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

Fast training of recurrent networks based on the EM algorithm.

S Ma1, C Ji

  • 1Department of Electrical, Computer, and Systems Engineering, Rensselaer Polytechnic Institute, Troy, NY 12180, 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

Sleep disturbance and brain health in professional association footballers.

Frontiers in sports and active living·2026
Same author

[Survey on the current status of hospitalized heart failure patients in China, 2023-2024].

Zhonghua xin xue guan bing za zhi·2026
Same author

[Evaluation of changes in serum measles antibody levels in children with tumors before and after chemotherapy and the necessity of revaccination with measles-containing vaccines].

Zhonghua yi xue za zhi·2025
Same author

[Genetic factors for risk stratification in dilated cardiomyopathy].

Zhonghua xin xue guan bing za zhi·2025
Same author

Alcohol, Anti-HIV Drugs, and/or Hippuric Acid Deteriorate Cellular Stresses in Senescent Hepatocytes and Aging Murine Liver.

Journal of addiction & prevention·2025
Same author

Development of linear IgA bullous dermatosis in a patient during an exacerbation of psoriasis.

Annales de dermatologie et de venereologie·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 new probabilistic model and fast training algorithm were developed for recurrent neural networks. This method significantly reduces training time by simplifying network training to individual feedforward neurons.

Area of Science:

  • Artificial Intelligence
  • Machine Learning
  • Computational Neuroscience

Background:

  • Recurrent neural networks (RNNs) are powerful tools for sequential data but suffer from computationally intensive training.
  • Existing training algorithms for RNNs can be slow and inefficient, limiting their practical application.
  • Probabilistic modeling offers a framework to potentially improve the efficiency and understanding of neural network dynamics.

Purpose of the Study:

  • To develop a novel, efficient training algorithm for recurrent neural networks.
  • To establish a probabilistic model for understanding recurrent network behavior.
  • To accelerate the training process of complex recurrent network architectures.

Main Methods:

  • A probabilistic model was established for recurrent networks.

Related Experiment Videos

  • The expectation-maximization (EM) algorithm was applied to derive the training method.
  • Mean-field approximation was used to simplify the network into individual feedforward neurons.
  • A linear weighted regression algorithm was employed for training these neurons.

    • Main Results:

    • A new, fast training algorithm for recurrent networks was successfully derived.
    • The algorithm converts complex recurrent network training into training individual feedforward neurons.
    • Training time was improved by five to 15 times on benchmark problems.
    • The approach offers a significant speed-up compared to traditional methods.

    Conclusions:

    • The developed probabilistic model and EM-based algorithm provide a computationally efficient method for training recurrent neural networks.
    • This approach drastically reduces training time, making complex RNNs more accessible for various applications.
    • The simplification to feedforward neurons offers a novel perspective on recurrent network training and analysis.