Partitioned schemes of mapping neural networks onto mesh-connected computers

Philip S. Liu, Luis Martinez

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Two new mapping methods for implementing neural networks on parallel mesh-connected processors are given. These are the rectangular region based mapping method and the no zero weight value mapping method. Implementation and performance comparisons with another existing method are carried out. One of these three methods can be used depending on system and problem parameters.

Original languageEnglish (US)
Title of host publicationConference Proceedings - IEEE SOUTHEASTCON
PublisherPubl by IEEE
Pages357-361
Number of pages5
ISBN (Print)078031798X
StatePublished - Jan 1 1994
EventProceedings of the IEEE SOUTHEASTCON '94 - Miami, FL, USA
Duration: Apr 10 1994Apr 13 1994

Publication series

NameConference Proceedings - IEEE SOUTHEASTCON
ISSN (Print)0734-7502

Other

OtherProceedings of the IEEE SOUTHEASTCON '94
CityMiami, FL, USA
Period4/10/944/13/94

ASJC Scopus subject areas

  • Electrical and Electronic Engineering

Fingerprint Dive into the research topics of 'Partitioned schemes of mapping neural networks onto mesh-connected computers'. Together they form a unique fingerprint.

Cite this