Impact of Synaptic Device Variations on Classification Accuracy in a Binarized Neural Network

Abstract Brain-inspired neuromorphic systems (hardware neural networks) are expected to be an energy-efficient computing architecture for solving cognitive tasks, which critically depend on the development of reliable synaptic weight storage (i.e., synaptic device). Although various nanoelectronic d...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autores principales:	Sungho Kim, Hee-Dong Kim, Sung-Jin Choi
Formato:	article
Lenguaje:	EN
Publicado:	Nature Portfolio 2019
Materias:	Medicine R Science Q
Acceso en línea:	https://doaj.org/article/7528dffafed64e5f861963b6b1186d67
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

id	oai:doaj.org-article:7528dffafed64e5f861963b6b1186d67
record_format	dspace
spelling	oai:doaj.org-article:7528dffafed64e5f861963b6b1186d672021-12-02T16:08:17ZImpact of Synaptic Device Variations on Classification Accuracy in a Binarized Neural Network10.1038/s41598-019-51814-52045-2322https://doaj.org/article/7528dffafed64e5f861963b6b1186d672019-10-01T00:00:00Zhttps://doi.org/10.1038/s41598-019-51814-5https://doaj.org/toc/2045-2322Abstract Brain-inspired neuromorphic systems (hardware neural networks) are expected to be an energy-efficient computing architecture for solving cognitive tasks, which critically depend on the development of reliable synaptic weight storage (i.e., synaptic device). Although various nanoelectronic devices have successfully reproduced the learning rules of biological synapses through their internal analog conductance states, the sustainability of such devices is still in doubt due to the variability common to all nanoelectronic devices. Alternatively, a neuromorphic system based on a relatively more reliable digital-type switching device has been recently demonstrated, i.e., a binarized neural network (BNN). The synaptic device is a more mature digital-type switching device, and the training/recognition algorithm developed for the BNN enables the task of facial image classification with a supervised training scheme. Here, we quantitatively investigate the effects of device parameter variations on the classification accuracy; the parameters include the number of weight states (N state ), the weight update margin (ΔG), and the weight update variation (G var ). This analysis demonstrates the feasibility of the BNN and introduces a practical neuromorphic system based on mature, conventional digital device technologies.Sungho KimHee-Dong KimSung-Jin ChoiNature PortfolioarticleMedicineRScienceQENScientific Reports, Vol 9, Iss 1, Pp 1-7 (2019)
institution	DOAJ
collection	DOAJ
language	EN
topic	Medicine R Science Q
spellingShingle	Medicine R Science Q Sungho Kim Hee-Dong Kim Sung-Jin Choi Impact of Synaptic Device Variations on Classification Accuracy in a Binarized Neural Network
description	Abstract Brain-inspired neuromorphic systems (hardware neural networks) are expected to be an energy-efficient computing architecture for solving cognitive tasks, which critically depend on the development of reliable synaptic weight storage (i.e., synaptic device). Although various nanoelectronic devices have successfully reproduced the learning rules of biological synapses through their internal analog conductance states, the sustainability of such devices is still in doubt due to the variability common to all nanoelectronic devices. Alternatively, a neuromorphic system based on a relatively more reliable digital-type switching device has been recently demonstrated, i.e., a binarized neural network (BNN). The synaptic device is a more mature digital-type switching device, and the training/recognition algorithm developed for the BNN enables the task of facial image classification with a supervised training scheme. Here, we quantitatively investigate the effects of device parameter variations on the classification accuracy; the parameters include the number of weight states (N state ), the weight update margin (ΔG), and the weight update variation (G var ). This analysis demonstrates the feasibility of the BNN and introduces a practical neuromorphic system based on mature, conventional digital device technologies.
format	article
author	Sungho Kim Hee-Dong Kim Sung-Jin Choi
author_facet	Sungho Kim Hee-Dong Kim Sung-Jin Choi
author_sort	Sungho Kim
title	Impact of Synaptic Device Variations on Classification Accuracy in a Binarized Neural Network
title_short	Impact of Synaptic Device Variations on Classification Accuracy in a Binarized Neural Network
title_full	Impact of Synaptic Device Variations on Classification Accuracy in a Binarized Neural Network
title_fullStr	Impact of Synaptic Device Variations on Classification Accuracy in a Binarized Neural Network
title_full_unstemmed	Impact of Synaptic Device Variations on Classification Accuracy in a Binarized Neural Network
title_sort	impact of synaptic device variations on classification accuracy in a binarized neural network
publisher	Nature Portfolio
publishDate	2019
url	https://doaj.org/article/7528dffafed64e5f861963b6b1186d67
work_keys_str_mv	AT sunghokim impactofsynapticdevicevariationsonclassificationaccuracyinabinarizedneuralnetwork AT heedongkim impactofsynapticdevicevariationsonclassificationaccuracyinabinarizedneuralnetwork AT sungjinchoi impactofsynapticdevicevariationsonclassificationaccuracyinabinarizedneuralnetwork
_version_	1718384537760694272

Impact of Synaptic Device Variations on Classification Accuracy in a Binarized Neural Network

Ejemplares similares