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

Magnified gradient function with deterministic weight modification in adaptive learning.

Sin-Chun Ng1, Chi-Chung Cheung, Shu-Hung Leung

  • 1School of Science and Technology, The Open University of Hong Kong, Hong Kong, China. scng@ouhk.edu.hk

IEEE Transactions on Neural Networks
|November 30, 2004
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

A feasibility study on active sound reduction across an acoustic plenum window by cancelling source clusters on internal periphery of the window cavity.

The Journal of the Acoustical Society of America·2024
Same author

A BERT Framework to Sentiment Analysis of Tweets.

Sensors (Basel, Switzerland)·2023
Same author

EEG-based emotion recognition using hybrid CNN and LSTM classification.

Frontiers in computational neuroscience·2022
Same author

I-Vector-Based Patient Adaptation of Deep Neural Networks for Automatic Heartbeat Classification.

IEEE journal of biomedical and health informatics·2019
Same author

Towards End-to-End ECG Classification With Raw Signal Extraction and Deep Neural Networks.

IEEE journal of biomedical and health informatics·2018
Same author

Filtering electrocardiographic signals using an unbiased and normalized adaptive noise reduction system.

Medical engineering & physics·2008
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

This study introduces two novel neural network algorithms, magnified gradient function backpropagation (MGFPROP) and deterministic weight modification (DWM), to enhance learning speed and global convergence. The combined MDPROP algorithm demonstrates superior performance over standard backpropagation for various learning tasks.

Area of Science:

  • Artificial Intelligence
  • Machine Learning
  • Neural Networks

Background:

  • Standard backpropagation (BP) algorithms face limitations in convergence rate and global convergence capability.
  • Improving the efficiency of neural network training is crucial for complex learning problems.

Purpose of the Study:

  • To introduce and evaluate two novel approaches, MGFPROP and DWM, for accelerating BP convergence and enhancing global convergence.
  • To investigate the performance of an integrated algorithm, MDPROP, combining MGFPROP and DWM.

Main Methods:

  • Magnified Gradient Function Backpropagation (MGFPROP): Magnifies the gradient function of the activation function to increase convergence rate.
  • Deterministic Weight Modification (DWM): Deterministically modifies network weights to reduce system error in multilayered feedforward neural networks.

Related Experiment Videos

  • MDPROP: Integrates MGFPROP and DWM to leverage the benefits of both approaches.

    • Main Results:

    • Both MGFPROP and DWM individually show improved performance compared to standard BP and other modified BP algorithms.
    • The integrated MDPROP algorithm consistently outperforms MGFPROP, DWM, and standard BP in terms of convergence rate and global convergence capability.
    • Simulation results validate the effectiveness of the proposed algorithms across multiple learning problems.

    Conclusions:

    • MGFPROP and DWM offer significant improvements over standard backpropagation for neural network training.
    • The MDPROP algorithm represents a substantial advancement, providing enhanced convergence rate and global convergence capability.
    • The proposed methods are effective for a range of learning problems, offering a more efficient alternative to existing BP algorithms.