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

A note on the low-dimensional display of multivariate data using neural networks

G Reibnegger1, G Werner-Felmayer, H Wachter

  • 1Institute for Medical Chemistry and Biochemistry, University of Innsbruck, Austria.

Journal of Molecular Graphics
|June 1, 1993
PubMed
Summary

A new neural network method for data visualization has limitations. It cannot uniquely reconstruct original data from a low-dimensional display due to linear dependencies, challenging previous claims.

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

Repetitive transcranial magnetic stimulation in the treatment of resistant depression: changes of specific neurotransmitter precursor amino acids.

Journal of neural transmission (Vienna, Austria : 1996)·2021
Same author

Worlds apart or two sides of the same coin? Attitudes, meanings, and motives of potential oocyte and sperm donors in Austria.

Journal of assisted reproduction and genetics·2020
Same author

Influence of acute normovolemic hemodilution on the pharmacokinetics of Cisatracurium Besylate.

Minerva anestesiologica·2013
Same author

Donor pretreatment with tetrahydrobiopterin saves pancreatic isografts from ischemia reperfusion injury in a mouse model.

American journal of transplantation : official journal of the American Society of Transplantation and the American Society of Transplant Surgeons·2010
Same author

Comment on 'Comparison between creatinine and cystatin C-based GFR equations in renal transplantation'.

Nephrology, dialysis, transplantation : official publication of the European Dialysis and Transplant Association - European Renal Association·2008
Same author

Substrate and cofactor requirements of indoleamine 2,3-dioxygenase in interferon-gamma-treated cells: utilization of oxygen rather than superoxide.

Current drug metabolism·2007

Area of Science:

  • Computational science
  • Data visualization
  • Machine learning

Background:

  • A novel neural network technique was proposed for low-dimensional display of multivariate data.
  • This method utilizes hidden neuron activity in feed-forward networks.
  • Previous claims suggested it could reconstruct input vectors from the display.

Purpose of the Study:

  • To evaluate the generalizability and reconstructive capabilities of the proposed neural network technique.
  • To investigate the validity of claims regarding unique data reconstruction from low-dimensional displays.
  • To identify limitations in the application of this neural network for multivariate data analysis.

Main Methods:

  • Analysis of a three-layer feed-forward neural network trained for low-dimensional data display.

Related Experiment Videos

  • Testing the network with previously unknown input vectors exhibiting linear relationships.
  • Comparing the projection of diverse input vectors onto the low-dimensional display.
  • Evaluating the reconstruction of multivariate vectors from display points.

    • Main Results:

    • The claim of unique reconstruction is unjustified in its general form.
    • Different input vectors with linear dependencies can map to the same low-dimensional point.
    • An infinite set of linearly dependent vectors projects onto a single display point.
    • Reconstruction from a display point leads to only one specific vector, not necessarily the original.

    Conclusions:

    • The neural network technique's ability to uniquely reconstruct multivariate data is limited.
    • Linear dependencies between input vector components can cause information loss in the display.
    • The method's utility for accurate data reconstruction, unlike principal components analysis or nonlinear mapping, is questionable under certain conditions.