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

Auto-association by multilayer perceptrons and singular value decomposition.

H Bourlard1, Y Kamp

  • 1Philips Research Laboratory, Brussels, Belgium.

Biological Cybernetics
|January 1, 1988
PubMed
Summary

For auto-association tasks, multilayer perceptrons do not require nonlinear hidden units. Optimal parameters for data compression and dimensionality reduction can be found using linear techniques like singular value decomposition.

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

Continuous speech recognition by connectionist statistical methods.

IEEE transactions on neural networks·1993
See all related articles

Area of Science:

  • Machine Learning
  • Information Processing
  • Artificial Neural Networks

Background:

  • Multilayer perceptrons (MLPs) in auto-association mode are explored for data compression and dimensionality reduction.
  • Traditional training methods often involve the error back-propagation algorithm.

Purpose of the Study:

  • To demonstrate that nonlinearities in hidden units are unnecessary for auto-association.
  • To present a linear technique for optimizing MLP parameters in auto-association.
  • To offer a clear interpretation of MLP parameters.

Main Methods:

  • Utilizing linear techniques based on singular value decomposition (SVD).
  • Employing low-rank matrix approximation methods.
  • Comparing the linear approach to the Karhunen-Loève transform (KLT).

Related Experiment Videos

Main Results:

  • Nonlinearities in hidden units are found to be redundant for auto-association tasks.
  • Optimal parameter values can be directly derived using linear methods.
  • The proposed linear approach offers an efficient alternative to error back-propagation.

Conclusions:

  • Linear techniques provide an effective and interpretable method for training MLPs in auto-association.
  • This approach simplifies the process of data compression and dimensionality reduction using MLPs.