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

SVD-NET: an algorithm that automatically selects network structure.

D C Psichogios1, L H Ungar

  • 1Dept. of Chem. Eng., Pennsylvania Univ., Philadelphia, PA.

IEEE Transactions on Neural Networks
|January 1, 1994
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

Identification of potential CSF biomarkers in ALS.

Neurology·2006
Same author

Chloroplast transit peptide prediction: a peek inside the black box.

Nucleic acids research·2001
Same author

Significance of conductances in Hodgkin-Huxley models.

Journal of neurophysiology·1993
Same author

Using radial basis functions to approximate a function and its error bounds.

IEEE transactions on neural networks·1992
Same journal

Universal perceptron and DNA-like learning algorithm for binary neural networks: LSBF and PBF implementations.

IEEE transactions on neural networks·2013
Same journal

Guest editorial: special section on white box nonlinear prediction models.

IEEE transactions on neural networks·2011
Same journal

Data-based fault-tolerant control of high-speed trains with traction/braking notch nonlinearities and actuator failures.

IEEE transactions on neural networks·2011
Same journal

Guest editorial: special section on data-based control, modeling, and optimization.

IEEE transactions on neural networks·2011
Same journal

Neural network-based multiple robot simultaneous localization and mapping.

IEEE transactions on neural networks·2011
Same journal

Data-driven model-free adaptive control for a class of MIMO nonlinear discrete-time systems.

IEEE transactions on neural networks·2011
See all related articles

A new algorithm uses singular value decomposition (SVD) to simplify feedforward neural networks by removing redundant nodes. This leads to smaller, more accurate models that generalize better, reducing the need for cross-validation.

Area of Science:

  • Artificial Intelligence
  • Machine Learning
  • Chemical Engineering

Background:

  • Feedforward neural networks (FNNs) are powerful tools for modeling complex systems.
  • Overfitting is a common challenge in FNN training, often addressed by cross-validation.
  • Redundant nodes in FNNs can lead to larger models and reduced generalization.

Purpose of the Study:

  • To develop an algorithm for training FNNs that minimizes network complexity.
  • To improve model generalization and reduce overfitting by eliminating redundant nodes.
  • To demonstrate the efficacy of the proposed method in a practical chemical engineering application.

Main Methods:

  • An algorithm utilizing singular value decomposition (SVD) was developed.
  • The algorithm identifies and removes redundant hidden nodes within the FNN architecture.

Related Experiment Videos

  • The method was applied to model a chemical reactor system.

    • Main Results:

    • The SVD-based algorithm successfully reduced the number of hidden nodes in the FNN.
    • The resulting smaller networks exhibited improved generalization capabilities.
    • The need for cross-validation to prevent overfitting was significantly reduced.

    Conclusions:

    • The developed SVD algorithm offers an effective approach to creating smaller, more generalizable FNNs.
    • Eliminating redundant nodes is a viable strategy for enhancing model performance and reducing computational cost.
    • The method shows promise for applications in chemical process modeling and other complex systems.