High-Precision Symmetric Weight Update of Memristor by Gate Voltage Ramping Method for Convolutional Neural Network Accelerator

被引：37

作者：

Chen, Jia ^{[1
]}

Pan, Wen-Qian ^{[1
]}

Li, Yi ^{[1
]}

Kuang, Rui ^{[1
]}

He, Yu-Hui ^{[1
]}

Lin, Chih-Yang ^{[2
]}

Duan, Nian ^{[1
]}

Feng, Gui-Rong ^{[1
]}

Zheng, Hao-Xuan ^{[2
]}

Chang, Ting-Chang ^{[2
]}

Sze, Simon M. ^{[3
]}

Miao, Xiang-Shui ^{[1
]}

机构：

[1] Huazhong Univ Sci & Technol, Sch Opt & Elect Informat, Wuhan Natl Lab Optoelect, Wuhan 430074, Peoples R China

[2] Natl Sun Yat Sen Univ, Dept Phys, Kaohsiung 80424, Taiwan

[3] Natl Chiao Tung Univ, Dept Elect Engn, Hsinchu 300, Taiwan

来源：

IEEE ELECTRON DEVICE LETTERS | 2020年 / 41卷 / 03期

基金：

中国国家自然科学基金;

关键词：

Memristor; symmetric weight update; convolutional neural network; MEMORY; CLASSIFICATION;

D O I：

10.1109/LED.2020.2968388

中图分类号：

TM [电工技术]; TN [电子技术、通信技术];

学科分类号：

0808 ; 0809 ;

摘要：

Memristor emerges as the key enabler for neural network accelerator. Here, we demonstrate high-precision symmetric weight update in a one transistor one resistor (1T1R) structure Ti/HfO2/TiN memristor using a gate voltage ramping method, with over 120-level states and low variation (< 4%). Incorporating all experimental non-idealities, the proposed mixed hardware-software convolutional neural network demonstrates over 92.79% online learning accuracy (against software equivalent 98.45%) for MNIST recognition task. The network also shows robustness to input image noises, array yield, and retention issues.

引用

页码：353 / 356

页数：4

共 30 条

[1]

Agarwal S, 2016, IEEE IJCNN, P929, DOI 10.1109/IJCNN.2016.7727298

[2] Equivalent-accuracy accelerated neural-network training using analogue memory [J].

Ambrogio, Stefano ;

Narayanan, Pritish ;

Tsai, Hsinyu ;

Shelby, Robert M. ;

Boybat, Irem ;

di Nolfo, Carmelo ;

Sidler, Severin ;

Giordano, Massimo ;

Bodini, Martina ;

Farinha, Nathan C. P. ;

Killeen, Benjamin ;

Cheng, Christina ;

Jaoudi, Yassine ;

Burr, Geoffrey W. .

NATURE, 2018, 558 (7708) :60-+

[3] Neuromorphic Learning and Recognition With One-Transistor-One-Resistor Synapses and Bistable Metal Oxide RRAM [J].

Ambrogio, Stefano ;

Balatti, Simone ;

Milo, Valerio ;

Carboni, Roberto ;

Wang, Zhong-Qiang ;

Calderoni, Alessandro ;

Ramaswamy, Nirmal ;

Ielmini, Daniele .

IEEE TRANSACTIONS ON ELECTRON DEVICES, 2016, 63 (04) :1508-1515

[4] Implementation of multilayer perceptron network with highly uniform passive memristive crossbar circuits [J].

Bayat, F. Merrikh ;

Prezioso, M. ;

Chakrabarti, B. ;

Nili, H. ;

Kataeva, I. ;

Strukov, D. .

NATURE COMMUNICATIONS, 2018, 9

[5]

Chang CC, 2017, INT EL DEVICES MEET

[6] LiSiOX-Based Analog Memristive Synapse for Neuromorphic Computing [J].

Chen, Jia ;

Lin, Chih-Yang ;

Li, Yi ;

Qin, Chao ;

Lu, Ke ;

Wang, Jie-Ming ;

Chen, Chun-Kuei ;

He, Yu-Hui ;

Chang, Ting-Chang ;

Sze, Simon M. ;

Miao, Xiang-Shui .

IEEE ELECTRON DEVICE LETTERS, 2019, 40 (04) :542-545

[7] SiGe epitaxial memory for neuromorphic computing with reproducible high performance based on engineered dislocations [J].

Choi, Shinhyun ;

Tan, Scott H. ;

Li, Zefan ;

Kim, Yunjo ;

Choi, Chanyeol ;

Chen, Pai-Yu ;

Yeon, Hanwool ;

Yu, Shimeng ;

Kim, Jeehwan .

NATURE MATERIALS, 2018, 17 (04) :335-+

[8] Perspective on training fully connected networks with resistive memories: Device requirements for multiple conductances of varying significance [J].

Cristiano, Giorgio ;

Giordano, Massimo ;

Ambrogio, Stefano ;

Romero, Louis P. ;

Cheng, Christina ;

Narayanan, Pritish ;

Tsai, Hsinyu ;

Shelby, Robert M. ;

Burr, Geoffrey W. .

JOURNAL OF APPLIED PHYSICS, 2018, 124 (15)

[9] Light-tuned selective photosynthesis of azo- and azoxy-aromatics using graphitic C3N4 [J].

Dai, Yitao ;

Li, Chao ;

Shen, Yanbin ;

Lim, Tingbin ;

Xu, Jian ;

Li, Yongwang ;

Niemantsverdriet, Hans ;

Besenbacher, Flemming ;

Lock, Nina ;

Su, Ren .

NATURE COMMUNICATIONS, 2018, 9

[10] Demonstration of Convolution Kernel Operation on Resistive Cross-Point Array [J].

Gao, Ligang ;

Chen, Pai-Yu ;

Yu, Shimeng .

IEEE ELECTRON DEVICE LETTERS, 2016, 37 (07) :870-873

← 1 2 3 →