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

Additive neural networks and periodic patterns.

Tomas Gedeon1

  • 1Department of Mathematical Sciences, Montana State University, Bozeman, MT, USA

Neural Networks : the Official Journal of the International Neural Network Society
|March 29, 2003
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

Cell geometry and membrane protein crowding constrain Escherichia coli growth rate, overflow metabolism, respiration, and maintenance energy.

FEBS letters·2026
Same author

Effect of transcriptional delay on ribosome abundance control.

Journal of mathematical biology·2026
Same author

Joint Realizability of Monotone Boolean Functions.

Theoretical computer science·2025
Same author

Oscillator death in coupled biochemical oscillators.

Mathematics of control, signals, and systems : MCSS·2025
Same author

Cellular Economics of Exchanged Metabolites Alter Ratios of Microbial Trading Partners in a Predictable Manner.

ACS synthetic biology·2025
Same author

Petri net on expanded networks computes the dynamics of finite multilevel models.

Interface focus·2025

This study introduces weight selection methods for additive neural networks, enabling them to learn periodic patterns. These networks dynamically adjust weights to ensure trajectories converge to predictable periodic orbits.

Area of Science:

  • Computational Neuroscience
  • Machine Learning

Background:

  • Additive neural networks are explored for their capacity to model complex data patterns.
  • Understanding the dynamics of neural networks in representing periodic data is crucial for advancing AI.

Purpose of the Study:

  • To investigate weight selection strategies for additive neural networks to effectively represent periodic patterns.
  • To analyze the convergence properties of neural network trajectories based on selected weights.

Main Methods:

  • A periodic set of vectors V(l) with components v(i)(l)=+/-1 was defined.
  • Correlation between vector components over time was measured to inform weight selection.
  • Two weight selection processes, discrete and continuous in time, were developed.

Related Experiment Videos

Main Results:

  • Additive neural networks with correlation-based weights demonstrate convergence to periodic orbits.
  • These orbits correspond to trajectories visiting orthants determined by the input vectors V(l).

Conclusions:

  • The proposed weight selection methods enable additive neural networks to accurately represent periodic patterns.
  • Dynamic weight selection processes offer a robust way to achieve desired network behaviors for periodic data.