Impact of Selector Devices in Analog RRAM-Based Crossbar Arrays for Inference and Training of Neuromorphic System

Jiyong Woo, Shimeng Yu

Research output: Contribution to journalArticle

Abstract

The impact of selector devices on the inference and training accuracy of a resistive random access memory (RRAM)-based neuromorphic computing system is rarely studied. In this paper, we analyze the weighted sum and weight update functions in a one-selector-one-RRAM (1S-1R)-based crossbar arrays. We first develop a Verilog-A model based on the lateral evolution of the filament to describe analog conductance tuning in the filamentary RRAM. We then perform an array-level SPICE simulation on the 1S-1R arrays, where the exponential and threshold selectors are employed. In the inference stage, the read-out current is vulnerable to the inevitable IR drop caused by the wire resistance. Our finding reveals that the use of a threshold selector allows the 1S-1R device to have a linear I-V relation, improving the immunity to the IR drop. On the other hand, the threshold selector distorts analog RRAM's linear weight update during the training stage. Instead, an introduction of exponential selector enables the desirable properties of analog RRAM to be maintained even in the 1S-1R device. These results indicate that different selectors suitable for each operation mode (inference or training) are preferred in the neuromorphic computing system.

Original languageEnglish (US)
Article number8734798
Pages (from-to)2205-2212
Number of pages8
JournalIEEE Transactions on Very Large Scale Integration (VLSI) Systems
Volume27
Issue number9
DOIs
StatePublished - Sep 1 2019

Fingerprint

Data storage equipment
Computer hardware description languages
SPICE
Tuning
Wire
RRAM

Keywords

  • Crossbar array
  • neuromorphic computing
  • resistive random access memory (RRAM)
  • selector device
  • synaptic device
  • weight update
  • weighted sum

ASJC Scopus subject areas

  • Software
  • Hardware and Architecture
  • Electrical and Electronic Engineering

Cite this

Impact of Selector Devices in Analog RRAM-Based Crossbar Arrays for Inference and Training of Neuromorphic System. / Woo, Jiyong; Yu, Shimeng.

In: IEEE Transactions on Very Large Scale Integration (VLSI) Systems, Vol. 27, No. 9, 8734798, 01.09.2019, p. 2205-2212.

Research output: Contribution to journalArticle

@article{0be438927657433aacff16db5aa95786,
title = "Impact of Selector Devices in Analog RRAM-Based Crossbar Arrays for Inference and Training of Neuromorphic System",
abstract = "The impact of selector devices on the inference and training accuracy of a resistive random access memory (RRAM)-based neuromorphic computing system is rarely studied. In this paper, we analyze the weighted sum and weight update functions in a one-selector-one-RRAM (1S-1R)-based crossbar arrays. We first develop a Verilog-A model based on the lateral evolution of the filament to describe analog conductance tuning in the filamentary RRAM. We then perform an array-level SPICE simulation on the 1S-1R arrays, where the exponential and threshold selectors are employed. In the inference stage, the read-out current is vulnerable to the inevitable IR drop caused by the wire resistance. Our finding reveals that the use of a threshold selector allows the 1S-1R device to have a linear I-V relation, improving the immunity to the IR drop. On the other hand, the threshold selector distorts analog RRAM's linear weight update during the training stage. Instead, an introduction of exponential selector enables the desirable properties of analog RRAM to be maintained even in the 1S-1R device. These results indicate that different selectors suitable for each operation mode (inference or training) are preferred in the neuromorphic computing system.",
keywords = "Crossbar array, neuromorphic computing, resistive random access memory (RRAM), selector device, synaptic device, weight update, weighted sum",
author = "Jiyong Woo and Shimeng Yu",
year = "2019",
month = "9",
day = "1",
doi = "10.1109/TVLSI.2019.2917764",
language = "English (US)",
volume = "27",
pages = "2205--2212",
journal = "IEEE Transactions on Very Large Scale Integration (VLSI) Systems",
issn = "1063-8210",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "9",

}

TY - JOUR

T1 - Impact of Selector Devices in Analog RRAM-Based Crossbar Arrays for Inference and Training of Neuromorphic System

AU - Woo, Jiyong

AU - Yu, Shimeng

PY - 2019/9/1

Y1 - 2019/9/1

N2 - The impact of selector devices on the inference and training accuracy of a resistive random access memory (RRAM)-based neuromorphic computing system is rarely studied. In this paper, we analyze the weighted sum and weight update functions in a one-selector-one-RRAM (1S-1R)-based crossbar arrays. We first develop a Verilog-A model based on the lateral evolution of the filament to describe analog conductance tuning in the filamentary RRAM. We then perform an array-level SPICE simulation on the 1S-1R arrays, where the exponential and threshold selectors are employed. In the inference stage, the read-out current is vulnerable to the inevitable IR drop caused by the wire resistance. Our finding reveals that the use of a threshold selector allows the 1S-1R device to have a linear I-V relation, improving the immunity to the IR drop. On the other hand, the threshold selector distorts analog RRAM's linear weight update during the training stage. Instead, an introduction of exponential selector enables the desirable properties of analog RRAM to be maintained even in the 1S-1R device. These results indicate that different selectors suitable for each operation mode (inference or training) are preferred in the neuromorphic computing system.

AB - The impact of selector devices on the inference and training accuracy of a resistive random access memory (RRAM)-based neuromorphic computing system is rarely studied. In this paper, we analyze the weighted sum and weight update functions in a one-selector-one-RRAM (1S-1R)-based crossbar arrays. We first develop a Verilog-A model based on the lateral evolution of the filament to describe analog conductance tuning in the filamentary RRAM. We then perform an array-level SPICE simulation on the 1S-1R arrays, where the exponential and threshold selectors are employed. In the inference stage, the read-out current is vulnerable to the inevitable IR drop caused by the wire resistance. Our finding reveals that the use of a threshold selector allows the 1S-1R device to have a linear I-V relation, improving the immunity to the IR drop. On the other hand, the threshold selector distorts analog RRAM's linear weight update during the training stage. Instead, an introduction of exponential selector enables the desirable properties of analog RRAM to be maintained even in the 1S-1R device. These results indicate that different selectors suitable for each operation mode (inference or training) are preferred in the neuromorphic computing system.

KW - Crossbar array

KW - neuromorphic computing

KW - resistive random access memory (RRAM)

KW - selector device

KW - synaptic device

KW - weight update

KW - weighted sum

UR - http://www.scopus.com/inward/record.url?scp=85071363070&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85071363070&partnerID=8YFLogxK

U2 - 10.1109/TVLSI.2019.2917764

DO - 10.1109/TVLSI.2019.2917764

M3 - Article

VL - 27

SP - 2205

EP - 2212

JO - IEEE Transactions on Very Large Scale Integration (VLSI) Systems

JF - IEEE Transactions on Very Large Scale Integration (VLSI) Systems

SN - 1063-8210

IS - 9

M1 - 8734798

ER -