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

Training two-layered feedforward networks with variable projection method.

C T Kim1, J J Lee

  • 1Department of Computer Science and Electrical Engineering, Korea Advanced Institute of Science and Technology, Daejon, Korea. ctkim@odyssey.kaist.ac.kr

IEEE Transactions on Neural Networks
|February 14, 2008
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

Effects of multiagent chemotherapy and independent risk factors in the treatment of high-risk GTT -25 years experiences of KRI-TRD.

International journal of gynaecology and obstetrics: the official organ of the International Federation of Gynaecology and Obstetrics·2018
Same author

Continuous ohmic heating of commercially processed apple juice using five sequential electric fields results in rapid inactivation of Alicyclobacillus acidoterrestris spores.

International journal of food microbiology·2017
Same author

Destruction of Bacillus cereus spores in a thick soy bean paste (doenjang) by continuous ohmic heating with five sequential electrodes.

Letters in applied microbiology·2016
Same author

Inactivation of Bacillus cereus spores in a tsuyu sauce using continuous ohmic heating with five sequential elbow-type electrodes.

Journal of applied microbiology·2015
Same author

Clinical outcomes in Asian elderly kidney transplant recipients: a multicenter cohort study.

Transplantation proceedings·2015
Same author

Vertebrobasilar dolichoectasia as a risk factor for cerebral microbleeds.

European journal of neurology·2013
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 variable projection (VP) method enhances the Levenberg-Marquardt algorithm for training neural networks. This efficient approach offers faster training compared to existing methods like ELM.

Area of Science:

  • Computational mathematics
  • Artificial intelligence
  • Machine learning

Background:

  • Training two-layered feedforward neural networks often involves complex optimization.
  • Separable nonlinear least squares (SNLLS) problems are common in this training process.
  • Existing optimization algorithms can be computationally intensive.

Purpose of the Study:

  • To introduce an efficient optimization method for training neural networks.
  • To integrate the variable projection (VP) method with the Levenberg-Marquardt algorithm.
  • To demonstrate the computational advantages of the proposed approach.

Main Methods:

  • The variable projection (VP) method is applied to separable nonlinear least squares (SNLLS) problems.
  • The VP method is incorporated into the Levenberg-Marquardt optimization algorithm.

Related Experiment Videos

  • Jacobian computation is achieved through a modified backpropagation algorithm.

    • Main Results:

    • The proposed algorithm demonstrates significant efficiency gains.
    • The method provides a computationally effective way to train two-layered feedforward neural networks.
    • Performance is superior to conventional Levenberg-Marquardt algorithm (LMA), hybrid gradient algorithm (HGA), and extreme learning machine (ELM).

    Conclusions:

    • The integration of VP with Levenberg-Marquardt offers an efficient training strategy.
    • Modified backpropagation provides an effective way to compute Jacobians for VP networks.
    • This method presents a promising alternative for accelerating neural network training.