Super-steep synapses based on positive feedback devices for reliable binary neural networks

Dongseok Kwon; Hyeongsu Kim; Kyu Ho Lee; Joon Hwang; Wonjun Shin; Jong Ho Bae; Sung Yun Woo; Jong Ho Lee

doi:10.1063/5.0131235

Super-steep synapses based on positive feedback devices for reliable binary neural networks

Dongseok Kwon, Hyeongsu Kim, Kyu Ho Lee, Joon Hwang, Wonjun Shin, Jong Ho Bae, Sung Yun Woo, Jong Ho Lee

Research output: Contribution to journal › Article › peer-review

2 Scopus citations

Abstract

This work proposes positive feedback (PF) device-based synaptic devices for reliable binary neural networks (BNNs). Due to PF operation, the fabricated PF device shows a high on/off current ratio (2.69 × 107). The PF device has a charge-trap layer by which the turn-on voltage (Von) of the device can be adjusted by program/erase operations and a long-term memory function is implemented. Also, due to the steep switching characteristics of the PF device, the conductance becomes tolerant to the retention time and the variation in turn-on voltage. Simulations show that high accuracy (88.44% for CIFAR-10 image classification) can be achieved in hardware-based BNNs using PF devices with these properties as synapses.

Original language	English
Article number	102101
Journal	Applied Physics Letters
Volume	122
Issue number	10
DOIs	https://doi.org/10.1063/5.0131235
State	Published - 6 Mar 2023

Access to Document

10.1063/5.0131235

Cite this

@article{cc92d65802634ae7bf4b5917db24e784,

title = "Super-steep synapses based on positive feedback devices for reliable binary neural networks",

abstract = "This work proposes positive feedback (PF) device-based synaptic devices for reliable binary neural networks (BNNs). Due to PF operation, the fabricated PF device shows a high on/off current ratio (2.69 × 107). The PF device has a charge-trap layer by which the turn-on voltage (Von) of the device can be adjusted by program/erase operations and a long-term memory function is implemented. Also, due to the steep switching characteristics of the PF device, the conductance becomes tolerant to the retention time and the variation in turn-on voltage. Simulations show that high accuracy (88.44% for CIFAR-10 image classification) can be achieved in hardware-based BNNs using PF devices with these properties as synapses.",

author = "Dongseok Kwon and Hyeongsu Kim and Lee, {Kyu Ho} and Joon Hwang and Wonjun Shin and Bae, {Jong Ho} and Woo, {Sung Yun} and Lee, {Jong Ho}",

note = "Publisher Copyright: {\textcopyright} 2023 Author(s).",

year = "2023",

month = mar,

day = "6",

doi = "10.1063/5.0131235",

language = "English",

volume = "122",

journal = "Applied Physics Letters",

issn = "0003-6951",

publisher = "American Institute of Physics",

number = "10",

}

TY - JOUR

T1 - Super-steep synapses based on positive feedback devices for reliable binary neural networks

AU - Kwon, Dongseok

AU - Kim, Hyeongsu

AU - Lee, Kyu Ho

AU - Hwang, Joon

AU - Shin, Wonjun

AU - Bae, Jong Ho

AU - Woo, Sung Yun

AU - Lee, Jong Ho

PY - 2023/3/6

Y1 - 2023/3/6

N2 - This work proposes positive feedback (PF) device-based synaptic devices for reliable binary neural networks (BNNs). Due to PF operation, the fabricated PF device shows a high on/off current ratio (2.69 × 107). The PF device has a charge-trap layer by which the turn-on voltage (Von) of the device can be adjusted by program/erase operations and a long-term memory function is implemented. Also, due to the steep switching characteristics of the PF device, the conductance becomes tolerant to the retention time and the variation in turn-on voltage. Simulations show that high accuracy (88.44% for CIFAR-10 image classification) can be achieved in hardware-based BNNs using PF devices with these properties as synapses.

AB - This work proposes positive feedback (PF) device-based synaptic devices for reliable binary neural networks (BNNs). Due to PF operation, the fabricated PF device shows a high on/off current ratio (2.69 × 107). The PF device has a charge-trap layer by which the turn-on voltage (Von) of the device can be adjusted by program/erase operations and a long-term memory function is implemented. Also, due to the steep switching characteristics of the PF device, the conductance becomes tolerant to the retention time and the variation in turn-on voltage. Simulations show that high accuracy (88.44% for CIFAR-10 image classification) can be achieved in hardware-based BNNs using PF devices with these properties as synapses.

UR - http://www.scopus.com/inward/record.url?scp=85149852173&partnerID=8YFLogxK

U2 - 10.1063/5.0131235

DO - 10.1063/5.0131235

M3 - Article

AN - SCOPUS:85149852173

SN - 0003-6951

VL - 122

JO - Applied Physics Letters

JF - Applied Physics Letters

IS - 10

M1 - 102101

ER -

Super-steep synapses based on positive feedback devices for reliable binary neural networks

Abstract

Access to Document

Other files and links

Fingerprint

Cite this