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

A fast, compact approximation of the exponential function.

N N Schraudolph1

  • 1IDSIA, Corso Elvezia 36, CH-6900, Lugano, Switzerland. nic@idsia.ch

Neural Computation
|May 5, 1999
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

Processing images by semi-linear predictability minimization.

Network (Bristol, England)ยท1999
See all related articles

This study introduces a fast approximation for exponential functions in neural network simulations. By manipulating floating-point representations, this method significantly speeds up computation time compared to standard libraries.

Area of Science:

  • Computational neuroscience
  • Computer science

Background:

  • Neural network simulations require extensive computation of exponential functions.
  • Standard mathematical library exponentiation routines are often performance bottlenecks.

Purpose of the Study:

  • To develop a faster approximation for exponential functions in neural network simulations.
  • To reduce overall computation time by optimizing exponentiation routines.

Main Methods:

  • Approximating exponentiation by manipulating components of the IEEE-754 floating-point representation.
  • Modeling the exponential function using this novel floating-point manipulation technique.

Main Results:

  • The proposed method models the exponential function effectively.

Related Experiment Videos

  • Achieved significantly faster computation times compared to standard lookup table methods with linear interpolation.
  • Demonstrated a more compact implementation.

    • Conclusions:

    • Manipulating floating-point representations offers a highly efficient alternative for approximating exponential functions in neural networks.
    • This approach can lead to substantial performance improvements in computational neuroscience and machine learning.