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

Ultimate performance of QEM classifiers.

P Comon1, G Bienvenu

  • 1Thomson Sintra, Sophia-Antipolis.

IEEE Transactions on Neural Networks
|January 1, 1996
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

SISSY: An efficient and automatic algorithm for the analysis of EEG sources based on structured sparsity.

NeuroImage·2017
Same author

Immune Cells and Molecular Networks in Experimentally Induced Pulpitis.

Journal of dental research·2015
Same author

EEG extended source localization: tensor-based vs. conventional methods.

NeuroImage·2014
Same author

Dysregulation of energy homeostasis in mice overexpressing insulin-like growth factor-binding protein 6 in the brain.

Diabetologia·2005
Same author

Insulin-like growth factor binding protein-6 inhibits neuroblastoma cell proliferation and tumour development.

European journal of cancer (Oxford, England : 1990)·2002
Same author

Vascular endothelial growth factor in pleural fluid.

Chest·1999
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
See all related articles

Quadratic error minimization (QEM) in supervised learning classifiers can approximate Bayesian rules, even with varied losses. This method, shared by multilayer perceptrons (MLP), allows performance assessment using finite datasets and kernel density estimation links.

Area of Science:

  • Machine Learning
  • Statistical Learning Theory

Background:

  • Supervised learning classifiers commonly use quadratic error minimization (QEM).
  • QEM is theoretically better suited for nonlinear regression than classification tasks.
  • The effectiveness of QEM for classification, especially with non-uniform losses, warrants further investigation.

Purpose of the Study:

  • To demonstrate that quadratic error minimization (QEM) in supervised learning can yield Bayesian discriminating rules.
  • To show that this property holds even with non-uniform loss functions when desired responses are appropriately set.
  • To establish a connection between the performance of classifiers trained with QEM and kernel density estimators.

Main Methods:

  • Analyzing the mapping generated by quadratic error minimization (QEM).

Related Experiment Videos

  • Investigating the behavior of multilayer perceptrons (MLP) under QEM.
  • Establishing theoretical links between QEM performance and kernel estimators of density.

    • Main Results:

    • The mapping produced by QEM tends to approximate Bayesian discriminating rules, even under non-uniform losses.
    • This Bayesian property is observed in multilayer perceptrons (MLP).
    • Performance assessment on finite learning sets is possible by relating QEM to kernel density estimation.

    Conclusions:

    • Quadratic error minimization (QEM) is a viable method for building classifiers that approximate Bayesian decision rules.
    • Multilayer perceptrons (MLP) exhibit this desirable property when trained with QEM.
    • The study provides a theoretical framework for evaluating classifier performance using finite data and kernel density estimation.