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

Embedding recurrent neural networks into predator-prey models.

Yves Moreau1, Stephane Louiès, Joos Vandewalle

  • 1Katholieke Universiteit Leuven, Department of Electrical Engineering, Kardinaal Mercierlaan 94, B-3001, Leuven, Belgium

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

Researchers transformed continuous-time recurrent neural networks into Lotka-Volterra systems. This allows applying predator-prey model analysis techniques to neural networks, demonstrating Lotka-Volterra systems

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

Client Participation per Round in Federated Learning for Multiple Sclerosis with Real-World Data.

Studies in health technology and informatics·2026
Same author

Good for All, Not Good Enough for One: Reuse Dilemma in Federated Learning.

Studies in health technology and informatics·2026
Same author

A Comparative Study of QSPR Methods on a Unique Multitask PAMPA Data Set.

Journal of chemical information and modeling·2026
Same author

WiNGS-API: a federated genome/phenome data sharing platform enabling gene discovery and variant classification for rare diseases.

Genome medicine·2026
Same author

JINet: easy and secure private data analysis for everyone.

BMC bioinformatics·2025
Same author

MINDDS-connect: a federated data platform integrating biobanks for meta cohort building and analysis.

European journal of human genetics : EJHG·2025

Area of Science:

  • Dynamical Systems
  • Computational Neuroscience
  • Theoretical Neuroscience

Background:

  • Continuous-time recurrent neural networks (RNNs) are powerful models for sequential data.
  • Analyzing the complex dynamics of RNNs can be challenging.
  • Predator-prey models, or Lotka-Volterra systems, offer a well-understood framework for studying dynamical systems.

Purpose of the Study:

  • To develop a method for embedding RNN dynamics into Lotka-Volterra systems.
  • To enable the application of established Lotka-Volterra analysis techniques to RNNs.
  • To explore the universal approximation capabilities of Lotka-Volterra systems for dynamical systems.

Main Methods:

  • Transformation of RNN ordinary differential equations (ODEs) into a quasi-monomial form.

Related Experiment Videos

  • Further transformation of the quasi-monomial form into Lotka-Volterra equations.
  • Utilizing activation functions like hyperbolic tangent or logistic sigmoid for transformation.

    • Main Results:

    • Successfully embedded RNN dynamics into higher-dimensional Lotka-Volterra systems.
    • Demonstrated that the dynamics of the first variables in the Lotka-Volterra system are equivalent to the original RNN.
    • Showcased that Lotka-Volterra systems can serve as universal approximators for dynamical systems.

    Conclusions:

    • The transformation provides a novel approach to analyze RNNs using Lotka-Volterra system theory.
    • This work bridges the gap between neural network dynamics and classical ecological modeling.
    • Lotka-Volterra systems exhibit universal approximation properties akin to continuous-time neural networks.