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

Designing a connectionist network supercomputer

K Asanović1, J Beck, J Feldman

  • 1University of California, Berkeley.

International Journal of Neural Systems
|December 1, 1993
PubMed
Summary

Researchers are designing a supercomputer for artificial neural networks, considering factors beyond just processing speed, inspired by Amdahl's Law and prior machine-building experience.

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

Spin environment of a superconducting qubit in high magnetic fields.

Nature communications·2025
Same author

Guiding clinical management of patients with CNS lymphomas by minimal-invasive detection of ctDNA in cerebrospinal fluid.

Leukemia·2025
Same author

Clinical symptoms and surgical outcome of colloid cysts of the third ventricle: A multicenter retrospective study.

Brain & spine·2025
Same author

"Making it work in the face of extreme adversity" - Exploring perceptions for the future of the imaging and oncology workforce using 'soundbite' interviews.

Radiography (London, England : 1995)·2024
Same author

Different Glymphatic Kinetics in Spontaneous Intracranial Hypotension.

AJNR. American journal of neuroradiology·2024
Same author

Spinal dementia: Don't miss it, it's treatable.

Neuroradiology·2024

Area of Science:

  • Computer Science
  • Artificial Intelligence
  • Supercomputing

Background:

  • Previous experience building simpler machines informed design choices.
  • Amdahl's Law was observed to significantly impact system performance.
  • Focus extended beyond raw computational power to other critical factors.

Purpose of the Study:

  • To develop a supercomputer optimized for artificial neural network (ANN) applications.
  • To incorporate lessons learned from previous machine designs.
  • To address limitations highlighted by Amdahl's Law in high-performance computing.

Main Methods:

  • Analyzing the influence of various design factors on performance.
  • Applying principles derived from Amdahl's Law to system architecture.
  • Describing application targets, machine goals, and system architecture.

Main Results:

  • Identification of key factors influencing supercomputer design for ANNs.
  • Quantification of the impact of these factors through rough expressions.
  • Development of a system architecture tailored to specific application needs.

Conclusions:

  • Supercomputer design for ANNs requires a holistic approach, considering factors beyond arithmetic speed.
  • Amdahl's Law underscores the importance of balancing computational and other system resources.
  • The designed architecture aims to meet specific application targets efficiently.

Related Experiment Videos