Acceleration of Convolutional Networks Using Nanoscale Memristive Devices

Shruti R. Kulkarni; Anakha V. Babu; Bipin Rajendran

doi:10.1007/978-3-319-98204-5_20

Acceleration of Convolutional Networks Using Nanoscale Memristive Devices

Shruti R. Kulkarni
, Anakha V. Babu
, Bipin Rajendran

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

1 Scopus citations

Abstract

We discuss a convolutional neural network for handwritten digit classification and its hardware acceleration as an inference engine using nanoscale memristive devices in the spike domain. We study the impact of device programming variability on the spiking neural network’s (SNN) inference accuracy and benchmark its performance with an equivalent artificial neural network (ANN). We demonstrate optimization strategies to implement these networks with memristive devices with an on-off ratio as low as 10 and only 32 levels of resolution. Further, close to baseline accuracies can be maintained for the networks even if such memristive devices are used to duplicate the pre-determined kernel weights to enable parallel execution of the convolution operation.

Original language	English
Title of host publication	Engineering Applications of Neural Networks - 19th International Conference, EANN 2018, Proceedings
Editors	Elias Pimenidis, Chrisina Jayne
Publisher	Springer Verlag
Pages	240-251
Number of pages	12
ISBN (Print)	9783319982038
DOIs	https://doi.org/10.1007/978-3-319-98204-5_20
State	Published - 2018
Externally published	Yes
Event	19th International Conference on Engineering Applications of Neural Networks, EANN 2018 - Bristol, United Kingdom Duration: Sep 3 2018 → Sep 5 2018

Publication series

Name	Communications in Computer and Information Science
Volume	893
ISSN (Print)	1865-0929

Conference

Conference	19th International Conference on Engineering Applications of Neural Networks, EANN 2018
Country/Territory	United Kingdom
City	Bristol
Period	09/3/18 → 09/5/18

Funding

Acknowledgments. This research was supported in part by the CAMPUSENSE project grant from CISCO Systems Inc, the Semiconductor Research Corporation (2016-SD-2717), and the National Science Foundation grant 1710009.

Keywords

Artificial neural networks
Memristors
Non-volatile memory devices
Programming variability
Spiking neural networks

Access to Document

10.1007/978-3-319-98204-5_20

Cite this

Kulkarni, S. R., Babu, A. V., & Rajendran, B. (2018). Acceleration of Convolutional Networks Using Nanoscale Memristive Devices. In E. Pimenidis, & C. Jayne (Eds.), Engineering Applications of Neural Networks - 19th International Conference, EANN 2018, Proceedings (pp. 240-251). (Communications in Computer and Information Science; Vol. 893). Springer Verlag. https://doi.org/10.1007/978-3-319-98204-5_20

@inproceedings{e5d54d8306ed4bc58f8a35261c4b4118,

title = "Acceleration of Convolutional Networks Using Nanoscale Memristive Devices",

abstract = "We discuss a convolutional neural network for handwritten digit classification and its hardware acceleration as an inference engine using nanoscale memristive devices in the spike domain. We study the impact of device programming variability on the spiking neural network{\textquoteright}s (SNN) inference accuracy and benchmark its performance with an equivalent artificial neural network (ANN). We demonstrate optimization strategies to implement these networks with memristive devices with an on-off ratio as low as 10 and only 32 levels of resolution. Further, close to baseline accuracies can be maintained for the networks even if such memristive devices are used to duplicate the pre-determined kernel weights to enable parallel execution of the convolution operation.",

keywords = "Artificial neural networks, Memristors, Non-volatile memory devices, Programming variability, Spiking neural networks",

author = "Kulkarni, \{Shruti R.\} and Babu, \{Anakha V.\} and Bipin Rajendran",

note = "Publisher Copyright: {\textcopyright} 2018, Springer Nature Switzerland AG.; 19th International Conference on Engineering Applications of Neural Networks, EANN 2018 ; Conference date: 03-09-2018 Through 05-09-2018",

year = "2018",

doi = "10.1007/978-3-319-98204-5\_20",

language = "English",

isbn = "9783319982038",

series = "Communications in Computer and Information Science",

publisher = "Springer Verlag",

pages = "240--251",

editor = "Elias Pimenidis and Chrisina Jayne",

booktitle = "Engineering Applications of Neural Networks - 19th International Conference, EANN 2018, Proceedings",

}

Kulkarni, SR, Babu, AV & Rajendran, B 2018, Acceleration of Convolutional Networks Using Nanoscale Memristive Devices. in E Pimenidis & C Jayne (eds), Engineering Applications of Neural Networks - 19th International Conference, EANN 2018, Proceedings. Communications in Computer and Information Science, vol. 893, Springer Verlag, pp. 240-251, 19th International Conference on Engineering Applications of Neural Networks, EANN 2018, Bristol, United Kingdom, 09/3/18. https://doi.org/10.1007/978-3-319-98204-5_20

Acceleration of Convolutional Networks Using Nanoscale Memristive Devices. / Kulkarni, Shruti R.; Babu, Anakha V.; Rajendran, Bipin.
Engineering Applications of Neural Networks - 19th International Conference, EANN 2018, Proceedings. ed. / Elias Pimenidis; Chrisina Jayne. Springer Verlag, 2018. p. 240-251 (Communications in Computer and Information Science; Vol. 893).

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

TY - GEN

T1 - Acceleration of Convolutional Networks Using Nanoscale Memristive Devices

AU - Kulkarni, Shruti R.

AU - Babu, Anakha V.

AU - Rajendran, Bipin

PY - 2018

Y1 - 2018

N2 - We discuss a convolutional neural network for handwritten digit classification and its hardware acceleration as an inference engine using nanoscale memristive devices in the spike domain. We study the impact of device programming variability on the spiking neural network’s (SNN) inference accuracy and benchmark its performance with an equivalent artificial neural network (ANN). We demonstrate optimization strategies to implement these networks with memristive devices with an on-off ratio as low as 10 and only 32 levels of resolution. Further, close to baseline accuracies can be maintained for the networks even if such memristive devices are used to duplicate the pre-determined kernel weights to enable parallel execution of the convolution operation.

AB - We discuss a convolutional neural network for handwritten digit classification and its hardware acceleration as an inference engine using nanoscale memristive devices in the spike domain. We study the impact of device programming variability on the spiking neural network’s (SNN) inference accuracy and benchmark its performance with an equivalent artificial neural network (ANN). We demonstrate optimization strategies to implement these networks with memristive devices with an on-off ratio as low as 10 and only 32 levels of resolution. Further, close to baseline accuracies can be maintained for the networks even if such memristive devices are used to duplicate the pre-determined kernel weights to enable parallel execution of the convolution operation.

KW - Artificial neural networks

KW - Memristors

KW - Non-volatile memory devices

KW - Programming variability

KW - Spiking neural networks

UR - https://www.scopus.com/pages/publications/85052887581

U2 - 10.1007/978-3-319-98204-5_20

DO - 10.1007/978-3-319-98204-5_20

M3 - Conference contribution

AN - SCOPUS:85052887581

SN - 9783319982038

T3 - Communications in Computer and Information Science

SP - 240

EP - 251

BT - Engineering Applications of Neural Networks - 19th International Conference, EANN 2018, Proceedings

A2 - Pimenidis, Elias

A2 - Jayne, Chrisina

PB - Springer Verlag

T2 - 19th International Conference on Engineering Applications of Neural Networks, EANN 2018

Y2 - 3 September 2018 through 5 September 2018

ER -

Acceleration of Convolutional Networks Using Nanoscale Memristive Devices

Abstract

Publication series

Conference

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this