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

Decision feedback recurrent neural equalization with fast convergence rate.

Jongsoo Choi1, Martin Bouchard, Tet Hin Yeap

  • 1School of Information Technology and Engineering, University of Ottawa, Ottawa, ON K1N 6N5, Canada. jongsoo.choi@ieee.org

IEEE Transactions on Neural Networks
|June 9, 2005
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

Working in a relational way is everything: Perceptions of power and value in a drug policy-making network.

Health research policy and systems·2024
Same author

Structuring adaptations: Resilience, restrictive deterrence, and the Cunningham precursor control papers.

The International journal on drug policy·2024
Same author

Sociometric network analysis in illicit drugs research: A scoping review.

PloS one·2023
Same author

Using Topic Modeling Methods for Short-Text Data: A Comparative Analysis.

Frontiers in artificial intelligence·2021
Same author

Network Integration within a Prison-Based Therapeutic Community.

Social networks·2020
Same author

Leveraging the value of online data and methods: Drug policy research at the cutting edge.

The International journal on drug policy·2019

This study introduces Extended Kalman Filter (EKF) algorithms for training decision feedback recurrent neural equalizers (DFRNEs). These EKF-based methods offer faster convergence and better tracking than traditional real-time recurrent learning (RTRL) for complex-valued signal processing.

Area of Science:

  • Signal Processing
  • Machine Learning
  • Artificial Neural Networks

Background:

  • Real-time recurrent learning (RTRL) trains recurrent neural networks (RNNs) but suffers from slow convergence.
  • Decision feedback recurrent neural equalizers (DFRNEs) using RTRL require extensive training data for optimal performance.
  • This limitation hinders efficient training, especially in complex-valued signal processing applications.

Purpose of the Study:

  • To present Extended Kalman Filter (EKF) algorithms for training DFRNEs.
  • To adapt RTRL-based DFRNEs using state-space formulation for complex-valued signals.
  • To enhance convergence speed and tracking performance in DFRNEs.

Main Methods:

  • Developed state-space formulations for DFRNEs incorporating RTRL.

Related Experiment Videos

  • Introduced global and decoupled Extended Kalman Filter (EKF) algorithms.
  • Evaluated DFRNE performance through nonlinear channel equalization simulations.

    • Main Results:

    • EKF-based DFRNE algorithms demonstrated significantly faster convergence compared to RTRL.
    • The proposed EKF methods exhibited superior tracking performance.
    • Performance was validated in complex-valued signal processing and nonlinear channel equalization.

    Conclusions:

    • EKF algorithms provide an efficient alternative to RTRL for training DFRNEs.
    • The proposed methods overcome the slow convergence drawback of RTRL.
    • EKF-based DFRNEs are suitable for applications requiring fast adaptation and accurate equalization.