Towards Efficient RRAM-based Quantized Neural Networks Hardware: State-of-the-art and Open Issues

O. Krestinskaya; L. Zhang; K. N. Salama

doi:10.1109/NANO54668.2022.9928590

Towards Efficient RRAM-based Quantized Neural Networks Hardware: State-of-the-art and Open Issues

O. Krestinskaya, L. Zhang, K. N. Salama

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

3 Scopus citations

Abstract

The increasing amount of data processed on edge and demand for reducing the energy consumption for large neural network architectures have initiated the transition from traditional von Neumann architectures towards in-memory computing paradigms. Quantization is one of the methods to reduce power and computation requirements for neural networks by limiting bit precision. Resistive Random Access Memory (RRAM) devices are great candidates for Quantized Neural Networks (QNN) implementations. As the number of possible conductive states in RRAMs is limited, a certain level of quantization is always considered when designing RRAM-based neural networks. In this work, we provide a comprehensive analysis of state-of-the-art RRAM-based QNN implementations, showing where RRAMs stand in terms of satisfying the criteria of efficient QNN hardware. We cover hardware and device challenges related to QNNs and show the main unsolved issues and possible future research directions.

Original language	English (US)
Title of host publication	2022 IEEE 22nd International Conference on Nanotechnology, NANO 2022
Publisher	IEEE Computer Society
Pages	465-468
Number of pages	4
ISBN (Electronic)	9781665452250
DOIs	https://doi.org/10.1109/NANO54668.2022.9928590
State	Published - 2022
Event	22nd IEEE International Conference on Nanotechnology, NANO 2022 - Palma de Mallorca, Spain Duration: Jul 4 2022 → Jul 8 2022

Publication series

Name	Proceedings of the IEEE Conference on Nanotechnology
Volume	2022-July
ISSN (Print)	1944-9399
ISSN (Electronic)	1944-9380

Conference

Conference	22nd IEEE International Conference on Nanotechnology, NANO 2022
Country/Territory	Spain
City	Palma de Mallorca
Period	07/4/22 → 07/8/22

Bibliographical note

Publisher Copyright:
© 2022 IEEE.

Keywords

Hardware
QNN
Quantization
RRAM

ASJC Scopus subject areas

Bioengineering
Electrical and Electronic Engineering
Materials Chemistry
Condensed Matter Physics

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1109/NANO54668.2022.9928590

Cite this

Krestinskaya, O., Zhang, L., & Salama, K. N. (2022). Towards Efficient RRAM-based Quantized Neural Networks Hardware: State-of-the-art and Open Issues. In 2022 IEEE 22nd International Conference on Nanotechnology, NANO 2022 (pp. 465-468). (Proceedings of the IEEE Conference on Nanotechnology; Vol. 2022-July). IEEE Computer Society. https://doi.org/10.1109/NANO54668.2022.9928590

@inproceedings{9b11df074b4a48deaee78a7228787107,

title = "Towards Efficient RRAM-based Quantized Neural Networks Hardware: State-of-the-art and Open Issues",

abstract = "The increasing amount of data processed on edge and demand for reducing the energy consumption for large neural network architectures have initiated the transition from traditional von Neumann architectures towards in-memory computing paradigms. Quantization is one of the methods to reduce power and computation requirements for neural networks by limiting bit precision. Resistive Random Access Memory (RRAM) devices are great candidates for Quantized Neural Networks (QNN) implementations. As the number of possible conductive states in RRAMs is limited, a certain level of quantization is always considered when designing RRAM-based neural networks. In this work, we provide a comprehensive analysis of state-of-the-art RRAM-based QNN implementations, showing where RRAMs stand in terms of satisfying the criteria of efficient QNN hardware. We cover hardware and device challenges related to QNNs and show the main unsolved issues and possible future research directions.",

keywords = "Hardware, QNN, Quantization, RRAM",

author = "O. Krestinskaya and L. Zhang and Salama, {K. N.}",

note = "Publisher Copyright: {\textcopyright} 2022 IEEE.; 22nd IEEE International Conference on Nanotechnology, NANO 2022 ; Conference date: 04-07-2022 Through 08-07-2022",

