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

Back-propagation learning of infinite-dimensional dynamical systems.

Isao Tokuda1, Ryuji Tokunaga, Kazuyuki Aihara

  • 1Department of Computer Science and Systems Engineering, Muroran Institute of Technology, Muroran, 050-0071 Hokkaido, Japan. tokuda@csse.muroran-it.ac.jp

Neural Networks : the Official Journal of the International Neural Network Society
|September 19, 2003
PubMed
Summary

This study explores back-propagation learning for delayed recurrent neural networks (DRNNs), finding DRNNs effective for spatio-temporal dynamics despite infinite-dimensional challenges. Comparisons reveal DRNNs

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

Energy landscapes and synergetic state transitions in frustrated Stuart-Landau oscillator networks: a homotopy continuation study.

Frontiers in network physiology·2026
Same author

CauFinder: Steering Cell-State and Phenotype Transitions by Causal Disentanglement Learning.

Advanced science (Weinheim, Baden-Wurttemberg, Germany)·2026
Same author

Identifying the optimal rapid antigen test for screening and determining the end of isolation: A modeling study.

PLoS computational biology·2026
Same author

Force Learning in Balanced Cortical E-I Networks.

Neural computation·2026
Same author

Psychological distress among Japanese high school students during the COVID-19 pandemic: An energy landscape analysis.

PLoS medicine·2026
Same author

Stratification of viral shedding patterns in saliva of COVID-19 patients.

eLife·2026

Area of Science:

  • Computational Neuroscience
  • Machine Learning

Background:

  • Recurrent neural networks (RNNs) are powerful for sequential data.
  • Ordinary RNNs (ORNNs) lack mechanisms to handle time-delayed feedback.
  • Delayed Recurrent Neural Networks (DRNNs) incorporate time delays, leading to infinite-dimensional dynamics.

Purpose of the Study:

  • To investigate the application of back-propagation learning to DRNNs.
  • To develop and compare learning algorithms for DRNNs.
  • To evaluate the learning capability and robustness of DRNNs against noise.

Main Methods:

  • Development of two distinct back-propagation learning algorithms tailored for DRNNs.
  • Numerical simulations using chaotic signals from the Mackey-Glass and Rössler equations.
  • Comparative analysis of DRNNs against ORNNs and time-delay neural networks.

Related Experiment Videos

Main Results:

  • DRNNs demonstrate capability in learning spatio-temporal dynamics.
  • The infinite-dimensional nature of DRNNs influences learning algorithms and capabilities.
  • Performance, robustness against noise, and limitations of DRNNs were elucidated through comparative studies.

Conclusions:

  • DRNNs offer a viable approach for modeling systems with time-delayed feedback.
  • Specific learning algorithms are crucial for effective DRNN training.
  • Understanding the trade-offs between DRNNs, ORNNs, and other delay networks is essential for application selection.