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

The cascade-correlation learning: a projection pursuit learning perspective.

J N Hwang1, S S You, S R Lay

  • 1Dept. of Electr. Eng., Washington Univ., Seattle, WA.

IEEE Transactions on Neural Networks
|January 1, 1996
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

[Changes of menstruation patterns and adverse effects during the treatment of LNG-IUS for symptomatic adenomyosis].

Zhonghua fu chan ke za zhi·2016
Same author

[A prospective cohort study on the impact of placement timing of LNG-IUS for adenomyosis].

Zhonghua yi xue za zhi·2016
Same author

[A prospective study on the effects of levonorgestrel-releasing intrauterine system for adenomyosis with menorrhagia].

Zhonghua fu chan ke za zhi·2016
Same author

[A prospective cohort study on effects of levonorgestrel-releasing intrauterine system for adenomyosis with severe dysmenorrhea].

Zhonghua fu chan ke za zhi·2016
Same author

Functional relation between the uterus and the corpus luteum.

The Journal of endocrinology·2010
Same author

Gonad stimulation by androgens in hypophysectomized pigeons.

The Journal of endocrinology·2010
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

Cascade-correlation (Cascor) networks dynamically grow hidden neurons but may saturate, favoring classification over regression. Projection pursuit learning networks (PPLN) offer an alternative for approximating nonlinearity.

Area of Science:

  • Artificial Intelligence
  • Machine Learning
  • Neural Networks

Background:

  • Cascade-correlation (Cascor) is a supervised learning architecture that dynamically grows hidden neuron layers.
  • Projection Pursuit Learning Networks (PPLN) also dynamically grow hidden neurons, similar to Cascor.

Purpose of the Study:

  • To analyze the training criteria of Cascor and PPLN.
  • To compare the suitability of Cascor and PPLN for classification and regression tasks.

Main Methods:

  • Analysis of the maximum correlation training criterion in Cascor.
  • Comparison of Cascor's residual error approximation with PPLN's approach using trainable nonlinear activations.
  • Simulation results to evaluate network performance.

Related Experiment Videos

Main Results:

  • Cascor's maximum correlation training criterion tends to cause hidden units to saturate.
  • This saturation makes Cascor more suitable for classification than regression tasks.
  • The identified weakness in Cascor may also impact its performance on certain classification tasks.

Conclusions:

  • Cascor's architecture is better suited for classification tasks due to hidden unit saturation.
  • PPLN offers an alternative for approximating high-order nonlinearity, potentially overcoming Cascor's limitations.
  • Further investigation into Cascor's limitations on classification tasks is warranted.