A Highly Robust Binary Neural Network Inference Accelerator Based on Binary Memristors

Since memristor was found, it has shown great application potential in neuromorphic computing. Currently, most neural networks based on memristors deploy the special analog characteristics of memristor. However, owing to the limitation of manufacturing process, non-ideal characteristics such as non-...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autores principales:	Yiyang Zhao, Yongjia Wang, Ruibo Wang, Yuan Rong, Xianyang Jiang
Formato:	article
Lenguaje:	EN
Publicado:	MDPI AG 2021
Materias:	memristor binary neural network (BNN) non-ideal factors Verilog-AMS Electronics TK7800-8360
Acceso en línea:	https://doaj.org/article/950bc862b0af4d5fae37b64127904c44
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

id	oai:doaj.org-article:950bc862b0af4d5fae37b64127904c44
record_format	dspace
spelling	oai:doaj.org-article:950bc862b0af4d5fae37b64127904c442021-11-11T15:37:36ZA Highly Robust Binary Neural Network Inference Accelerator Based on Binary Memristors10.3390/electronics102126002079-9292https://doaj.org/article/950bc862b0af4d5fae37b64127904c442021-10-01T00:00:00Zhttps://www.mdpi.com/2079-9292/10/21/2600https://doaj.org/toc/2079-9292Since memristor was found, it has shown great application potential in neuromorphic computing. Currently, most neural networks based on memristors deploy the special analog characteristics of memristor. However, owing to the limitation of manufacturing process, non-ideal characteristics such as non-linearity, asymmetry, and inconsistent device periodicity appear frequently and definitely, therefore, it is a challenge to employ memristor in a massive way. On the contrary, a binary neural network (BNN) requires its weights to be either +1 or −1, which can be mapped by digital memristors with high technical maturity. Upon this, a highly robust BNN inference accelerator with binary sigmoid activation function is proposed. In the accelerator, the inputs of each network layer are either +1 or 0, which can facilitate feature encoding and reduce the peripheral circuit complexity of memristor hardware. The proposed two-column reference memristor structure together with current controlled voltage source (CCVS) circuit not only solves the problem of mapping positive and negative weights on memristor array, but also eliminates the sneak current effect under the minimum conductance status. Being compared to the traditional differential pair structure of BNN, the proposed two-column reference scheme can reduce both the number of memristors and the latency to refresh the memristor array by nearly 50%. The influence of non-ideal factors of memristor array such as memristor array yield, memristor conductance fluctuation, and reading noise on the accuracy of BNN is investigated in detail based on a newly memristor circuit model with non-ideal characteristics. The experimental results demonstrate that when the array yield <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mi>α</mi></semantics></math></inline-formula> ≥ 5%, or the reading noise <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mi>σ</mi></semantics></math></inline-formula> ≤ 0.25, a recognition accuracy greater than 97% on the MNIST data set is achieved.Yiyang ZhaoYongjia WangRuibo WangYuan RongXianyang JiangMDPI AGarticlememristorbinary neural network (BNN)non-ideal factorsVerilog-AMSElectronicsTK7800-8360ENElectronics, Vol 10, Iss 2600, p 2600 (2021)
institution	DOAJ
collection	DOAJ
language	EN
topic	memristor binary neural network (BNN) non-ideal factors Verilog-AMS Electronics TK7800-8360
spellingShingle	memristor binary neural network (BNN) non-ideal factors Verilog-AMS Electronics TK7800-8360 Yiyang Zhao Yongjia Wang Ruibo Wang Yuan Rong Xianyang Jiang A Highly Robust Binary Neural Network Inference Accelerator Based on Binary Memristors
description	Since memristor was found, it has shown great application potential in neuromorphic computing. Currently, most neural networks based on memristors deploy the special analog characteristics of memristor. However, owing to the limitation of manufacturing process, non-ideal characteristics such as non-linearity, asymmetry, and inconsistent device periodicity appear frequently and definitely, therefore, it is a challenge to employ memristor in a massive way. On the contrary, a binary neural network (BNN) requires its weights to be either +1 or −1, which can be mapped by digital memristors with high technical maturity. Upon this, a highly robust BNN inference accelerator with binary sigmoid activation function is proposed. In the accelerator, the inputs of each network layer are either +1 or 0, which can facilitate feature encoding and reduce the peripheral circuit complexity of memristor hardware. The proposed two-column reference memristor structure together with current controlled voltage source (CCVS) circuit not only solves the problem of mapping positive and negative weights on memristor array, but also eliminates the sneak current effect under the minimum conductance status. Being compared to the traditional differential pair structure of BNN, the proposed two-column reference scheme can reduce both the number of memristors and the latency to refresh the memristor array by nearly 50%. The influence of non-ideal factors of memristor array such as memristor array yield, memristor conductance fluctuation, and reading noise on the accuracy of BNN is investigated in detail based on a newly memristor circuit model with non-ideal characteristics. The experimental results demonstrate that when the array yield <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mi>α</mi></semantics></math></inline-formula> ≥ 5%, or the reading noise <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mi>σ</mi></semantics></math></inline-formula> ≤ 0.25, a recognition accuracy greater than 97% on the MNIST data set is achieved.
format	article
author	Yiyang Zhao Yongjia Wang Ruibo Wang Yuan Rong Xianyang Jiang
author_facet	Yiyang Zhao Yongjia Wang Ruibo Wang Yuan Rong Xianyang Jiang
author_sort	Yiyang Zhao
title	A Highly Robust Binary Neural Network Inference Accelerator Based on Binary Memristors
title_short	A Highly Robust Binary Neural Network Inference Accelerator Based on Binary Memristors
title_full	A Highly Robust Binary Neural Network Inference Accelerator Based on Binary Memristors
title_fullStr	A Highly Robust Binary Neural Network Inference Accelerator Based on Binary Memristors
title_full_unstemmed	A Highly Robust Binary Neural Network Inference Accelerator Based on Binary Memristors
title_sort	highly robust binary neural network inference accelerator based on binary memristors
publisher	MDPI AG
publishDate	2021
url	https://doaj.org/article/950bc862b0af4d5fae37b64127904c44
work_keys_str_mv	AT yiyangzhao ahighlyrobustbinaryneuralnetworkinferenceacceleratorbasedonbinarymemristors AT yongjiawang ahighlyrobustbinaryneuralnetworkinferenceacceleratorbasedonbinarymemristors AT ruibowang ahighlyrobustbinaryneuralnetworkinferenceacceleratorbasedonbinarymemristors AT yuanrong ahighlyrobustbinaryneuralnetworkinferenceacceleratorbasedonbinarymemristors AT xianyangjiang ahighlyrobustbinaryneuralnetworkinferenceacceleratorbasedonbinarymemristors AT yiyangzhao highlyrobustbinaryneuralnetworkinferenceacceleratorbasedonbinarymemristors AT yongjiawang highlyrobustbinaryneuralnetworkinferenceacceleratorbasedonbinarymemristors AT ruibowang highlyrobustbinaryneuralnetworkinferenceacceleratorbasedonbinarymemristors AT yuanrong highlyrobustbinaryneuralnetworkinferenceacceleratorbasedonbinarymemristors AT xianyangjiang highlyrobustbinaryneuralnetworkinferenceacceleratorbasedonbinarymemristors
_version_	1718434872704368640

A Highly Robust Binary Neural Network Inference Accelerator Based on Binary Memristors

Ejemplares similares