Share Email Print

Proceedings Paper

Systolic Implementation Of Neural Network
Author(s): A J De Groot; S. R. Parker
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

The backpropagation algorithm for error gradient calculations in multilayer, feed-forward neural networks is derived in matrix form involving inner and outer products. It is demonstrated that these calculations can be carried out efficiently using systolic processing techniques [3], particularly using the SPRINT, a 64-element systolic processor developed at Lawrence Livermore National Laboratory. This machine contains one million synapses, and forward-propagates 12 million connections per second, using 100 watts of power. When executing the algorithm, each SPRINT processor performs useful work 97% of the time. The theory and applications are confirmed by some nontrivial examples involving seismic signal recognition.

Paper Details

Date Published: 17 May 1989
PDF: 10 pages
Proc. SPIE 1058, High Speed Computing II, (17 May 1989); doi: 10.1117/12.951681
Show Author Affiliations
A J De Groot, Lawrence Livermore National Laboratory (United States)
S. R. Parker, Lawrence Livermore National Laboratory (United States)

Published in SPIE Proceedings Vol. 1058:
High Speed Computing II
Keith Bromley, Editor(s)

© SPIE. Terms of Use
Back to Top