Weight tuning of resistive memories and convolution kernel operation on cross-point array for neuro-inspired computing

Ligang Gao; Pai Yu Chen; Shimeng Yu

doi:10.1109/ICSICT.2016.7998889

Weight tuning of resistive memories and convolution kernel operation on cross-point array for neuro-inspired computing

Ligang Gao, Pai Yu Chen, Shimeng Yu

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

1 Scopus citations

Abstract

Analog conductance of resistive memories is attractive for implementing the weights in neuro-inspired algorithms. One of the most popular deep learning algorithms is the convolutional neural network (CNN). In this paper, we review our recent progress on using resistive memories for neuro-inspired computing. First, we optimized the iterative programming protocol to tune the weights of HfO_x based resistive memories by adjusting the pulse amplitude incremental steps, the pulse width incremental steps, and the start voltages. Then, we demonstrated the key operation in the CNN-the convolution kernel on a 12×12 cross-point array. As a proof-of-concept demonstration, we use the offline trained edge filters to detect both horizontal and vertical edges of the 50×50 pixels of a grayscale dog image. The experimental kernel operation matches the simulation results.

Original language	English (US)
Title of host publication	2016 13th IEEE International Conference on Solid-State and Integrated Circuit Technology, ICSICT 2016 - Proceedings
Editors	Yu-Long Jiang, Ting-Ao Tang, Ru Huang
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	247-250
Number of pages	4
ISBN (Electronic)	9781467397179
DOIs	https://doi.org/10.1109/ICSICT.2016.7998889
State	Published - 2016
Event	13th IEEE International Conference on Solid-State and Integrated Circuit Technology, ICSICT 2016 - Hangzhou, China Duration: Oct 25 2016 → Oct 28 2016

Publication series

Name	2016 13th IEEE International Conference on Solid-State and Integrated Circuit Technology, ICSICT 2016 - Proceedings

Other

Other	13th IEEE International Conference on Solid-State and Integrated Circuit Technology, ICSICT 2016
Country/Territory	China
City	Hangzhou
Period	10/25/16 → 10/28/16

ASJC Scopus subject areas

Electrical and Electronic Engineering
Safety, Risk, Reliability and Quality
Electronic, Optical and Magnetic Materials

Access to Document

10.1109/ICSICT.2016.7998889

Cite this

Gao, L., Chen, P. Y., & Yu, S. (2016). Weight tuning of resistive memories and convolution kernel operation on cross-point array for neuro-inspired computing. In Y.-L. Jiang, T.-A. Tang, & R. Huang (Eds.), 2016 13th IEEE International Conference on Solid-State and Integrated Circuit Technology, ICSICT 2016 - Proceedings (pp. 247-250). Article 7998889 (2016 13th IEEE International Conference on Solid-State and Integrated Circuit Technology, ICSICT 2016 - Proceedings). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ICSICT.2016.7998889

Weight tuning of resistive memories and convolution kernel operation on cross-point array for neuro-inspired computing. / Gao, Ligang; Chen, Pai Yu; Yu, Shimeng.
2016 13th IEEE International Conference on Solid-State and Integrated Circuit Technology, ICSICT 2016 - Proceedings. ed. / Yu-Long Jiang; Ting-Ao Tang; Ru Huang. Institute of Electrical and Electronics Engineers Inc., 2016. p. 247-250 7998889 (2016 13th IEEE International Conference on Solid-State and Integrated Circuit Technology, ICSICT 2016 - Proceedings).

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

Gao, L, Chen, PY & Yu, S 2016, Weight tuning of resistive memories and convolution kernel operation on cross-point array for neuro-inspired computing. in Y-L Jiang, T-A Tang & R Huang (eds), 2016 13th IEEE International Conference on Solid-State and Integrated Circuit Technology, ICSICT 2016 - Proceedings., 7998889, 2016 13th IEEE International Conference on Solid-State and Integrated Circuit Technology, ICSICT 2016 - Proceedings, Institute of Electrical and Electronics Engineers Inc., pp. 247-250, 13th IEEE International Conference on Solid-State and Integrated Circuit Technology, ICSICT 2016, Hangzhou, China, 10/25/16. https://doi.org/10.1109/ICSICT.2016.7998889

Gao L, Chen PY, Yu S. Weight tuning of resistive memories and convolution kernel operation on cross-point array for neuro-inspired computing. In Jiang YL, Tang TA, Huang R, editors, 2016 13th IEEE International Conference on Solid-State and Integrated Circuit Technology, ICSICT 2016 - Proceedings. Institute of Electrical and Electronics Engineers Inc. 2016. p. 247-250. 7998889. (2016 13th IEEE International Conference on Solid-State and Integrated Circuit Technology, ICSICT 2016 - Proceedings). doi: 10.1109/ICSICT.2016.7998889

Gao, Ligang ; Chen, Pai Yu ; Yu, Shimeng. / Weight tuning of resistive memories and convolution kernel operation on cross-point array for neuro-inspired computing. 2016 13th IEEE International Conference on Solid-State and Integrated Circuit Technology, ICSICT 2016 - Proceedings. editor / Yu-Long Jiang ; Ting-Ao Tang ; Ru Huang. Institute of Electrical and Electronics Engineers Inc., 2016. pp. 247-250 (2016 13th IEEE International Conference on Solid-State and Integrated Circuit Technology, ICSICT 2016 - Proceedings).

