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

Dynamical features simulated by recurrent neural networks.

F Botelho1

  • 1Department of Mathematical Sciences, University of Memphis, Memphis, TN, USA

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

The study reveals that two-dimensional neural networks with specific structures behave like simple interval maps. Their dynamics range from predictable convergence to complex chaotic behavior with long cycles.

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

The oncological impact of complete transurethral resection before neoadjuvant chemotherapy in muscle-invasive bladder cancer.

Actas urologicas espanolas·2025
Same author

The influence of different metabolic syndrome definitions in predicting vasculogenic erectile dysfunction: is there a role for the index of central obesity?

The aging male : the official journal of the International Society for the Study of the Aging Male·2013
Same author

Endothelial function in patients with metabolic syndrome and erectile dysfunction: a question of angiopoietin imbalance?

Andrology·2013
Same author

Model-based patterns in prostate cancer mortality worldwide.

British journal of cancer·2013
Same author

Erectile dysfunction in obstructive sleep apnea syndrome--prevalence and determinants.

Revista portuguesa de pneumologia·2012
Same author

Predictive factors of a successful testicular biopsy and subsequent clinical pregnancy.

Andrologia·2012

Area of Science:

  • Computational neuroscience
  • Dynamical systems theory
  • Artificial neural networks

Background:

  • Two-dimensional neural networks with rank one connecting matrices and saturated linear transfer functions are a simplified yet relevant model.
  • Understanding the iterative behavior of such networks is crucial for analyzing their computational capabilities.

Purpose of the Study:

  • To establish a dynamic equivalence between these neural network models and piecewise linear maps on an interval.
  • To characterize the range of iterative behaviors exhibited by these neural networks.

Main Methods:

  • Establishing a formal dynamic equivalence between the specified neural network architecture and piecewise linear maps.
  • Analyzing the iterative dynamics of these maps to identify different behavioral regimes.

Related Experiment Videos

Main Results:

  • Demonstrated that the evolution of these neural networks is dynamically equivalent to piecewise linear maps on an interval.
  • Characterized the iterative behavior, showing a spectrum from predictable fixed-point convergence to chaotic dynamics.
  • Identified the existence of chaotic regions with cycles of arbitrarily large periods.

Conclusions:

  • The simplified neural network models exhibit complex dynamical behaviors previously observed in interval maps.
  • This equivalence provides a powerful framework for studying the stability and complexity of neural network dynamics.
  • The findings highlight the potential for rich, emergent behavior even in constrained neural network architectures.