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

Improving generalization performance of natural gradient learning using optimized regularization by NIC.

Hyeyoung Park1, Noboru Murata, Shun-Ichi Amari

  • 1Brain Science Institute, RIKEN, Saitama, Japan. hypark@brain.riken.go.jp

Neural Computation
|March 10, 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

Amantadine: a first option for symptomatic relief in young-onset Parkinson's disease?

Parkinsonism & related disorders·2026
Same author

First Korean Case of <i>C9orf72</i>-Related Amyotrophic Lateral Sclerosis With Progressive Supranuclear Palsy-Like Features.

Journal of clinical neurology (Seoul, Korea)·2026
Same author

SHANK2 establishes auditory hair bundle architecture essential for mammalian hearing.

Proceedings of the National Academy of Sciences of the United States of America·2025
Same author

University of Massachusetts Amherst's nursing PhD program Equity-centered community design of a new curriculum to transform the discipline, achieve health justice, and sustain planetary health.

Nursing outlook·2025
Same author

Sensorial flavor characteristics and volatile Maillard reaction products of <i>Tenebrio molitor</i> mealworm based-reaction flavors with beef broth-like flavor.

Food science and biotechnology·2025
Same author

Association between Loneliness and Memory Function through White Matter Hyperintensities in Older Adults: The Moderating Role of Gender.

Behavioral sciences (Basel, Switzerland)·2023
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 learning accelerates neural network training but risks overfitting. Introducing regularization optimized by the Network Information Criterion (NIC) improves generalization performance and computational efficiency in real-world applications.

Area of Science:

  • Machine Learning
  • Neural Networks
  • Optimization

Background:

  • Natural gradient learning accelerates training by escaping plateaus, a common cause of slow neural network learning.
  • Adaptive natural gradient methods have demonstrated advantages in practical, real-world problem-solving.
  • However, rapid parameter fitting in natural gradient learning can lead to overfitting and poor generalization.

Purpose of the Study:

  • To investigate and improve the generalization performance of natural gradient learning.
  • To address the overfitting issue inherent in rapid parameter fitting during natural gradient learning.
  • To propose an efficient method for optimizing regularization in natural gradient learning.

Main Methods:

  • Introduced a regularization term into the natural gradient learning process.

Related Experiment Videos

  • Proposed an optimization method for the regularization scale using the generalized Akaike information criterion (Network Information Criterion - NIC).
  • Conducted theoretical analysis and computer simulations to study the properties of NIC-optimized regularization strength.

    • Main Results:

    • The proposed method effectively balances fitting training data and preventing overfitting.
    • Optimized regularization strength via NIC demonstrated beneficial properties.
    • Computational experiments on benchmark problems confirmed the method's efficiency and generalization capabilities.

    Conclusions:

    • The integration of NIC-based regularization significantly enhances the generalization performance of natural gradient learning.
    • The proposed method offers a computationally efficient solution for improving neural network generalization.
    • This approach is validated for practical applications, showing improved performance on benchmark datasets.