@inproceedings{c2ba5ee3347143c7a5b7185b444ec694,

title = "Weight tuning of resistive memories and convolution kernel operation on cross-point array for neuro-inspired computing",

abstract = "Analog conductance of resistive memories is attractive for implementing the weights in neuro-inspired algorithms. One of the most popular deep learning algorithms is the convolutional neural network (CNN). In this paper, we review our recent progress on using resistive memories for neuro-inspired computing. First, we optimized the iterative programming protocol to tune the weights of HfOx based resistive memories by adjusting the pulse amplitude incremental steps, the pulse width incremental steps, and the start voltages. Then, we demonstrated the key operation in the CNN-the convolution kernel on a 12×12 cross-point array. As a proof-of-concept demonstration, we use the offline trained edge filters to detect both horizontal and vertical edges of the 50×50 pixels of a grayscale dog image. The experimental kernel operation matches the simulation results.",

author = "Ligang Gao and Chen, {Pai Yu} and Shimeng Yu",

note = "Publisher Copyright: {\textcopyright} 2016 IEEE.; 13th IEEE International Conference on Solid-State and Integrated Circuit Technology, ICSICT 2016 ; Conference date: 25-10-2016 Through 28-10-2016",

year = "2016",

doi = "10.1109/ICSICT.2016.7998889",

language = "English (US)",

series = "2016 13th IEEE International Conference on Solid-State and Integrated Circuit Technology, ICSICT 2016 - Proceedings",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "247--250",

editor = "Yu-Long Jiang and Ting-Ao Tang and Ru Huang",

booktitle = "2016 13th IEEE International Conference on Solid-State and Integrated Circuit Technology, ICSICT 2016 - Proceedings",

}

TY - GEN

T1 - Weight tuning of resistive memories and convolution kernel operation on cross-point array for neuro-inspired computing

AU - Gao, Ligang

AU - Chen, Pai Yu

AU - Yu, Shimeng

PY - 2016

Y1 - 2016

N2 - Analog conductance of resistive memories is attractive for implementing the weights in neuro-inspired algorithms. One of the most popular deep learning algorithms is the convolutional neural network (CNN). In this paper, we review our recent progress on using resistive memories for neuro-inspired computing. First, we optimized the iterative programming protocol to tune the weights of HfOx based resistive memories by adjusting the pulse amplitude incremental steps, the pulse width incremental steps, and the start voltages. Then, we demonstrated the key operation in the CNN-the convolution kernel on a 12×12 cross-point array. As a proof-of-concept demonstration, we use the offline trained edge filters to detect both horizontal and vertical edges of the 50×50 pixels of a grayscale dog image. The experimental kernel operation matches the simulation results.

AB - Analog conductance of resistive memories is attractive for implementing the weights in neuro-inspired algorithms. One of the most popular deep learning algorithms is the convolutional neural network (CNN). In this paper, we review our recent progress on using resistive memories for neuro-inspired computing. First, we optimized the iterative programming protocol to tune the weights of HfOx based resistive memories by adjusting the pulse amplitude incremental steps, the pulse width incremental steps, and the start voltages. Then, we demonstrated the key operation in the CNN-the convolution kernel on a 12×12 cross-point array. As a proof-of-concept demonstration, we use the offline trained edge filters to detect both horizontal and vertical edges of the 50×50 pixels of a grayscale dog image. The experimental kernel operation matches the simulation results.

UR - http://www.scopus.com/inward/record.url?scp=85028646210&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85028646210&partnerID=8YFLogxK

U2 - 10.1109/ICSICT.2016.7998889

DO - 10.1109/ICSICT.2016.7998889

M3 - Conference contribution

AN - SCOPUS:85028646210

T3 - 2016 13th IEEE International Conference on Solid-State and Integrated Circuit Technology, ICSICT 2016 - Proceedings

SP - 247

EP - 250

BT - 2016 13th IEEE International Conference on Solid-State and Integrated Circuit Technology, ICSICT 2016 - Proceedings

A2 - Jiang, Yu-Long

A2 - Tang, Ting-Ao

A2 - Huang, Ru

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 13th IEEE International Conference on Solid-State and Integrated Circuit Technology, ICSICT 2016

Y2 - 25 October 2016 through 28 October 2016

ER -

Weight tuning of resistive memories and convolution kernel operation on cross-point array for neuro-inspired computing

Abstract

Publication series

Other

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this