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

Honest assessments of automatic learning algorithm performance.

M Revow1, D Maclean

  • 1Department of Computer Science, University of Toronto, Ontario, Canada.

Analytical and Quantitative Cytology and Histology
|May 9, 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

Identifying predictive markers for survival in malignant biliary obstruction following percutaneous transhepatic biliary drainage.

Clinical radiology·2025
Same author

Machine Learning to Predict Prostate Artery Embolization Outcomes.

Cardiovascular and interventional radiology·2024
Same author

A retrospective cohort study comparing a novel, spherical, resorbable particle against five established embolic agents for uterine fibroid embolisation.

Clinical radiology·2021
Same author

Factors Predicting a Good Symptomatic Outcome After Prostate Artery Embolisation (PAE).

Cardiovascular and interventional radiology·2018
Same author

The Treatment of Haemorrhoids.

British medical journal·2010
Same author

Diseases of the Chest in Children: Their Treatment by Blisters.

British medical journal·2010
Same journal

Micronuclei and other nuclear anomalies in buccal mucosa following exposure to X-ray radiation.

Analytical and quantitative cytology and histology·2012
Same journal

Color deconvolution. Optimizing handling of 3D unitary optical density vectors with polar coordinates.

Analytical and quantitative cytology and histology·2012
Same journal

Nuclear morphometry in the differential diagnosis of eosinophilic renal epithelial tumors.

Analytical and quantitative cytology and histology·2012
Same journal

Spectrum of prostatic nonepithelial tumors and tumorlike conditions with spindle cell features.

Analytical and quantitative cytology and histology·2012
Same journal

A framework for quantitative assessment of Ki67 distribution in preneoplastic bronchial epithelial lesions.

Analytical and quantitative cytology and histology·2012
Same journal

Classification and grading of noninvasive and invasive neoplasms of the urothelium.

Analytical and quantitative cytology and histology·2012
See all related articles

Evaluating machine learning algorithms requires careful consideration of data splitting methods. Standard cross-validation can lead to biased results, highlighting the need for robust assessment techniques in predictive modeling.

Area of Science:

  • Machine Learning
  • Computational Biology
  • Medical Image Analysis

Background:

  • Evaluating the performance of predictive models is crucial in machine learning.
  • Standard methods like cross-validation have limitations in assessing algorithm performance accurately.
  • The choice of data splitting significantly impacts the reliability of performance evaluations.

Purpose of the Study:

  • To compare different methodologies for evaluating probabilistic predictors in machine learning systems.
  • To assess the performance of automatic learning algorithms in classifying cervical cells from digital images.
  • To investigate the trade-offs between statistical rigor and data requirements in model evaluation.

Main Methods:

  • Four machine learning algorithms were evaluated using four distinct methodologies.

Related Experiment Videos

  • Methodologies included standard and modified cross-validation techniques.
  • Assessments focused on separating normal squamous intermediate cervical cells from other image objects.

    • Main Results:

    • Cross-validation, while data-efficient, can yield misleading performance assessments due to bias and variance.
    • A modified cross-validation approach offered more reliable assessments but could still be misleading.
    • Statistical rigor in evaluation is inversely related to the cost of data collection.

    Conclusions:

    • Users of machine learning algorithms must carefully select evaluation methods.
    • Judicious care is needed to avoid bias and large variance in performance assessments.
    • Reliable evaluation is essential for trustworthy machine learning applications.