Fully parallel write/read in resistive synaptic array for accelerating on-chip learning

Ligang Gao, I. Ting Wang, Pai Yu Chen, Sarma Vrudhula, Jae sun Seo, Yu Cao, Tuo Hung Hou, Shimeng Yu

Research output: Contribution to journalArticle

Abstract

A neuro-inspired computing paradigm beyond the von Neumann architecture is emerging and it generally takes advantage of massive parallelism and is aimed at complex tasks that involve intelligence and learning. The cross-point array architecture with synaptic devices has been proposed for on-chip implementation of the weighted sum and weight update in the learning algorithms. In this work, forming-free, silicon-process-compatible Ta/TaOx/TiO2/Ti synaptic devices are fabricated, in which >200 levels of conductance states could be continuously tuned by identical programming pulses. In order to demonstrate the advantages of parallelism of the cross-point array architecture, a novel fully parallel write scheme is designed and experimentally demonstrated in a small-scale crossbar array to accelerate the weight update in the training process, at a speed that is independent of the array size. Compared to the conventional row-by-row write scheme, it achieves >30× speed-up and >30× improvement in energy efficiency as projected in a large-scale array. If realistic synaptic device characteristics such as device variations are taken into an array-level simulation, the proposed array architecture is able to achieve ∼95% recognition accuracy of MNIST handwritten digits, which is close to the accuracy achieved by software using the ideal sparse coding algorithm.

Original languageEnglish (US)
Pages (from-to)455204
Number of pages1
JournalNanotechnology
Volume26
Issue number45
StatePublished - Nov 13 2015
Externally publishedYes

Fingerprint

Learning
Equipment and Supplies
Silicon
Learning algorithms
Energy efficiency
Weights and Measures
Intelligence
Software
Efficiency
Recognition (Psychology)

ASJC Scopus subject areas

  • Medicine(all)

Cite this

Fully parallel write/read in resistive synaptic array for accelerating on-chip learning. / Gao, Ligang; Wang, I. Ting; Chen, Pai Yu; Vrudhula, Sarma; Seo, Jae sun; Cao, Yu; Hou, Tuo Hung; Yu, Shimeng.

In: Nanotechnology, Vol. 26, No. 45, 13.11.2015, p. 455204.

Research output: Contribution to journalArticle

Gao, Ligang ; Wang, I. Ting ; Chen, Pai Yu ; Vrudhula, Sarma ; Seo, Jae sun ; Cao, Yu ; Hou, Tuo Hung ; Yu, Shimeng. / Fully parallel write/read in resistive synaptic array for accelerating on-chip learning. In: Nanotechnology. 2015 ; Vol. 26, No. 45. pp. 455204.
@article{d75cc2e57056447c8a0759742c211727,
title = "Fully parallel write/read in resistive synaptic array for accelerating on-chip learning",
abstract = "A neuro-inspired computing paradigm beyond the von Neumann architecture is emerging and it generally takes advantage of massive parallelism and is aimed at complex tasks that involve intelligence and learning. The cross-point array architecture with synaptic devices has been proposed for on-chip implementation of the weighted sum and weight update in the learning algorithms. In this work, forming-free, silicon-process-compatible Ta/TaOx/TiO2/Ti synaptic devices are fabricated, in which >200 levels of conductance states could be continuously tuned by identical programming pulses. In order to demonstrate the advantages of parallelism of the cross-point array architecture, a novel fully parallel write scheme is designed and experimentally demonstrated in a small-scale crossbar array to accelerate the weight update in the training process, at a speed that is independent of the array size. Compared to the conventional row-by-row write scheme, it achieves >30× speed-up and >30× improvement in energy efficiency as projected in a large-scale array. If realistic synaptic device characteristics such as device variations are taken into an array-level simulation, the proposed array architecture is able to achieve ∼95{\%} recognition accuracy of MNIST handwritten digits, which is close to the accuracy achieved by software using the ideal sparse coding algorithm.",
author = "Ligang Gao and Wang, {I. Ting} and Chen, {Pai Yu} and Sarma Vrudhula and Seo, {Jae sun} and Yu Cao and Hou, {Tuo Hung} and Shimeng Yu",
year = "2015",
month = "11",
day = "13",
language = "English (US)",
volume = "26",
pages = "455204",
journal = "Nanotechnology",
issn = "0957-4484",
publisher = "IOP Publishing Ltd.",
number = "45",

}

TY - JOUR

T1 - Fully parallel write/read in resistive synaptic array for accelerating on-chip learning

AU - Gao, Ligang

AU - Wang, I. Ting

AU - Chen, Pai Yu

AU - Vrudhula, Sarma

AU - Seo, Jae sun

AU - Cao, Yu

AU - Hou, Tuo Hung

AU - Yu, Shimeng

PY - 2015/11/13

Y1 - 2015/11/13

N2 - A neuro-inspired computing paradigm beyond the von Neumann architecture is emerging and it generally takes advantage of massive parallelism and is aimed at complex tasks that involve intelligence and learning. The cross-point array architecture with synaptic devices has been proposed for on-chip implementation of the weighted sum and weight update in the learning algorithms. In this work, forming-free, silicon-process-compatible Ta/TaOx/TiO2/Ti synaptic devices are fabricated, in which >200 levels of conductance states could be continuously tuned by identical programming pulses. In order to demonstrate the advantages of parallelism of the cross-point array architecture, a novel fully parallel write scheme is designed and experimentally demonstrated in a small-scale crossbar array to accelerate the weight update in the training process, at a speed that is independent of the array size. Compared to the conventional row-by-row write scheme, it achieves >30× speed-up and >30× improvement in energy efficiency as projected in a large-scale array. If realistic synaptic device characteristics such as device variations are taken into an array-level simulation, the proposed array architecture is able to achieve ∼95% recognition accuracy of MNIST handwritten digits, which is close to the accuracy achieved by software using the ideal sparse coding algorithm.

AB - A neuro-inspired computing paradigm beyond the von Neumann architecture is emerging and it generally takes advantage of massive parallelism and is aimed at complex tasks that involve intelligence and learning. The cross-point array architecture with synaptic devices has been proposed for on-chip implementation of the weighted sum and weight update in the learning algorithms. In this work, forming-free, silicon-process-compatible Ta/TaOx/TiO2/Ti synaptic devices are fabricated, in which >200 levels of conductance states could be continuously tuned by identical programming pulses. In order to demonstrate the advantages of parallelism of the cross-point array architecture, a novel fully parallel write scheme is designed and experimentally demonstrated in a small-scale crossbar array to accelerate the weight update in the training process, at a speed that is independent of the array size. Compared to the conventional row-by-row write scheme, it achieves >30× speed-up and >30× improvement in energy efficiency as projected in a large-scale array. If realistic synaptic device characteristics such as device variations are taken into an array-level simulation, the proposed array architecture is able to achieve ∼95% recognition accuracy of MNIST handwritten digits, which is close to the accuracy achieved by software using the ideal sparse coding algorithm.

UR - http://www.scopus.com/inward/record.url?scp=84979796621&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84979796621&partnerID=8YFLogxK

M3 - Article

VL - 26

SP - 455204

JO - Nanotechnology

JF - Nanotechnology

SN - 0957-4484

IS - 45

ER -