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

An epsilon-margin nonlinear classifier based on fuzzy if-then rules.

Jacek M Leski1

  • 1Institute of Electronics, Silesian University of Technology, 44-101 Gliwice, Poland. jl@boss.iele.polsl.gliwice.pl

IEEE Transactions on Systems, Man, and Cybernetics. Part B, Cybernetics : a Publication of the IEEE Systems, Man, and Cybernetics Society
|September 17, 2004
PubMed
Summary

This study presents novel classifier design methods, enhancing linear classifiers with absolute loss for better error approximation and outlier robustness. It also introduces nonlinear classifiers using fuzzy if-then rules and ensemble averaging for improved generalization.

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

Detection of Atrial Fibrillation Episodes in Long-Term Heart Rhythm Signals Using a Support Vector Machine.

Sensors (Basel, Switzerland)·2020
Same author

Epsilon-insensitive fuzzy c-regression models: introduction to epsilon-insensitive fuzzy modeling.

IEEE transactions on systems, man, and cybernetics. Part B, Cybernetics : a publication of the IEEE Systems, Man, and Cybernetics Society·2004
Same author

Computationally effective algorithm for robust weighted averaging.

IEEE transactions on bio-medical engineering·2004
Same author

Robust weighted averaging.

IEEE transactions on bio-medical engineering·2002

Area of Science:

  • Machine Learning
  • Pattern Recognition
  • Data Mining

Background:

  • Classical Ho-Kashyap procedure is a foundational method for designing linear classifiers.
  • Minimizing misclassification error and ensuring robustness to outliers are key challenges in classifier design.
  • Generalization ability is crucial for a classifier's performance on unseen data.

Purpose of the Study:

  • To introduce novel classifier design methods based on a modified Ho-Kashyap procedure.
  • To develop a linear classifier using absolute loss for improved error approximation and outlier robustness.
  • To extend classifier design to nonlinear models using fuzzy if-then rules and ensemble averaging.

Main Methods:

  • Modification of the classical Ho-Kashyap procedure.

Related Experiment Videos

  • Design of linear classifiers using absolute loss instead of squared loss.
  • Minimization of the Vapnik-Chervonenkis dimension for generalization control.
  • Ensemble averaging of fuzzy if-then rule-based classifiers (Takagi-Sugeno-Kang form).
  • Parameter estimation using local (independent) and global (simultaneous) approaches.

    • Main Results:

    • The proposed absolute loss method provides a better approximation of misclassification error.
    • The new linear classifier design demonstrates enhanced robustness to outliers.
    • Easy control over generalization ability is achieved through Vapnik-Chervonenkis dimension minimization.
    • The nonlinear classifier extension using ensemble averaging is presented.
    • Demonstrated validity of the introduced methods through examples.

    Conclusions:

    • The modified Ho-Kashyap procedure offers effective methods for designing both linear and nonlinear classifiers.
    • The use of absolute loss and Vapnik-Chervonenkis dimension minimization enhances classifier performance and generalization.
    • Ensemble averaging of fuzzy rule-based classifiers provides a viable approach for nonlinear classification.