year = "2022",

doi = "10.1109/NANO54668.2022.9928590",

language = "English (US)",

series = "Proceedings of the IEEE Conference on Nanotechnology",

publisher = "IEEE Computer Society",

pages = "465--468",

booktitle = "2022 IEEE 22nd International Conference on Nanotechnology, NANO 2022",

address = "United States",

}

Krestinskaya, O , Zhang, L & Salama, KN 2022, Towards Efficient RRAM-based Quantized Neural Networks Hardware: State-of-the-art and Open Issues. in 2022 IEEE 22nd International Conference on Nanotechnology, NANO 2022. Proceedings of the IEEE Conference on Nanotechnology, vol. 2022-July, IEEE Computer Society, pp. 465-468, 22nd IEEE International Conference on Nanotechnology, NANO 2022, Palma de Mallorca, Spain, 07/4/22. https://doi.org/10.1109/NANO54668.2022.9928590

Towards Efficient RRAM-based Quantized Neural Networks Hardware: State-of-the-art and Open Issues. / Krestinskaya, O.; Zhang, L.; Salama, K. N.
2022 IEEE 22nd International Conference on Nanotechnology, NANO 2022. IEEE Computer Society, 2022. p. 465-468 (Proceedings of the IEEE Conference on Nanotechnology; Vol. 2022-July).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Towards Efficient RRAM-based Quantized Neural Networks Hardware

T2 - 22nd IEEE International Conference on Nanotechnology, NANO 2022

AU - Krestinskaya, O.

AU - Zhang, L.

AU - Salama, K. N.

PY - 2022

Y1 - 2022

N2 - The increasing amount of data processed on edge and demand for reducing the energy consumption for large neural network architectures have initiated the transition from traditional von Neumann architectures towards in-memory computing paradigms. Quantization is one of the methods to reduce power and computation requirements for neural networks by limiting bit precision. Resistive Random Access Memory (RRAM) devices are great candidates for Quantized Neural Networks (QNN) implementations. As the number of possible conductive states in RRAMs is limited, a certain level of quantization is always considered when designing RRAM-based neural networks. In this work, we provide a comprehensive analysis of state-of-the-art RRAM-based QNN implementations, showing where RRAMs stand in terms of satisfying the criteria of efficient QNN hardware. We cover hardware and device challenges related to QNNs and show the main unsolved issues and possible future research directions.

AB - The increasing amount of data processed on edge and demand for reducing the energy consumption for large neural network architectures have initiated the transition from traditional von Neumann architectures towards in-memory computing paradigms. Quantization is one of the methods to reduce power and computation requirements for neural networks by limiting bit precision. Resistive Random Access Memory (RRAM) devices are great candidates for Quantized Neural Networks (QNN) implementations. As the number of possible conductive states in RRAMs is limited, a certain level of quantization is always considered when designing RRAM-based neural networks. In this work, we provide a comprehensive analysis of state-of-the-art RRAM-based QNN implementations, showing where RRAMs stand in terms of satisfying the criteria of efficient QNN hardware. We cover hardware and device challenges related to QNNs and show the main unsolved issues and possible future research directions.

KW - Hardware

KW - QNN

KW - Quantization

KW - RRAM

UR - http://www.scopus.com/inward/record.url?scp=85142930048&partnerID=8YFLogxK

U2 - 10.1109/NANO54668.2022.9928590

DO - 10.1109/NANO54668.2022.9928590

M3 - Conference contribution

AN - SCOPUS:85142930048

T3 - Proceedings of the IEEE Conference on Nanotechnology

SP - 465

EP - 468

BT - 2022 IEEE 22nd International Conference on Nanotechnology, NANO 2022

PB - IEEE Computer Society

Y2 - 4 July 2022 through 8 July 2022

ER -

Towards Efficient RRAM-based Quantized Neural Networks Hardware: State-of-the-art and Open Issues

Abstract

Publication series

Conference

Bibliographical note

Keywords

ASJC Scopus subject areas

UN SDGs

Access to Document

Other files and links

Fingerprint

Cite this