Hardware requirements for neural network pattern classifiers

A case study and implementation

Bernhard E. Boser, Eduard Sackinger, Yann LeCun, Yann leCun, Lawrence D. Jackel

Research output: Contribution to journalArticle

Abstract

A special-purpose chip, optimized for computational needs of neural networks and performing over 2000 multiplications and additions simultaneously, is described. Its data path is particularly suitable for the convolutional architectures typical in pattern classification networks but can also be configured for fully connected or feedback topologies. A development system permits rapid prototyping of new applications and analysis of the impact of the specialized hardware on system performance. The power and flexibility of the processor are demonstrated with a neural network for handwritten character recognition containing over 133,000 connections.

Original languageEnglish (US)
Pages (from-to)32-40
Number of pages9
JournalIEEE Micro
Volume12
Issue number1
DOIs
StatePublished - Feb 1992

Fingerprint

Classifiers
Neural networks
Hardware
Character recognition
Rapid prototyping
Pattern recognition
Topology
Feedback

ASJC Scopus subject areas

  • Computer Graphics and Computer-Aided Design
  • Hardware and Architecture
  • Software

Cite this

Hardware requirements for neural network pattern classifiers : A case study and implementation. / Boser, Bernhard E.; Sackinger, Eduard; LeCun, Yann; leCun, Yann; Jackel, Lawrence D.

In: IEEE Micro, Vol. 12, No. 1, 02.1992, p. 32-40.

Research output: Contribution to journalArticle

Boser, Bernhard E. ; Sackinger, Eduard ; LeCun, Yann ; leCun, Yann ; Jackel, Lawrence D. / Hardware requirements for neural network pattern classifiers : A case study and implementation. In: IEEE Micro. 1992 ; Vol. 12, No. 1. pp. 32-40.
@article{704079dee11a443a89fc1091a753834a,
title = "Hardware requirements for neural network pattern classifiers: A case study and implementation",
abstract = "A special-purpose chip, optimized for computational needs of neural networks and performing over 2000 multiplications and additions simultaneously, is described. Its data path is particularly suitable for the convolutional architectures typical in pattern classification networks but can also be configured for fully connected or feedback topologies. A development system permits rapid prototyping of new applications and analysis of the impact of the specialized hardware on system performance. The power and flexibility of the processor are demonstrated with a neural network for handwritten character recognition containing over 133,000 connections.",
author = "Boser, {Bernhard E.} and Eduard Sackinger and Yann LeCun and Yann leCun and Jackel, {Lawrence D.}",
year = "1992",
month = "2",
doi = "10.1109/40.124378",
language = "English (US)",
volume = "12",
pages = "32--40",
journal = "IEEE Micro",
issn = "0272-1732",
publisher = "IEEE Computer Society",
number = "1",

}

TY - JOUR

T1 - Hardware requirements for neural network pattern classifiers

T2 - A case study and implementation

AU - Boser, Bernhard E.

AU - Sackinger, Eduard

AU - LeCun, Yann

AU - leCun, Yann

AU - Jackel, Lawrence D.

PY - 1992/2

Y1 - 1992/2

N2 - A special-purpose chip, optimized for computational needs of neural networks and performing over 2000 multiplications and additions simultaneously, is described. Its data path is particularly suitable for the convolutional architectures typical in pattern classification networks but can also be configured for fully connected or feedback topologies. A development system permits rapid prototyping of new applications and analysis of the impact of the specialized hardware on system performance. The power and flexibility of the processor are demonstrated with a neural network for handwritten character recognition containing over 133,000 connections.

AB - A special-purpose chip, optimized for computational needs of neural networks and performing over 2000 multiplications and additions simultaneously, is described. Its data path is particularly suitable for the convolutional architectures typical in pattern classification networks but can also be configured for fully connected or feedback topologies. A development system permits rapid prototyping of new applications and analysis of the impact of the specialized hardware on system performance. The power and flexibility of the processor are demonstrated with a neural network for handwritten character recognition containing over 133,000 connections.

UR - http://www.scopus.com/inward/record.url?scp=0026818499&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0026818499&partnerID=8YFLogxK

U2 - 10.1109/40.124378

DO - 10.1109/40.124378

M3 - Article

VL - 12

SP - 32

EP - 40

JO - IEEE Micro

JF - IEEE Micro

SN - 0272-1732

IS - 1

ER -