Author

Casey S. Bahr

Date

August 1988

Document Type

Thesis

Degree Name

M.S.

Department

Dept. of Computer Science and Engineering

Institution

Oregon Graduate Center

Abstract

ANNE is a neural network simulation system designed for an MIMD distributed-memory computer and implemented on Intel's iPSC. ANNE is one part of a neural network hardware development system that includes a network description language, a standardized network structural format, and a utility to map network connection nodes to physical processor nodes. ANNE features user-variable, message-driven synchronization between iPSC nodes. This synchronization technique relies on a replication of the simulation clock among the iPSC nodes and the cube host processor. Each node clock runs independently of other node clocks for some number of cycles, before synchronizing with the global host clock. Messages between network connection nodes that must cross processor boundaries are packaged according to one of two methods: synchronous packetization (SP) or asynchronous packetization (AP). These messages are timestamped according to the clock of the local physical node, and this timestamp is used to determine a message's "alignment" upon arrival. Several aspects of ANNE's performance were examined via tests with several back-propagation and "receptive field" networks. Performance results from the iPSC are presented as well as preliminary results from a direct porting of ANNE to the iPSC/2.

Identifier

doi:10.6083/M43J39X7

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.