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

An incremental network for on-line unsupervised classification and topology learning.

Shen Furao1, Osamu Hasegawa

  • 1Department of Computational Intelligence and Systems Science, Tokyo Institute of TechnologyR2-52, 4259 Nagatsuta, Midori-ku, Yokohama 226-8503, Japan. furaoshen@isl.titech.ac.jp

Neural Networks : the Official Journal of the International Neural Network Society
|September 13, 2005
PubMed
Summary

This study introduces an online unsupervised learning system that effectively handles noisy, unlabeled data. It dynamically adapts to changing data patterns and identifies clusters without needing prior knowledge of the data structure.

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

Differential diagnosis of retroperitoneal lymphangioma mimicking an ovarian tumor in the right adnexal region in a woman with a giant fibroid: a case report and literature review.

Fukushima journal of medical science·2026
Same author

Hemiparkinsonism Associated with Venous Anomaly-related Basal Ganglia Calcification.

Internal medicine (Tokyo, Japan)·2026
Same author

Beta-Catenin Positive Giant Cell Tumor of Bone with Pathological Fracture Treated with Neoadjuvant Denosumab Therapy for Visualizing Extraskeletal Lesions.

Case reports in oncology·2025
Same author

Cellular Angiofibroma of the Perineal Region With Tortuous Vessels: A Report of Three Cases.

Cureus·2025
Same author

Global and Regional Reduction of Myocardial Perfusion in Patients with Transthyretin Type of Cardiac Amyloidosis: A Dual SPECT Study Using <sup>99m</sup>Tc Pyrophosphate and <sup>201</sup>Thallium.

Annals of nuclear cardiology·2024
Same author

Case report: Atypical lipomatous tumor of the thigh in a four-year-old girl.

Frontiers in oncology·2024

Area of Science:

  • Machine Learning
  • Artificial Intelligence
  • Data Science

Background:

  • Unsupervised learning typically requires labeled data or predefined cluster numbers.
  • Handling noisy and non-stationary data distributions presents a significant challenge for existing algorithms.

Purpose of the Study:

  • To develop an online unsupervised learning mechanism capable of processing noisy, unlabeled data.
  • To enable incremental system growth and adaptation to non-stationary data distributions.
  • To automatically determine the optimal number of clusters and representative patterns.

Main Methods:

  • Utilizes a similarity threshold-based and local error-based insertion criterion for incremental learning.
  • Introduces an 'error-radius' parameter to learn the required number of nodes.

Related Experiment Videos

  • Employs a novel node removal technique in low probability density regions to eliminate noise and separate overlapping clusters.

    • Main Results:

    • The system successfully learns from unlabeled, noisy data in an online fashion.
    • It dynamically adjusts to non-stationary data distributions.
    • Accurately identifies the number of clusters and provides representative patterns without prior assumptions.

    Conclusions:

    • The proposed two-layer neural network architecture effectively represents topological structures of online data.
    • This method offers a robust solution for unsupervised learning with noisy and evolving data.
    • It eliminates the need for pre-specifying the number of clusters or initial codebooks.