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

Analog neural nets with gaussian or other common noise distribution cannot recognize arbitrary regular languages.

W Maass1, E D Sontag

  • 1Institute for Theoretical Computer Science, Technische Universität Graz, Klosterwiesgasse 32/2, A-8010, Graz, Austria, maass@igi. tu-graz.ac.at

Neural Computation
|March 23, 1999
PubMed
Summary

Many regular languages are not recognizable by noisy recurrent analog neural networks. This study precisely characterizes recognizable languages, highlighting limitations for robust network construction against analog noise.

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

Monotone bifurcation graphs.

Journal of biological dynamics·2012
Same author

Influence of pinholes on MgO-tunnel junction barrier parameters obtained from current-voltage characteristics.

Journal of nanoscience and nanotechnology·2010
Same author

Remarks on feedforward circuits, adaptation, and pulse memory.

IET systems biology·2009
Same author

Proximity of intracellular regulatory networks to monotone systems.

IET systems biology·2008
Same author

Some new directions in control theory inspired by systems biology.

Systems biology·2006
Same author

Methods of robustness analysis for Boolean models of gene control networks.

Systems biology·2006

Area of Science:

  • Computational neuroscience
  • Theoretical computer science
  • Machine learning theory

Background:

  • Recurrent analog neural networks are susceptible to noise in their gate outputs.
  • Understanding the impact of noise on neural network computability is crucial for practical applications.
  • Previous research has explored noise robustness in digital neural networks.

Purpose of the Study:

  • To investigate the limitations of recurrent analog neural networks in recognizing regular languages under noisy conditions.
  • To precisely characterize the class of languages that can be recognized by such networks.
  • To explore methods for constructing noise-robust analog neural networks.

Main Methods:

  • Mathematical analysis of recurrent analog neural network models with Gaussian and other noise distributions.

Related Experiment Videos

  • Formal language theory to define and analyze language recognizability.
  • Development of theoretical frameworks to characterize noise-robust language recognition.

    • Main Results:

    • Demonstration that many regular languages cannot be recognized by recurrent analog neural networks with significant noise.
    • Precise characterization of the set of languages that are recognizable by these noisy networks.
    • Identification of severe constraints on building robust recurrent analog neural networks.

    Conclusions:

    • The presence of realistic analog noise imposes fundamental limitations on the capabilities of recurrent analog neural networks.
    • The findings necessitate a re-evaluation of architectures for noise-robust computation.
    • A method for constructing noise-robust feedforward analog neural networks is presented as an alternative.