Rochester Connectionist Simulation Environment

Research output: Chapter in Book/Report/Conference proceedingChapter

Abstract

The Rochester Connectionist Simulator, one of the earliest public domain neural network simulation environments, was designed for research into structured connectionist modeling. The main design criteria have been flexibility and generality. The network model consists of units, sites and links, each unit having a number of sites at which incoming links arrive. The network data structures are user-extensible allowing arbitrary state variables to be associated with each unit, site and link. The base simulator is written in C and the user has a choice of C, C++, Lisp or Scheme as the language for network specification, using simulator primitives to create units, sites and links. Update functions are taken from the library or written by the user. The X11 graphics interface provides multiple display windows, customizable mouse actions and the ability to arbitrarily extend functionality. Other forms of control are exercised via a user-extensible command language. A parallel version exhibiting near-linear speedup on a shared-memory multiprocessor is described. Some examples are given of the wide variety of research projects the simulator has been used for.
Original languageEnglish
Title of host publicationNeural Network Simulation Environments
EditorsJosef Skrzypek
Place of PublicationBoston, MA
PublisherSpringer
Pages187-207
Number of pages21
ISBN (Electronic)978-1-4615-2736-7
ISBN (Print)978-1-4613-6180-0
DOIs
Publication statusPublished - 1994

Publication series

NameThe Kluwer International Series in Engineering and Computer Science
PublisherSpringer US
Volume254
ISSN (Print)0893-3405

Fingerprint

Dive into the research topics of 'Rochester Connectionist Simulation Environment'. Together they form a unique fingerprint.

Cite this