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

Architectures for high-performance FPGA implementations of neural models.

Randall K Weinstein1, Robert H Lee

  • 1Laboratory for Neuroengineering, School of Electrical and Computer Engineering, Georgia Institute of Technology, Atlanta, GA 30332, USA.

Journal of Neural Engineering
|March 3, 2006
PubMed
Summary

Field-Programmable Gate Arrays (FPGAs) offer a powerful solution for complex neural modeling, providing high performance and cost-effectiveness. This technology enables efficient implementation of diverse neural models, overcoming limitations of traditional software approaches.

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

Inter-Observer Variability of Gerd-Related Metrics on High-Resolution Manometry and Calculation of the Milan Score.

Neurogastroenterology and motility·2026
Same author

A novel G13-RAPGEF2-RAP1 signaling pathway critical for platelet adhesion and aggregation.

Blood·2026
Same author

The fibrinogen αC region promotes arterial thrombosis in the context of hypofibrinogenemia.

Blood·2026
Same author

RAP1-RHO small GTPase cross-talk mediates integrin-dependent and -independent platelet procoagulant response.

Haematologica·2026
Same author

Platelet-mediated activation of perivascular mast cells triggers progression of sepsis to septic shock in mice.

Nature communications·2025
Same author

RAP1-RHO small GTPase cross-talk mediates integrin-dependent and -independent platelet procoagulant response.

bioRxiv : the preprint server for biology·2025

Area of Science:

  • Computational Neuroscience
  • Hardware Acceleration
  • Neuromorphic Engineering

Background:

  • Increasing complexity of neural models (larger populations, varied ionic conductances, detailed morphologies) challenges traditional software-based simulations.
  • Existing software models struggle to achieve desired performance levels for large-scale neural simulations.

Purpose of the Study:

  • To describe the implementation of neural models using Field-Programmable Gate Arrays (FPGAs).
  • To demonstrate how FPGAs can achieve high performance comparable to custom analogue circuits or computer clusters while maintaining software reconfigurability and affordability.

Main Methods:

  • Utilizing FPGAs to implement neural models as parallel processed data paths.
  • Developing generalized architectures for efficient modeling of first-order, nonlinear differential equations in throughput-maximizing or latency-minimizing configurations.

Related Experiment Videos

  • Exploiting model homogeneity in population and multi-compartment models to create deep pipelines for enhanced performance.

    • Main Results:

    • FPGAs enable the implementation of a wide range of single-compartment, multi-compartment, and population neural models.
    • Achieved performance levels rival custom analogue circuits or computer clusters at a cost comparable to personal computers.
    • Demonstrated efficient data-path configurations for differential equation modeling and pipeline architectures for homogeneous models.

    Conclusions:

    • FPGAs offer a viable and high-performance solution for complex neural modeling, overcoming the scalability limitations of traditional software.
    • The parallel processing capabilities and reconfigurability of FPGAs make them suitable for diverse neural simulation needs.
    • Future research should explore further limitations and advancements in FPGA architectures for neuromorphic computing.