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

Optical finite impulse response neural networks.

Paulo E X Silveira1, G S Pati, Kelvin H Wagner

  • 1University of Colorado, Optoelectronic Computing Systems Center, Department of Electrical and Computer Engineering, Boulder 80309-0525, USA. paulo.silveira@networkphotonics.com

Applied Optics
|July 27, 2002
PubMed
Summary

This study details the finite impulse response neural network and introduces a novel delayed-feedback back-propagation algorithm. Optoelectronic processors using adaptive holograms enable efficient temporal processing for advanced neural network architectures.

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

Fourier-basis active structured illumination sensing lidar imaging using crossed acousto-optic devices.

Applied optics·2026
Same author

Time-shifted pseudo-noise multibeam lidar array using acousto-optic deflectors.

Applied optics·2025
Same author

Monopulse ladar: super-resolved 3D localization with Si-photonic serpentine optical phased arrays.

Applied optics·2022
Same author

Computational optical sensing and imaging 2021: feature issue introduction.

Optics express·2022
Same author

Computational Optical Sensing and Imaging 2021: introduction to the feature issue.

Applied optics·2022
Same author

Afocal catadioptric optical assembly for Fourier-sampling computational microscopy.

Applied optics·2021

Area of Science:

  • Neuroscience
  • Optical Engineering
  • Computer Science

Background:

  • Finite impulse response (FIR) neural networks offer unique computational properties.
  • Temporal back-propagation is crucial for training recurrent neural networks.
  • Optoelectronic processors can accelerate complex computations.

Purpose of the Study:

  • To provide a detailed description of FIR neural networks.
  • To introduce and analyze a novel delayed-feedback back-propagation algorithm.
  • To present and evaluate optoelectronic processor architectures for temporal processing.

Main Methods:

  • Detailed description of finite impulse response neural network.
  • Analysis of temporal back-propagation algorithms, including a new delayed-feedback variant.

Related Experiment Videos

  • Presentation and analysis of optoelectronic processors utilizing adaptive volume holograms and 3D optical processing.
  • Development of single-layer (input/output delay plane) and multi-layer architectures.

    • Main Results:

    • Two single-layer architectures (input delay plane, output delay plane) are presented.
    • Two multi-layer architectures are proposed, implementing both conventional and delayed-feedback back-propagation.
    • Demonstration of optoelectronic processors capable of implementing these architectures.

    Conclusions:

    • The proposed delayed-feedback back-propagation algorithm offers a novel approach to temporal processing.
    • Adaptive volume holograms and 3D optical processing are viable for implementing advanced neural network architectures.
    • The presented architectures facilitate efficient forward and backward propagation in FIR neural networks.