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

Posterior probability support vector machines for unbalanced data.

Qing Tao1, Gao-Wei Wu, Fei-Yue Wang

  • 1Key Laboratory of Complex Systems and Intelligence Science, Institute of Automation, Chinese Academy of Sciences, Beijing 100080, P.R. China. qing.tao@mail.ia.ac.cn

IEEE Transactions on Neural Networks
|December 14, 2005
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

Embodied cognition-driven interpretable trajectory prediction of autonomous systems.

Nature communications·2026
Same author

SR-LLM: An incremental symbolic regression framework driven by LLM-based retrieval-augmented generation.

Proceedings of the National Academy of Sciences of the United States of America·2025
Same author

Generative AI-Driven Ergonomics: A Virtual-Real Hybrid Experiment for Human Factors Engineering.

IEEE transactions on cybernetics·2025
Same author

A Visual Benchmark for Autonomous Driving in Open-Pit Mines.

IEEE transactions on pattern analysis and machine intelligence·2025
Same author

A NIR dual-channel fluorescent probe for detecting viscosity and ONOO<sup>-</sup> in vitro and vivo.

Spectrochimica acta. Part A, Molecular and biomolecular spectroscopy·2025
Same author

Scene as Occupancy and Reconstruction: A Comprehensive Dataset for Unstructured Scene Understanding.

Scientific data·2025

This study introduces posterior probability support vector machines (PPSVMs) for weighted data, offering improved classification accuracy closer to Bayes optimal. PPSVMs provide a natural extension of SVMs, outperforming standard SVMs in certain scenarios.

Area of Science:

  • Machine Learning
  • Statistical Learning Theory
  • Pattern Recognition

Background:

  • Support Vector Machines (SVMs) are powerful classification tools, but standard formulations may not optimally handle weighted or unbalanced datasets.
  • Existing methods like fuzzy SVMs offer extensions, but a more statistically grounded approach is desirable.

Purpose of the Study:

  • To propose a novel framework for posterior probability support vector machines (PPSVMs) to address weighted training samples and unbalanced classification problems.
  • To introduce a new optimization problem and concept of support vectors within the PPSVM framework.
  • To develop a soft PPSVM with an interpretable parameter and an empirical method for posterior probability estimation.

Main Methods:

  • Modified concepts of risks, linear separability, margin, and optimal hyperplane were used to formulate the PPSVM framework.

Related Experiment Videos

  • A new optimization problem was defined for unbalanced classification.
  • A soft PPSVM with parameter 'v' was derived, analogous to v-SVM, with an empirical method for determining 'v' via posterior probability estimation.

    • Main Results:

    • The PPSVM framework was validated using synthetic and real-world classification examples.
    • PPSVMs demonstrated logical correctness and relationships to regular SVMs and Bayesian methods.
    • Empirical results showed PPSVMs outperforming regular SVMs in specific cases and offering a natural extension of SVMs based on statistical learning theory.

    Conclusions:

    • The proposed PPSVM framework offers a statistically sound and analytically natural extension of regular SVMs.
    • PPSVMs achieve performance closer to the Bayes optimal classifier without requiring prior distribution knowledge.
    • The method provides an interpretable parameter 'v' and an effective approach for handling weighted and unbalanced data in classification tasks.