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

Bounding the effect of noise in multiobjective learning classifier systems.

Xavier Llorà1, David E Goldberg

  • 1Illinois Genetic Algorithms Laboratory, National Center for Supercomputing Applications, University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA. xllora@illigal.ge.uiuc.edu

Evolutionary Computation
|October 16, 2003
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

Optimized nanospherical layered alternating metal-dielectric probes for optical sensing.

Optics express·2010
Same author

Optimally designed nanolayered metal-dielectric particles as probes for massively multiplexed and ultrasensitive molecular assays.

Proceedings of the National Academy of Sciences of the United States of America·2010
Same author

Dependency structure matrix, genetic algorithms, and effective recombination.

Evolutionary computation·2009
Same author

Invariability of central metabolic flux distribution in Shewanella oneidensis MR-1 under environmental or genetic perturbations.

Biotechnology progress·2009
Same author

Metabolic flux analysis of Shewanella spp. reveals evolutionary robustness in central carbon metabolism.

Biotechnology and bioengineering·2008
Same author

The crowding approach to niching in genetic algorithms.

Evolutionary computation·2008

This study examines noisy data's effect on Pittsburgh-style learning classifier systems. Researchers developed a model to predict errors and identify overfitting, improving algorithm performance in noisy domains.

Area of Science:

  • Machine Learning
  • Artificial Intelligence
  • Data Science

Background:

  • Pittsburgh-style learning classifier systems are susceptible to performance degradation when trained on noisy datasets.
  • Understanding the impact of data noise is crucial for developing robust machine learning algorithms.
  • Multiobjective selection is a key component in the analyzed learning classifier systems.

Purpose of the Study:

  • To analyze the impact of noisy datasets on Pittsburgh-style learning classifier systems.
  • To characterize the behavior of multiobjective selection-based learning classifier systems in noisy domains.
  • To develop a theoretical model for predicting minimal achievable error in noisy environments.

Main Methods:

  • Development of a theoretical model to predict the minimal achievable error in noisy domains.

Related Experiment Videos

  • Application of multiobjective selection techniques within the learning classifier system.
  • Graphical representation of evolved hypotheses to bound system behavior.
  • Analysis of overfitting conditions in learned hypotheses.

    • Main Results:

    • A theoretical model was established for predicting minimal achievable error in noisy domains.
    • The study successfully bounded the behavior of the learning classifier system.
    • Overfitting conditions leading to poor generalization were identified.
    • The research provides insights into algorithm performance with imperfect data.

    Conclusions:

    • The developed theoretical model aids in understanding and predicting the performance of learning classifier systems in noisy data.
    • Multiobjective techniques combined with theoretical modeling effectively bound algorithm behavior and identify overfitting.
    • This research contributes to the development of more robust and generalizable machine learning models for real-world applications.