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

Efficient greedy learning of gaussian mixture models.

J J Verbeek1, N Vlassis, B Kröse

  • 1Informatics Institute, University of Amsterdam, 1098 SJ Amsterdam, The Netherlands. jverbeek@science.uva.nl

Neural Computation
|February 20, 2003
PubMed
Summary

This study introduces a greedy learning algorithm for Gaussian mixtures, efficiently finding optimal components without sensitivity to initial values. It

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

Efficient source adaptivity in independent component analysis.

IEEE transactions on neural networks·2008
Same author

Fetal loss following ultrasound diagnosis of a live fetus at 6-10 weeks of gestation.

Ultrasound in obstetrics & gynecology : the official journal of the International Society of Ultrasound in Obstetrics and Gynecology·2003
Same author

Down syndrome in a population of elderly mentally retarded patients: genetic-diagnostic survey and implications for medical care.

American journal of medical genetics·1999
Same author

Visual texture perception. Features and spatial filters.

Nature·1988

Area of Science:

  • Machine Learning
  • Statistical Modeling
  • Artificial Intelligence

Background:

  • Gaussian mixture models are widely used for density estimation and clustering.
  • Traditional methods like Expectation-Maximization (EM) can be sensitive to initialization.
  • Determining the optimal number of components is often challenging.

Purpose of the Study:

  • To develop a novel greedy learning algorithm for Gaussian mixtures.
  • To address the initialization sensitivity of existing methods.
  • To provide an efficient algorithm for scenarios where the number of components is unknown.

Main Methods:

  • A heuristic approach for selecting optimal mixture components.
  • Randomized generation of candidate components.
  • Iterative insertion of locally optimal components into the mixture.

Main Results:

  • The proposed greedy algorithm demonstrates robustness to initialization.
  • Achieves linear time complexity with respect to data points.
  • Exhibits quadratic time complexity concerning the number of mixture components.
  • Outperforms existing methods in density estimation and texture segmentation tasks.

Conclusions:

  • The greedy learning approach offers a viable alternative to EM for Gaussian mixtures.
  • The algorithm is particularly advantageous when the number of components is not predefined.
  • Experimental validation confirms its effectiveness and efficiency.

Related Experiment Videos