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

A fast U-D factorization-based learning algorithm with applications to nonlinear system modeling and identification.

Y Zhang1, X R Li

  • 1Department of Electrical and Computer Engineering, The University of Western Ontario, London, Ontario N6A 5B9, Canada.

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

First Detection of Ultrahigh Energy Emission from Gamma-Ray Binary LS I +61° 303.

Physical review letters·2026
Same author

Evidence of Cosmic-Ray Acceleration up to Sub-PeV Energies in the Supernova Remnant IC 443.

Physical review letters·2026
Same author

[Clinicopathologic features and treatment of 13 cases of Castleman's disease].

Zhonghua nei ke za zhi·2026
Same author

Precise Measurement of the Cosmic Ray Helium Spectrum above 0.1 PeV.

Physical review letters·2026
Same author

[Research progress on the mechanisms and optimization strategies of mesenchymal stem cell-derived exosomes in the treatment of liver diseases].

Zhonghua gan zang bing za zhi = Zhonghua ganzangbing zazhi = Chinese journal of hepatology·2026
Same author

All-Sky Search for Individual Primordial Black Hole Bursts with LHAASO.

Physical review letters·2025
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

A novel U-D factorization-based fading memory extended Kalman filter (FMEKF) algorithm enhances neural network training speed and accuracy. This method offers improved convergence, stability, and generalization compared to traditional backpropagation and EKF algorithms.

Area of Science:

  • Artificial Intelligence
  • Machine Learning
  • Neural Networks

Background:

  • Training multilayer feedforward neural networks (FNNs) can be computationally intensive.
  • Existing algorithms like backpropagation (BP) and standard Extended Kalman Filters (EKF) have limitations in convergence speed and accuracy.

Purpose of the Study:

  • To present a fast learning algorithm for FNN training using a fading memory extended Kalman filter (FMEKF).
  • To introduce a U-D factorization-based FMEKF for improved learning rate and accuracy.
  • To demonstrate the algorithm's superiority over BP and existing EKF methods.

Main Methods:

  • Development of a fading memory extended Kalman filter (FMEKF) with a self-adjusting time-varying forgetting factor.
  • Implementation of a U-D factorization-based FMEKF for enhanced numerical stability and efficiency.

Related Experiment Videos

  • Comparative analysis against backpropagation (BP) and other EKF-based learning algorithms.

    • Main Results:

    • The U-D factorization-based FMEKF algorithm achieved significantly more accurate learning results with fewer hidden nodes.
    • Demonstrated improved convergence rate, enhanced numerical stability, and robustness.
    • Showed reduced sensitivity to initial parameters and data randomness, along with good generalization ability.

    Conclusions:

    • The proposed U-D factorization-based FMEKF algorithm offers a more effective and efficient approach for training FNNs.
    • This method requires less training time to reach desired accuracy levels.
    • Simulation results confirm its effectiveness in modeling and identification of nonlinear dynamic systems.