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

Multiobjective genetic optimization of diagnostic classifiers with implications for generating receiver operating

M A Kupinski1, M A Anastasio

  • 1Kurt Rossmann Laboratories, Department of Radiology, The University of Chicago, IL 60637, USA.

IEEE Transactions on Medical Imaging
|October 26, 1999
PubMed
Summary

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

ADAPTIVE SMALL-ANIMAL SPECT/CT.

IEEE Nuclear Science Symposium conference record. Nuclear Science Symposium·2015
Same author

Three-Dimensional Optoacoustic and Laser-Induced Ultrasound Tomography System for Preclinical Research in Mice: Design and Phantom Validation.

Ultrasonic imaging·2015
Same author

Characterization of speckle in lung images acquired with a benchtop in-line x-ray phase-contrast system.

Physics in medicine and biology·2013
Same author

An initial evaluation of analyser-based phase-contrast X-ray imaging of carotid plaque microstructure.

The British journal of radiology·2012
Same author

Full- and minimal-scan reconstruction algorithms for fan-beam diffraction tomography.

Applied optics·2008
Same author

A new reconstruction approach for reflection mode diffraction tomography.

IEEE transactions on image processing : a publication of the IEEE Signal Processing Society·2008

This study introduces a niched Pareto multiobjective genetic algorithm (GA) for optimizing binary classifiers. This approach bypasses traditional objective function aggregation, offering a superior set of optimal solutions for sensitivity and specificity.

Area of Science:

  • Machine Learning
  • Computational Biology
  • Bioinformatics

Background:

  • Binary classifiers possess inherent performance metrics: sensitivity and specificity.
  • Conventional training methods aggregate these metrics into a single objective, requiring prior information.
  • This aggregation can limit the classifier's overall performance and adaptability.

Purpose of the Study:

  • To investigate the application of a niched Pareto multiobjective genetic algorithm (GA) for optimizing binary classifiers.
  • To eliminate the need for aggregating objective functions during classifier training.
  • To leverage post-optimization selection for incorporating prior knowledge.

Main Methods:

  • Utilized a niched Pareto multiobjective genetic algorithm (GA) to optimize an objective vector instead of a scalar function.

Related Experiment Videos

  • Applied the technique to train both linear classifiers and artificial neural networks (ANNs).
  • Used simulated datasets for training and evaluation.

    • Main Results:

    • The niched Pareto GA generated a set of optimal solutions representing various sensitivity-specificity trade-offs.
    • These solutions correspond to operating points on a receiver operating characteristic (ROC) curve.
    • The generated ROC curves were found to be superior or equal to those from conventional methods.

    Conclusions:

    • Multiobjective optimization using niched Pareto GA offers a more effective approach to binary classifier training.
    • This method avoids premature objective aggregation, providing a richer set of optimal solutions.
    • The technique enables flexible selection of classifier operating points post-optimization based on specific needs.