DMCA
Fast Neural Net Simulation with a DSP Processor Array (1993)
| Citations: | 7 - 0 self |
BibTeX
@MISC{Müller93fastneural,
author = {Urs A. Müller and Anton Gunzinger and Walter Guggenbühl},
title = {Fast Neural Net Simulation with a DSP Processor Array},
year = {1993}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
This paper describes the implementation of a fast neural net simulator on a novel parallel distributed-memory computer. A 60-processor system, named MUSIC, 1 is operational and runs the back-propagation algorithm at a speed of 247 million connection updates per second (continuous weight update) using 32 bit floating-point precision. This is equal to 1 Gflops sustained performance. The complete system with 3.6 Gflops peak performance consumes less than 800 Watt of electrical power and fits into a 19-inch rack. While reaching the speed of modern supercomputers, MUSIC still can be used as a personal desktop computer at a researchers own disposal. In neural net simulation, this gives a computing performance to a single user which was unthinkable before. The system's real-time interfaces make it especially useful for embedded applications. 1. Introduction A major limitation in the research of learning behavior and learning algorithms of artificial neural networks are the long c...
Keyphrases
dsp processor array fast neural net simulation 19-inch rack single user electrical power researcher disposal major limitation 60-processor system real-time interface continuous weight update gflops peak performance consumes artificial neural network connection update complete system embedded application modern supercomputer novel parallel distributed-memory computer back-propagation algorithm bit floating-point precision personal desktop computer neural net simulation fast neural net simulator
