Ferroelectric NAND for efficient hardware bayesian neural networks

Abstract The rapid advancement of artificial intelligence has enabled breakthroughs in diverse fields, including autonomous systems and medical diagnostics. However, conventional deterministic neural networks struggle to capture uncertainty, limiting their reliability when handling real-world data,...

Full description

Saved in:
Bibliographic Details
Main Authors: Minsuk Song, Ryun-Han Koo, Jangsaeng Kim, Chang-Hyeon Han, Jiyong Yim, Jonghyun Ko, Sijung Yoo, Duk-hyun Choe, Sangwook Kim, Wonjun Shin, Daewoong Kwon
Format: Article
Language:English
Published: Nature Portfolio 2025-07-01
Series:Nature Communications
Online Access:https://doi.org/10.1038/s41467-025-61980-y
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849331631819587584
author Minsuk Song
Ryun-Han Koo
Jangsaeng Kim
Chang-Hyeon Han
Jiyong Yim
Jonghyun Ko
Sijung Yoo
Duk-hyun Choe
Sangwook Kim
Wonjun Shin
Daewoong Kwon
author_facet Minsuk Song
Ryun-Han Koo
Jangsaeng Kim
Chang-Hyeon Han
Jiyong Yim
Jonghyun Ko
Sijung Yoo
Duk-hyun Choe
Sangwook Kim
Wonjun Shin
Daewoong Kwon
author_sort Minsuk Song
collection DOAJ
description Abstract The rapid advancement of artificial intelligence has enabled breakthroughs in diverse fields, including autonomous systems and medical diagnostics. However, conventional deterministic neural networks struggle to capture uncertainty, limiting their reliability when handling real-world data, which are often noisy, imbalanced, or scarce. Bayesian neural networks address this limitation by representing weights as probabilistic distributions, allowing for natural uncertainty quantification and improved robustness. Despite their advantages, hardware-based implementations face significant challenges due to the difficulty of independently tuning both the mean and variance of weight distributions. Herein, we propose a 3D ferroelectric NAND-based Bayesian neural network system that leverages incremental step pulse programming technology to achieve efficient and scalable probabilistic weight control. The page-level programming capabilities and intrinsic device-to-device variations enable gaussian weight distributions in a single programming step, without structural modifications. By modulating the incremental step pulse programming voltage step, we achieve precise weight distribution control. The proposed system demonstrates successful uncertainty estimation, enhanced energy efficiency, and robustness to external noise for medical images.
format Article
id doaj-art-2a54777f7fe540eaa81087f03973d1fe
institution Kabale University
issn 2041-1723
language English
publishDate 2025-07-01
publisher Nature Portfolio
record_format Article
series Nature Communications
spelling doaj-art-2a54777f7fe540eaa81087f03973d1fe2025-08-20T03:46:28ZengNature PortfolioNature Communications2041-17232025-07-0116111410.1038/s41467-025-61980-yFerroelectric NAND for efficient hardware bayesian neural networksMinsuk Song0Ryun-Han Koo1Jangsaeng Kim2Chang-Hyeon Han3Jiyong Yim4Jonghyun Ko5Sijung Yoo6Duk-hyun Choe7Sangwook Kim8Wonjun Shin9Daewoong Kwon10Department of Nanoscale Semiconductor Engineering, Hanyang UniversityDepartment of Electrical and Computer Engineering and Inter-university Semiconductor Research Center, Seoul National UniversityDepartment of Electronic Engineering, Sogang UniversityDepartment of Electrical Engineering, Hanyang UniversityDepartment of Electrical Engineering, Hanyang UniversityDepartment of Electrical and Computer Engineering and Inter-university Semiconductor Research Center, Seoul National UniversityThin Film Technical Unit, Device Research Center (DRC), Samsung Advanced Institute of TechnologyThin Film Technical Unit, Device Research Center (DRC), Samsung Advanced Institute of TechnologyThin Film Technical Unit, Device Research Center (DRC), Samsung Advanced Institute of TechnologyDepartment of Semiconductor Convergence Engineering, Sungkyunkwan UniversityDepartment of Nanoscale Semiconductor Engineering, Hanyang UniversityAbstract The rapid advancement of artificial intelligence has enabled breakthroughs in diverse fields, including autonomous systems and medical diagnostics. However, conventional deterministic neural networks struggle to capture uncertainty, limiting their reliability when handling real-world data, which are often noisy, imbalanced, or scarce. Bayesian neural networks address this limitation by representing weights as probabilistic distributions, allowing for natural uncertainty quantification and improved robustness. Despite their advantages, hardware-based implementations face significant challenges due to the difficulty of independently tuning both the mean and variance of weight distributions. Herein, we propose a 3D ferroelectric NAND-based Bayesian neural network system that leverages incremental step pulse programming technology to achieve efficient and scalable probabilistic weight control. The page-level programming capabilities and intrinsic device-to-device variations enable gaussian weight distributions in a single programming step, without structural modifications. By modulating the incremental step pulse programming voltage step, we achieve precise weight distribution control. The proposed system demonstrates successful uncertainty estimation, enhanced energy efficiency, and robustness to external noise for medical images.https://doi.org/10.1038/s41467-025-61980-y
spellingShingle Minsuk Song
Ryun-Han Koo
Jangsaeng Kim
Chang-Hyeon Han
Jiyong Yim
Jonghyun Ko
Sijung Yoo
Duk-hyun Choe
Sangwook Kim
Wonjun Shin
Daewoong Kwon
Ferroelectric NAND for efficient hardware bayesian neural networks
Nature Communications
title Ferroelectric NAND for efficient hardware bayesian neural networks
title_full Ferroelectric NAND for efficient hardware bayesian neural networks
title_fullStr Ferroelectric NAND for efficient hardware bayesian neural networks
title_full_unstemmed Ferroelectric NAND for efficient hardware bayesian neural networks
title_short Ferroelectric NAND for efficient hardware bayesian neural networks
title_sort ferroelectric nand for efficient hardware bayesian neural networks
url https://doi.org/10.1038/s41467-025-61980-y
work_keys_str_mv AT minsuksong ferroelectricnandforefficienthardwarebayesianneuralnetworks
AT ryunhankoo ferroelectricnandforefficienthardwarebayesianneuralnetworks
AT jangsaengkim ferroelectricnandforefficienthardwarebayesianneuralnetworks
AT changhyeonhan ferroelectricnandforefficienthardwarebayesianneuralnetworks
AT jiyongyim ferroelectricnandforefficienthardwarebayesianneuralnetworks
AT jonghyunko ferroelectricnandforefficienthardwarebayesianneuralnetworks
AT sijungyoo ferroelectricnandforefficienthardwarebayesianneuralnetworks
AT dukhyunchoe ferroelectricnandforefficienthardwarebayesianneuralnetworks
AT sangwookkim ferroelectricnandforefficienthardwarebayesianneuralnetworks
AT wonjunshin ferroelectricnandforefficienthardwarebayesianneuralnetworks
AT daewoongkwon ferroelectricnandforefficienthardwarebayesianneuralnetworks