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

Neural networks for estimating intrinsic dimension.

A Potapov1, M K Ali

  • 1Department of Physics, The University of Lethbridge, 4401 University Dr. W. Lethbridge, Alberta, Canada T1K 3M4. alexei.potapov@uleth.ca

Physical Review. E, Statistical, Nonlinear, and Soft Matter Physics
|May 15, 2002
PubMed
Summary

This study introduces a novel neural network approach for feature extraction and dimensionality reduction. The method uses topological mapping to preserve data relationships, offering a faster alternative to traditional autoassociative networks for time series analysis.

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 intact and fibrotic parenchyma in pancreatic ductal adenocarcinomas using compression optical coherence elastography.

Scientific reports·2026
Same author

Cell hiding in colorectal cancer: correlation with response to chemotherapy in vitro and in vivo.

Scientific reports·2024
Same author

COVID-19 and tobacco cessation: lessons from India.

Public health·2021
Same author

Effect of a multicomponent intervention on achievement and improvements in quality-of-care indices among people with Type 2 diabetes in South Asia: the CARRS trial.

Diabetic medicine : a journal of the British Diabetic Association·2019
Same author

Uterine exteriorization versus intraperitoneal repair: effect on intraoperative nausea and vomiting during repeat cesarean delivery - A randomized clinical trial.

Facts, views & vision in ObGyn·2019
Same author

Acceptability of a decision-support electronic health record system and its impact on diabetes care goals in South Asia: a mixed-methods evaluation of the CARRS trial.

Diabetic medicine : a journal of the British Diabetic Association·2018

Area of Science:

  • Computational neuroscience
  • Machine learning
  • Data science

Background:

  • Feature extraction and dimensionality reduction are crucial for analyzing complex observational data.
  • Autoassociative neural networks with bottleneck layers are commonly used but can be computationally intensive.
  • Determining the intrinsic dimensionality of data is essential for model simplification.

Purpose of the Study:

  • To propose a novel, efficient neural network-based method for feature extraction and intrinsic dimensionality determination.
  • To develop a technique that performs nonlinear, lower-dimensional data projection while preserving neighbor distances.
  • To demonstrate the application of this method for estimating the minimal model dimension from time series data.

Main Methods:

  • Utilizing a neural network to perform a topological mapping for data projection.

Related Experiment Videos

  • Implementing the topological mapping efficiently using radial basis function networks.
  • Comparing the proposed method's speed and effectiveness against autoassociative neural networks.

    • Main Results:

    • The proposed topological mapping technique effectively creates nonlinear, lower-dimensional data projections.
    • The method preserves the relative distances between neighboring data points.
    • Radial basis function network implementation offers significant speed advantages over autoassociative networks.
    • The technique successfully estimates the minimal mathematical model dimension from time series data.

    Conclusions:

    • The proposed topological mapping neural network provides an efficient and effective approach for feature extraction and dimensionality reduction.
    • This method is a viable and faster alternative to traditional autoassociative networks for analyzing observational and time series data.
    • The technique's ability to preserve neighbor distances and estimate intrinsic dimensionality makes it valuable for complex data analysis.