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

Online learning with ensembles.

R Urbanczik1

  • 1Neural Computing Research Group, Aston University, Aston Triangle, Birmingham B4 7ET, United Kingdom.

Physical Review. E, Statistical Physics, Plasmas, Fluids, and Related Interdisciplinary Topics
|November 23, 2000
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

Laboratory tests focusing on sputum.

The international journal of tuberculosis and lung disease : the official journal of the International Union against Tuberculosis and Lung Disease·2010
Same author

Scaling up tuberculosis culture services: a precautionary note.

The international journal of tuberculosis and lung disease : the official journal of the International Union against Tuberculosis and Lung Disease·2009
Same author

Enumerating constrained elementary flux vectors of metabolic networks.

IET systems biology·2007
Same author

Yield of serial sputum specimen examinations in the diagnosis of pulmonary tuberculosis: a systematic review.

The international journal of tuberculosis and lung disease : the official journal of the International Union against Tuberculosis and Lung Disease·2007
Same author

Functional stoichiometric analysis of metabolic networks.

Bioinformatics (Oxford, England)·2005
Same author

An improved algorithm for stoichiometric network analysis: theory and applications.

Bioinformatics (Oxford, England)·2004
Same journal

Efficient Monte Carlo simulations using a shuffled nested Weyl sequence random number generator.

Physical review. E, Statistical physics, plasmas, fluids, and related interdisciplinary topics·2002
Same journal

Spatiotemporal dynamics of electromagnetic pulses in saturating nonlinear optical media with normal group velocity dispersion.

Physical review. E, Statistical physics, plasmas, fluids, and related interdisciplinary topics·2002
Same journal

Soliton-breather reaction pathways.

Physical review. E, Statistical physics, plasmas, fluids, and related interdisciplinary topics·2002
Same journal

Calculation of electromagnetic properties of regular and random arrays of metallic and dielectric cylinders.

Physical review. E, Statistical physics, plasmas, fluids, and related interdisciplinary topics·2002
Same journal

Electromagnetic convective cells in a nonuniform dusty plasma.

Physical review. E, Statistical physics, plasmas, fluids, and related interdisciplinary topics·2002
Same journal

Stability of neural networks and solitons of field theory.

Physical review. E, Statistical physics, plasmas, fluids, and related interdisciplinary topics·2002
See all related articles

Ensembles in supervised online learning, like those using the perceptron rule, offer generalization improvements similar to Gibbs learning. However, for optimized rules, ensembles provide no benefit due to a transform that equates single-student and ensemble performance.

Area of Science:

  • Machine Learning
  • Artificial Intelligence
  • Statistical Learning Theory

Background:

  • Supervised online learning algorithms update models iteratively using a stream of data.
  • Ensemble methods combine multiple models to improve predictive performance and robustness.
  • Generalization error measures how well a model performs on unseen data.

Purpose of the Study:

  • To analyze the generalization performance of ensembles in supervised online learning with randomized initial conditions.
  • To compare the effectiveness of ensembles with single models across different learning rules.
  • To understand the theoretical underpinnings of ensemble behavior in online learning settings.

Main Methods:

  • Analysis of supervised online learning algorithms with randomized initial conditions.

Related Experiment Videos

  • Mathematical derivation and comparison of generalization errors for ensemble and single-student models.
  • Investigation of the perceptron learning rule and more optimized learning rules.

    • Main Results:

    • Ensembles with randomized initial conditions show similar generalization error improvements as Gibbs learning for the perceptron rule.
    • For more optimized learning rules, ensembles do not yield significant improvements over single students.
    • A theoretical transformation (f) was identified, demonstrating that a single student using (f) can achieve the same generalization as an ensemble of (f) students.

    Conclusions:

    • The benefit of using ensembles in supervised online learning is dependent on the specific learning rule employed.
    • Randomization of initial conditions in ensembles provides benefits primarily for simpler learning rules like the perceptron.
    • Theoretical analysis reveals that for any learning rule, a single optimized student can match ensemble performance, questioning the universal utility of ensembles.