single three-layer MLP.

    Conclusions:

    • Fundamental changes in MLP optimization strategy can significantly enhance performance.
    • The proposed methods simplify training dynamics and improve classification accuracy.
    • Optimized MLPs offer a competitive alternative to PNNs with greater efficiency.