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

Two regularizers for recursive least squared algorithms in feedforward multilayered neural networks.

C S Leung1, A C Tsoi, L W Chan

  • 1Department of Electronic Engineering, City University of Hong Kong, Hong Kong.

IEEE Transactions on Neural Networks
|February 6, 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

The incidence of arthrogenic muscle inhibition in acute knee injuries: a prospective cohort study in a tertiary acute knee unit.

Asia-Pacific journal of sports medicine, arthroscopy, rehabilitation and technology·2025
Same author

Time Pattern of Presentation of Victims of High-Speed Passenger Ferry Mass Casualty Incidents to the Emergency Department.

Disaster medicine and public health preparedness·2024
Same author

Small beams, fast predictions: a comparison of machine learning dose prediction models for proton minibeam therapy.

Medical physics·2022
Same author

Prenatal diagnosis of pathogenic genomic imbalance in fetuses with increased nuchal translucency but normal karyotyping using chromosomal microarray.

Hong Kong medical journal = Xianggang yi xue za zhi·2019
Same author

FIR and IIR Synapses, a New Neural Network Architecture for Time Series Modeling.

Neural computation·2019
Same author

Ambulance use affects timely emergency treatment of acute ischaemic stroke.

Hong Kong medical journal = Xianggang yi xue za zhi·2018
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

New Recursive Least Squares (RLS) algorithms enhance feedforward multilayered neural network (FMNN) training. These methods maintain consistent weight decay, improving network generalization for better performance.

Area of Science:

  • Computational Neuroscience
  • Machine Learning
  • Artificial Neural Networks

Background:

  • Recursive Least Squares (RLS) algorithms are efficient for online training of feedforward multilayered neural networks (FMNNs).
  • The standard RLS algorithm exhibits diminishing weight decay effectiveness over training epochs, potentially hindering optimal network generalization.

Purpose of the Study:

  • To address the diminishing weight decay issue in standard RLS algorithms for FMNN training.
  • To develop modified RLS algorithms that provide consistent weight decay and improve network generalization capabilities.

Main Methods:

  • Derivation of the True Weight Decay RLS (TWDRLS) algorithm, employing a modified energy function for constant weight decay.
  • Development of the Input Perturbation RLS (IPRLS) algorithm, focusing on prediction robustness to input perturbations.

Related Experiment Videos

  • Comparative simulations to evaluate the performance and generalization of the proposed algorithms against standard RLS.

    • Main Results:

    • Both TWDRLS and IPRLS algorithms demonstrated improved generalization capabilities in trained FMNNs.
    • The TWDRLS algorithm ensures a constant weight decay effect, irrespective of the number of training epochs.
    • The IPRLS algorithm enhances prediction performance robustness against input perturbations.

    Conclusions:

    • Modified RLS algorithms (TWDRLS and IPRLS) effectively overcome the limitations of standard RLS regarding weight decay.
    • These novel algorithms offer significant improvements in the generalization ability of feedforward multilayered neural networks.
    • The proposed methods provide valuable tools for enhancing the training and performance of neural networks in various applications.