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

VLSI implementation of neural networks.

B M Wilamowski, J Binfet, M O Kaynak

    International Journal of Neural Systems
    |September 30, 2000
    PubMed
    Summary

    Neural network hardware implementations offer superior control surfaces with fewer errors than fuzzy controllers. Despite challenges with CMOS activation functions and weight quantization, neural networks show promising results for VLSI applications.

    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

    Analog implementation of pulse-coupled neural networks.

    IEEE transactions on neural networksยท2008
    See all related articles

    Area of Science:

    • Electrical Engineering
    • Computer Science
    • Artificial Intelligence

    Background:

    • Fuzzy controllers are widely used for hardware implementation due to their design simplicity.
    • Neural controllers offer superior performance with reduced error but present implementation challenges.

    Purpose of the Study:

    • To address the challenges of implementing neural networks on VLSI chips.
    • To compare the performance of neural network hardware implementations against fuzzy controllers.

    Main Methods:

    • Developed approximation functions to address CMOS neural network activation function differences.
    • Trained neural networks using these approximation functions.
    • Investigated the impact of weight quantization on neural network performance in VLSI.

    Main Results:

    • Approximation functions successfully resolved activation function issues, leading to low-trained network errors.
    • Weight quantization significantly increased errors by an order of magnitude.
    • Despite quantization effects, neural network hardware implementations outperformed fuzzy systems.

    Conclusions:

    • Neural network hardware implementations are viable for complex control surfaces, offering better performance than fuzzy controllers.
    • Further research is needed to mitigate quantization effects in VLSI neural network design.

    Related Experiment Videos