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

Complexity issues in natural gradient descent method for training multilayer perceptrons

H H Yang1, S Amari

  • 1Oregon Graduate Institute, Computer Science Dept, Box 91000, Portland OR 97291, USA. hyang@cse.ogi.edu

Neural Computation
|November 6, 1998
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

Metropolitan air pollution and the upper and lower airway: chronic rhinosinusitis and asthma outpatient burden.

Rhinology·2026
Same author

Incorporation of transperineal ultrasonography and pelvic electromyography into the real clinical setting for the prediction of postpartum stress urinary incontinence.

Clinical radiology·2025
Same author

[Severity of SARS-CoV-2 infection in children with kidney disease undergoing immunosuppressive therapy].

Zhonghua er ke za zhi = Chinese journal of pediatrics·2025
Same author

[Progress in the application of exosomes in the diagnosis and treatments of pediatric kidney diseases].

Zhonghua er ke za zhi = Chinese journal of pediatrics·2025
Same author

[New biological and targeted therapies for the management of immune-related adverse events].

Zhonghua nei ke za zhi·2022
Same author

[Tocilizumab therapy for immune checkpoint inhibitor associated myocarditis: a case report].

Zhonghua xin xue guan bing za zhi·2022
Same journal

A Model-Free Reinforcement Learning Implementation of Decision Making Under Uncertainty by Sequential Sampling.

Neural computation·2026
Same journal

DROP: Distributional and Regular Optimism and Pessimism for Reinforcement Learning.

Neural computation·2026
Same journal

Hierarchical Active Inference Using Successor Representations.

Neural computation·2026
Same journal

W-Kernel and Its Principal Space for Frequentist Evaluation of Bayesian Estimators.

Neural computation·2026
Same journal

A Hidden Markov Model-Inspired Sequence Classification Method for Hyperdimensional Computing.

Neural computation·2026
Same journal

Sparse Graphical Modeling for Electrophysiological Phase-Based Connectivity Using Circular Statistics.

Neural computation·2026
See all related articles

Natural gradient descent trains n-m-1 perceptrons efficiently. A new algorithm computes the natural gradient for stochastic multilayer perceptrons without matrix inversion, achieving O(n) complexity when input dimension exceeds hidden neurons.

Area of Science:

  • Machine Learning
  • Artificial Intelligence
  • Neural Networks

Background:

  • The natural gradient descent method is a powerful optimization technique for training neural networks.
  • Efficient computation of the Fisher information matrix is crucial for natural gradient methods.
  • Existing methods for calculating natural gradients can be computationally intensive, especially for large input dimensions.

Purpose of the Study:

  • To develop a novel algorithm for computing the natural gradient in n-m-1 multilayer perceptrons.
  • To propose an efficient scheme for representing the Fisher information matrix in stochastic multilayer perceptrons.
  • To reduce the computational complexity of natural gradient calculation for high-dimensional inputs.

Main Methods:

  • Application of the natural gradient descent method to train an n-m-1 multilayer perceptron.

Related Experiment Videos

  • Development of an efficient scheme for representing the Fisher information matrix.
  • Algorithm design for calculating the natural gradient without explicit Fisher information matrix inversion.

    • Main Results:

    • A new algorithm for natural gradient calculation in n-m-1 stochastic multilayer perceptrons is proposed.
    • The algorithm avoids explicit inversion of the Fisher information matrix.
    • When the input dimension (n) is much larger than the number of hidden neurons (m), the time complexity for computing the natural gradient is O(n).

    Conclusions:

    • The proposed algorithm offers an efficient approach to natural gradient computation for specific neural network architectures.
    • This method is particularly beneficial for models with high-dimensional input spaces.
    • The reduced time complexity enhances the practicality of natural gradient descent in large-scale machine learning applications.