Summary
International Symposium on Nonlinear Theory and Its Applications
2015
Session Number:B4L-B
Session:
Number:B4L-B-1
Toward a Neural Network Computing: a Novel NN-SRAM
Ali Massoud Haidar, Nayif Saleh, Wassim Itani, Hiroyuki Shirahama,
pp.672-675
Publication Date:2015/12/1
Online ISSN:2188-5079
DOI:10.34385/proc.47.B4L-B-1
PDF download (396.1KB)
Summary:
A novel Neural Network Memory cell design is proposed. A Neural Network Static RAM (NN-SRAM) consists of internal latch that stores binary information. The stored information remains valid as long as power is applied to the NN-SRAM. The NN-SRAM is easy to use, where the simulation (using Matlab/Simulink) showed that the read and write cycles are short, reducing the energy consumption, thus it can be used in wireless sensor networks for extending the life time. The NN-SRAM is volatile, but no need to refresh. Also, this paper presents the neural network internal structure of m-words with n-bits SRAM chip. The chip is a combination of NN-SRAM cells and associated neural networks to select the word, read, write, input, and output; these neural networks are used to control the operation of the m x n NN-SRAM chip.