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

Implementing online natural gradient learning: problems and solutions.

Weishui Wan1

  • 1CED System Corporation, Jimbocho, Tokyo 101-0051, Japan. weisui@excite.co.jp

IEEE Transactions on Neural Networks
|March 29, 2006
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 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

This study introduces a novel online natural gradient learning algorithm, outperforming existing methods by addressing implementation challenges and improving learning efficiency. The research found that using online training error to adjust learning rates is ineffective.

Area of Science:

  • Machine Learning
  • Optimization Algorithms
  • Computational Statistics

Background:

  • Standard gradient descent methods suffer from slow learning speeds and poor performance.
  • Online natural gradient learning offers efficiency but faces implementation challenges.
  • Existing online learning algorithms include Almeida-Langlois-Amaral-Plakhov (ALAP), Vario-eta, and local adaptive methods.

Purpose of the Study:

  • To propose a new algorithm for online natural gradient learning that overcomes existing implementation problems.
  • To empirically compare the performance of the new algorithm against established online learning methods.
  • To analyze the strengths and weaknesses of various online learning algorithms.

Main Methods:

  • Development of a novel algorithm for online natural gradient learning.

Related Experiment Videos

  • Empirical evaluation using benchmark datasets: Proben1 and normalized handwritten digits (US Postal Service data).
  • Comparative analysis with algorithms such as ALAP, Vario-eta, local adaptive learning rate, and learning with momentum.

    • Main Results:

    • The proposed algorithm demonstrates superior performance compared to existing online learning algorithms.
    • Using online training error as a criterion for learning rate adjustment was found to be inappropriate.
    • Empirical analysis revealed distinct strengths and weaknesses of the compared algorithms.

    Conclusions:

    • The newly proposed online natural gradient learning algorithm offers improved performance.
    • The study highlights the limitations of using online training error for adaptive learning rate control.
    • The findings provide valuable insights into the practical implementation and comparative effectiveness of online learning strategies.