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

Recent advances in the MOBJ algorithm for training artificial neural networks.

R D Teixeira1, A P Braga, R H Takahashi

  • 1Department of Electronics Engineering, Federal University of Minas Gerais, Belo Horizonte, Minas Gerais, Brazil. roselito@cpdee.ufmg.br

International Journal of Neural Systems
|September 28, 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

Thy28 protects against anti-CD3-mediated thymic cell death in vivo.

Apoptosis : an international journal on programmed cell death·2014
Same author

Shifts in plant respiration and carbon use efficiency at a large-scale drought experiment in the eastern Amazon.

The New phytologist·2010
Same author

Ultrastructural study of spermatozoa of the neotropical lizards, Tropidurus semitaeniatus and Tropidurus torquatus (Squamata, Tropiduridae).

Tissue & cell·2008
Same author

A method for extracting plant roots from soil which facilitates rapid sample processing without compromising measurement accuracy.

The New phytologist·2007
Same author

Comparative study of sperm ultrastructure of five species of teiid lizards (Teiidae, Squamata), and Cercosaura ocellata (Gymnophthalmidae, Squamata).

Tissue & cell·2007
Same author

Pbhyd1 and Pbhyd2: two mycelium-specific hydrophobin genes from the dimorphic fungus Paracoccidioides brasiliensis.

Fungal genetics and biology : FG & B·2004

A new multi-objective optimization method trains MLPs by balancing training error and network weight norms. This approach yields superior generalization performance in classification and regression tasks compared to traditional methods.

Area of Science:

  • Machine Learning
  • Neural Networks
  • Optimization

Background:

  • Multi-objective optimization is crucial for balancing competing goals in machine learning.
  • Training Multilayer Perceptrons (MLPs) often involves trade-offs between accuracy and model complexity.
  • Existing methods like Weight Decay (WD), Support Vector Machines (SVMs), and Backpropagation (BP) have limitations.

Purpose of the Study:

  • To introduce a novel training scheme for MLPs using a relaxation method for multi-objective optimization.
  • To develop an algorithm that balances training error and the norm of network weight vectors.
  • To evaluate the proposed method's effectiveness in classification and regression problems.

Main Methods:

  • A relaxation method is employed for multi-objective optimization.

Related Experiment Videos

  • The algorithm identifies a reduced set of solutions.
  • The solution with the best generalization capability is selected from the reduced set.
  • The method balances training error against the network weight vector norm.

    • Main Results:

    • The proposed training scheme results in neural models with good generalization.
    • The method demonstrates superior performance compared to traditional techniques like WD, SVMs, and BP.
    • Effective application to both classification and regression tasks was shown.

    Conclusions:

    • The presented systematic training procedure offers a robust approach for developing well-generalizing neural models.
    • The multi-objective optimization strategy effectively balances key training objectives.
    • The novel method provides a competitive alternative to established machine learning algorithms.