Device and materials requirements for neuromorphic computing

Raisul Islam, Haitong Li, Pai Yu Chen, Weier Wan, Hong Yu Chen, Bin Gao, Huaqiang Wu, Shimeng Yu, Krishna Saraswat, H. S. Philip Wong

Research output: Contribution to journalReview article

3 Citations (Scopus)

Abstract

Energy efficient hardware implementation of artificial neural network is challenging due the 'memory-wall' bottleneck. Neuromorphic computing promises to address this challenge by eliminating data movement to and from off-chip memory devices. Emerging non-volatile memory (NVM) devices that exhibit gradual changes in resistivity are a key enabler of in-memory computing - a type of neuromorphic computing. In this paper, we present a review of some of the NVM devices (RRAM, CBRAM, PCM) commonly used in neuromorphic application. The review focuses on the trade-off between device parameters such as retention, endurance, device-to-device variation, speed and resistance levels, and the interplay with target applications. This work aims at providing guidance for finding the optimized resistive memory devices material stack suitable for neuromorphic application.

Original languageEnglish (US)
Article number113001
JournalJournal of Physics D: Applied Physics
Volume52
Issue number11
DOIs
StatePublished - Jan 18 2019

Fingerprint

Data storage equipment
requirements
chips (memory devices)
Pulse code modulation
endurance
emerging
hardware
Durability
Neural networks
Hardware
electrical resistivity
energy

Keywords

  • deep neural network
  • neuromorphic computing
  • non volatile memory

ASJC Scopus subject areas

  • Electronic, Optical and Magnetic Materials
  • Condensed Matter Physics
  • Acoustics and Ultrasonics
  • Surfaces, Coatings and Films

Cite this

Islam, R., Li, H., Chen, P. Y., Wan, W., Chen, H. Y., Gao, B., ... Philip Wong, H. S. (2019). Device and materials requirements for neuromorphic computing. Journal of Physics D: Applied Physics, 52(11), [113001]. https://doi.org/10.1088/1361-6463/aaf784

Device and materials requirements for neuromorphic computing. / Islam, Raisul; Li, Haitong; Chen, Pai Yu; Wan, Weier; Chen, Hong Yu; Gao, Bin; Wu, Huaqiang; Yu, Shimeng; Saraswat, Krishna; Philip Wong, H. S.

In: Journal of Physics D: Applied Physics, Vol. 52, No. 11, 113001, 18.01.2019.

Research output: Contribution to journalReview article

Islam, R, Li, H, Chen, PY, Wan, W, Chen, HY, Gao, B, Wu, H, Yu, S, Saraswat, K & Philip Wong, HS 2019, 'Device and materials requirements for neuromorphic computing', Journal of Physics D: Applied Physics, vol. 52, no. 11, 113001. https://doi.org/10.1088/1361-6463/aaf784
Islam, Raisul ; Li, Haitong ; Chen, Pai Yu ; Wan, Weier ; Chen, Hong Yu ; Gao, Bin ; Wu, Huaqiang ; Yu, Shimeng ; Saraswat, Krishna ; Philip Wong, H. S. / Device and materials requirements for neuromorphic computing. In: Journal of Physics D: Applied Physics. 2019 ; Vol. 52, No. 11.
@article{30f7b2fb659e44a49773ca9ede10722e,
title = "Device and materials requirements for neuromorphic computing",
abstract = "Energy efficient hardware implementation of artificial neural network is challenging due the 'memory-wall' bottleneck. Neuromorphic computing promises to address this challenge by eliminating data movement to and from off-chip memory devices. Emerging non-volatile memory (NVM) devices that exhibit gradual changes in resistivity are a key enabler of in-memory computing - a type of neuromorphic computing. In this paper, we present a review of some of the NVM devices (RRAM, CBRAM, PCM) commonly used in neuromorphic application. The review focuses on the trade-off between device parameters such as retention, endurance, device-to-device variation, speed and resistance levels, and the interplay with target applications. This work aims at providing guidance for finding the optimized resistive memory devices material stack suitable for neuromorphic application.",
keywords = "deep neural network, neuromorphic computing, non volatile memory",
author = "Raisul Islam and Haitong Li and Chen, {Pai Yu} and Weier Wan and Chen, {Hong Yu} and Bin Gao and Huaqiang Wu and Shimeng Yu and Krishna Saraswat and {Philip Wong}, {H. S.}",
year = "2019",
month = "1",
day = "18",
doi = "10.1088/1361-6463/aaf784",
language = "English (US)",
volume = "52",
journal = "Journal Physics D: Applied Physics",
issn = "0022-3727",
publisher = "IOP Publishing Ltd.",
number = "11",

}

TY - JOUR

T1 - Device and materials requirements for neuromorphic computing

AU - Islam, Raisul

AU - Li, Haitong

AU - Chen, Pai Yu

AU - Wan, Weier

AU - Chen, Hong Yu

AU - Gao, Bin

AU - Wu, Huaqiang

AU - Yu, Shimeng

AU - Saraswat, Krishna

AU - Philip Wong, H. S.

PY - 2019/1/18

Y1 - 2019/1/18

N2 - Energy efficient hardware implementation of artificial neural network is challenging due the 'memory-wall' bottleneck. Neuromorphic computing promises to address this challenge by eliminating data movement to and from off-chip memory devices. Emerging non-volatile memory (NVM) devices that exhibit gradual changes in resistivity are a key enabler of in-memory computing - a type of neuromorphic computing. In this paper, we present a review of some of the NVM devices (RRAM, CBRAM, PCM) commonly used in neuromorphic application. The review focuses on the trade-off between device parameters such as retention, endurance, device-to-device variation, speed and resistance levels, and the interplay with target applications. This work aims at providing guidance for finding the optimized resistive memory devices material stack suitable for neuromorphic application.

AB - Energy efficient hardware implementation of artificial neural network is challenging due the 'memory-wall' bottleneck. Neuromorphic computing promises to address this challenge by eliminating data movement to and from off-chip memory devices. Emerging non-volatile memory (NVM) devices that exhibit gradual changes in resistivity are a key enabler of in-memory computing - a type of neuromorphic computing. In this paper, we present a review of some of the NVM devices (RRAM, CBRAM, PCM) commonly used in neuromorphic application. The review focuses on the trade-off between device parameters such as retention, endurance, device-to-device variation, speed and resistance levels, and the interplay with target applications. This work aims at providing guidance for finding the optimized resistive memory devices material stack suitable for neuromorphic application.

KW - deep neural network

KW - neuromorphic computing

KW - non volatile memory

UR - http://www.scopus.com/inward/record.url?scp=85062610098&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85062610098&partnerID=8YFLogxK

U2 - 10.1088/1361-6463/aaf784

DO - 10.1088/1361-6463/aaf784

M3 - Review article

VL - 52

JO - Journal Physics D: Applied Physics

JF - Journal Physics D: Applied Physics

SN - 0022-3727

IS - 11

M1 - 113001

ER -