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

On the classification capability of a dynamic threshold neural network

C C Chiang1, H C Fu

  • 1Advanced Technology Center, Industrial Technology Research Institute, Chu-Tung, Taiwan, R.O.C.

International Journal of Neural Systems
|June 1, 1994
PubMed
Summary

This study introduces the Dynamic Threshold Neural Network (DTNN), a novel neural network architecture. DTNN demonstrates superior classification capabilities and faster learning compared to traditional sigmoidal neural networks.

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

Unequivocal Identification of Spin-Triplet and Spin-Singlet Superconductors with Upper Critical Field and Flux Quantization.

Physical review letters·2023
Same author

[Predicting response to non-small cell lung cancer immunotherapy using pre-treatment contrast-enhanced CT texture-based classification].

Zhonghua zhong liu za zhi [Chinese journal of oncology]·2021
Same author

Absence of Evidence of Electrical Switching of the Antiferromagnetic Néel Vector.

Physical review letters·2019
Same author

Geometric decoherence of valley excitons in monolayer transition metal dichachogenides.

Journal of physics. Condensed matter : an Institute of Physics journal·2017
Same author

Chiral topological excitons in the monolayer transition metal dichalcogenides.

Scientific reports·2017
Same author

Developmental and evolutionary novelty in the serrated teeth of theropod dinosaurs.

Scientific reports·2015

Area of Science:

  • Artificial Intelligence
  • Machine Learning
  • Computational Neuroscience

Background:

  • Conventional sigmoidal multilayer neural networks face limitations in classification capability.
  • Efficient learning algorithms are crucial for complex pattern recognition tasks.

Purpose of the Study:

  • To introduce and evaluate the Dynamic Threshold Neural Network (DTNN).
  • To demonstrate the theoretical and experimental superiority of DTNN over sigmoidal networks.
  • To develop a learning algorithm for DTNN.

Main Methods:

  • Theoretical analysis of parameter bounds for DTNN and sigmoidal networks.
  • Derivation of a backpropagation-like learning algorithm for DTNN.
  • Empirical evaluation using the Two-Spirals problem benchmark.

Related Experiment Videos

Main Results:

  • DTNN exhibits a more favorable upper bound on free parameters compared to sigmoidal networks.
  • The derived DTNN learning algorithm facilitates efficient training.
  • DTNN successfully learned the Two-Spirals problem in significantly fewer epochs than reported for sigmoidal networks.

Conclusions:

  • DTNN offers enhanced classification capability and learning efficiency.
  • The proposed DTNN architecture and learning algorithm represent a significant advancement in neural network design.
  • DTNN shows promise for tackling complex learning tasks where sigmoidal networks struggle.