17 May 1989 Systolic Implementation Of Neural Network
Author Affiliations +
Proceedings Volume 1058, High Speed Computing II; (1989) https://doi.org/10.1117/12.951681
Event: OE/LASE '89, 1989, Los Angeles, CA, United States
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.
© (1989) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
A J De Groot, S. R. Parker, "Systolic Implementation Of Neural Network", Proc. SPIE 1058, High Speed Computing II, (17 May 1989); doi: 10.1117/12.951681; https://doi.org/10.1117/12.951681

Back to Top