A dataflow processing element for neural network simulation

Author

Mutlaq, M. Abu ; Braham, R.

Author_Institution

Dept. of Comput. Eng., King Fahd Univ. of Pet. & Miner., Dhahran, Saudi Arabia

Volume

1

fYear

1995

fDate

Nov/Dec 1995

Firstpage

398

Abstract

Neural networks can be easily represented by macro dataflow graphs. Dataflow machines are thus suitable for simulation of these networks. In this paper, neural computing hardware considerations are first addressed. The architecture of a new argument-fetch dataflow processor dedicated to neural computing is then described. Backpropagation and Hopfield networks are transformed into dataflow graphs and simulated on the machine. Excellent performance results have been achieved

Keywords

Hopfield neural nets; backpropagation; data flow graphs; neural net architecture; performance evaluation; simulation; virtual machines; Hopfield networks; architecture; argument-fetch dataflow processor; backpropagation; dataflow processing; macro dataflow graphs; performance evaluation; simulation; Artificial neural networks; Backpropagation; Computational modeling; Computer architecture; Computer networks; Computer simulation; Concurrent computing; Distributed computing; Hardware; Neural network hardware; Neural networks; Neurons; Scalability;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Networks, 1995. Proceedings., IEEE International Conference on

Conference_Location

Perth, WA

Print_ISBN

0-7803-2768-3

Type

conf

DOI

10.1109/ICNN.1995.488133

Filename

488133