Variation-aware binarized memristive networks

Lammie, Corey, Krestinskaya, Olga, James, Alex, and Rahimi Azghadi, Mostafa (2019) Variation-aware binarized memristive networks. In: Proceedings of the 26th IEEE International Conference on Electronics, Circuits and Systems. pp. 490-493. From: 2019 ICECS: 26th IEEE International Conference on Electronics, Circuits and Systems, 27-29 November 2019, Genoa, Italy.

[img]
Preview
PDF (Published Version) - Published Version
Available under License Creative Commons Attribution.

Download (4MB) | Preview
View at Publisher Website: https://doi.org/10.1109/ICECS46596.2019....
 
36


Abstract

The quantization of weights to binary states in Deep Neural Networks (DNNs) can replace resource-hungry multiply accumulate operations with simple accumulations. Such Binarized Neural Networks (BNNs) exhibit greatly reduced resource and power requirements. In addition, memristors have been shown as promising synaptic weight elements in DNNs. In this paper, we propose and simulate novel Binarized Memristive Convolutional Neural Network (BMCNN) architectures employing hybrid weight and parameter representations. We train the proposed architectures offline and then map the trained parameters to our binarized memristive devices for inference. To take into account the variations in memristive devices, and to study their effect on the performance, we introduce variations in R ON and R OFF . Moreover, we introduce means to mitigate the adverse effect of memristive variations in our proposed networks. Finally, we benchmark our BMCNNs and variation-aware BMCNNs using the MNIST dataset.

Item ID: 62460
Item Type: Conference Item (Research - E1)
ISBN: 978-1-7281-0996-1
Copyright Information: Copyright © 2019, IEEE
Date Deposited: 11 Mar 2020 00:08
FoR Codes: 09 ENGINEERING > 0906 Electrical and Electronic Engineering > 090604 Microelectronics and Integrated Circuits @ 50%
09 ENGINEERING > 0906 Electrical and Electronic Engineering > 090601 Circuits and Systems @ 50%
SEO Codes: 97 EXPANDING KNOWLEDGE > 970109 Expanding Knowledge in Engineering @ 50%
97 EXPANDING KNOWLEDGE > 970110 Expanding Knowledge in Technology @ 50%
Downloads: Total: 36
Last 12 Months: 25
More Statistics

Actions (Repository Staff Only)

Item Control Page Item Control Page