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

Efficiently learning multilayer perceptrons.

C Bunzmann1, M Biehl, R Urbanczik

  • 1Institut für Theoretische Physik, Universität Würzburg, Am Hubland, D-97074 Würzburg, Germany.

Physical Review Letters
|April 6, 2001
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

IRMA: Machine learning-based harmonization of <math><mmultiscripts><mrow></mrow> <mrow></mrow> <mn>18</mn></mmultiscripts></math> F-FDG PET brain scans in multi-center studies.

European journal of nuclear medicine and molecular imaging·2025
Same author

Investigating the aspect of asymmetry in brain-first versus body-first Parkinson's disease.

NPJ Parkinson's disease·2024
Same author

[Bone Marrow Lesion - Risk-Factors and Pathophysiological Relevance: Systematic Review].

Zeitschrift fur Orthopadie und Unfallchirurgie·2016
Same author

Expression of chemokines CXCL4 and CXCL7 by synovial macrophages defines an early stage of rheumatoid arthritis.

Annals of the rheumatic diseases·2015
Same author

Assessment of acrosome state in boar spermatozoa heads using n-contours descriptor and RLVQ.

Computer methods and programs in biomedicine·2013
Same author

Simulation of self-assembled nanopatterns in strained 2D alloys on the face centered cubic (111) surface.

Journal of physics. Condensed matter : an Institute of Physics journal·2011
Same journal

Erratum: Bacterial Turbulence at Compressible Fluid Interfaces [Phys. Rev. Lett. 136, 138301 (2026)].

Physical review letters·2026
Same journal

Unveiling Light-Quark Yukawa Flavor Structure via Dihadron Fragmentation at Lepton Colliders.

Physical review letters·2026
Same journal

Adaptable Route to Fast Coherent State Transport via Bang-Bang-Bang Protocols.

Physical review letters·2026
Same journal

Topological Transition and Emergence of Elasticity of Dislocation in Skyrmion Lattice: Beyond Kittel's Magnetic-Polar Analogy.

Physical review letters·2026
Same journal

Pound-Drever-Hall Method for Superconducting-Qubit Readout.

Physical review letters·2026
Same journal

Coupling a ^{73}Ge Nuclear Spin to an Electrostatically Defined Quantum Dot in Silicon.

Physical review letters·2026
See all related articles

This study introduces a novel learning algorithm for multilayer perceptrons using principal component analysis. This method requires significantly fewer training examples for effective generalization in large networks compared to traditional approaches.

Area of Science:

  • Machine Learning
  • Artificial Intelligence
  • Computational Neuroscience

Background:

  • Multilayer perceptrons (MLPs) are fundamental neural network architectures.
  • Traditional on-line learning algorithms for MLPs can require extensive training data for generalization.
  • Efficient learning algorithms are crucial for advancing artificial intelligence and neural network applications.

Purpose of the Study:

  • To present a new learning algorithm for multilayer perceptrons.
  • To improve the data efficiency and generalization capabilities of MLPs, particularly for large networks.
  • To leverage principal component analysis for enhanced neural network training.

Main Methods:

  • The proposed algorithm utilizes principal component analysis (PCA).

Related Experiment Videos

  • A correlation matrix is computed from example inputs and their target outputs.
  • PCA is applied to identify key components for efficient learning.

    • Main Results:

    • The new algorithm demonstrates superior data efficiency compared to traditional on-line methods.
    • Fewer training examples are needed to achieve good generalization, especially in large MLP networks.
    • The principal component-based approach offers a more effective learning strategy.

    Conclusions:

    • The principal component-based learning algorithm provides a significant advancement for multilayer perceptrons.
    • This method offers a more efficient and effective way to train large neural networks.
    • The findings have implications for developing more scalable and data-efficient AI systems.