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

Boosted ARTMAP: modifications to fuzzy ARTMAP motivated by boosting theory.

Stephen J Verzi1, Gregory L Heileman, Michael Georgiopoulos

  • 1Computer Science Department University of New Mexico, Albuquerque, NM 87131, USA.

Neural Networks : the Official Journal of the International Neural Network Society
|December 14, 2005
PubMed
Summary

New Fuzzy ARTMAP neural network modifications improve classification in complex, noisy environments by reducing overfitting. Boosted ARTMAP and Structural Boosted ARTMAP enhance generalization performance through structural risk minimization techniques.

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

Computing with Spikes: The Advantage of Fine-Grained Timing.

Neural computation·2018
Same author

Potential Public Health Effects of Reducing Nicotine Levels in Cigarettes in the United States.

The New England journal of medicine·2018
Same author

Pareto-Optimal Model Selection via SPRINT-Race.

IEEE transactions on cybernetics·2017
Same author

Multi-Objective Model Selection via Racing.

IEEE transactions on cybernetics·2015
Same author

A Simple Method for Solving the SVM Regularization Path for Semidefinite Kernels.

IEEE transactions on neural networks and learning systems·2015
Same author

Modeling Evacuation of a Hospital without Electric Power.

Prehospital and disaster medicine·2015

Area of Science:

  • Artificial Intelligence
  • Machine Learning
  • Neural Networks

Background:

  • Fuzzy ARTMAP neural networks struggle with overfitting in complex, noisy data.
  • Structural risk minimization offers a framework to address overfitting by balancing training error and classifier complexity.

Purpose of the Study:

  • To propose modifications to Fuzzy ARTMAP for improved generalization in challenging environments.
  • To introduce novel architectures, Boosted ART and Boosted ARTMAP, leveraging structural risk minimization principles.

Main Methods:

  • Introducing Boosted ART for independent cluster spatial extent adjustment.
  • Developing Boosted ARTMAP to allow non-zero training error for reduced hypothesis complexity.
  • Proposing Structural Boosted ARTMAP for comparing online/offline and empirical/structural risk minimization.

Related Experiment Videos

Main Results:

  • Boosted ARTMAP generalizes better than Fuzzy ARTMAP by reducing overfitting.
  • Structural Boosted ARTMAP enables direct comparison of different learning paradigms.
  • Empirical and theoretical results support the enhanced understanding of these architectures.

Conclusions:

  • Modified Fuzzy ARTMAP architectures, Boosted ARTMAP and Structural Boosted ARTMAP, effectively improve generalization performance.
  • These novel approaches provide valuable insights into mitigating overfitting in complex classification tasks.
  • The proposed methods offer a robust alternative for noisy environments and advanced machine learning applications.