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

Confidence-based active learning.

Mingkun Li1, Ishwar K Sethi

  • 1DOE Joint Genome Institute, Lawrence Berkeley National Laboratory, 2800 Mitchell Drive, Walnut Creek, CA 94598, USA. mli@lbl.gov

IEEE Transactions on Pattern Analysis and Machine Intelligence
|August 5, 2006
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

Levetiracetam reverses temozolomide resistance in glioblastoma by blocking drug efflux through RAB5A/CD63-RAB35 axis.

Cancer letters·2026
Same author

Extensive individual and microorganism-specific circadian oscillations of the upper respiratory tract microbiome.

Cell reports·2026
Same author

Machine learning-based prediction of multi-level antimicrobial resistance in Klebsiella pneumoniae using whole-genome sequencing data.

International journal of antimicrobial agents·2026
Same author

Influence of the Fibrous Network Architecture on the Mechanical Properties of Melt-Blown Non-Woven Thermoplastic Polyurethane Fabrics.

Polymers·2026
Same author

Intratumoral <i>Parvimonas micra</i> promotes esophageal squamous cell carcinoma via <i>p</i>-cresol-induced T<sub>reg</sub> differentiation.

Science advances·2026
Same author

Machine-learning-decoupled absorption spectroscopy reveals bandgap shrinkage competition in Si-doped β-Ga<sub>2</sub>O<sub>3</sub>.

Optics letters·2026

This study introduces confidence-based active learning for classifier training, focusing on uncertain samples. This robust method effectively identifies and annotates crucial data points without extra computational cost.

Area of Science:

  • Machine Learning
  • Artificial Intelligence
  • Data Science

Background:

  • Classifier training often requires large labeled datasets, which are expensive and time-consuming to acquire.
  • Active learning aims to reduce labeling effort by strategically selecting informative samples for annotation.
  • Existing active learning methods may lack robustness or introduce computational overhead.

Purpose of the Study:

  • To propose a novel confidence-based active learning approach for efficient classifier training.
  • To develop a method for accurately estimating sample uncertainty using classifier output scores.
  • To demonstrate the effectiveness and robustness of the proposed approach compared to existing methods.

Main Methods:

  • Confidence-based active learning identifies uncertain samples by measuring their conditional error.

Related Experiment Videos

  • The approach calibrates classifier output scores to estimate sample uncertainty.
  • A dynamic bin width allocation method is used for accurate conditional error estimation, adapted to underlying probabilities.

    • Main Results:

    • The proposed confidence-based active learning approach demonstrates good performance in training various classifiers.
    • The method is robust and does not incur additional computational effort compared to existing techniques.
    • Experiments on synthetic and real datasets show competitive performance against the least certain active learning method.

    Conclusions:

    • Confidence-based active learning offers an effective and efficient strategy for reducing data labeling costs.
    • The proposed method's robustness and lack of computational overhead make it a practical choice for real-world applications.
    • The approach shows promise for improving the training of classifiers, particularly Support Vector Machines (SVMs).