Article
Computer Science, Information Systems
Min Suk Song, Hwiho Hwang, Geun Ho Lee, Suhyeon Ahn, Sungmin Hwang, Hyungjin Kim
Summary: In this study, we fabricated a flash memory device based on a 3D NAND flash architecture and used it as a synaptic device in a neuromorphic system. We proposed two kernel mapping methods for convolutional neural networks and verified their effectiveness through simulation. Finally, we performed off-chip learning using the MNIST dataset and compared the two schemes.
Article
Engineering, Electrical & Electronic
Gihun Choe, Wonbo Shim, Panni Wang, Jae Hur, Asif Islam Khan, Shimeng Yu
Summary: The study examines the use of 3-D NAND architecture based on FeFETs for in-memory computing, investigating the impact of phase distribution in ferroelectric Hafnia-based thin films on read-out current variation and optimizing bias conditions through TCAD simulations. The analysis further explores the array-level impact of phase variation on vector-matrix multiplication using SPICE simulations, confirming sufficient read-out accuracy for analog-to-digital conversion.
IEEE TRANSACTIONS ON ELECTRON DEVICES
(2021)
Article
Engineering, Electrical & Electronic
Geun Ho Lee, Min Suk Song, Sangwoo Kim, Jiyong Yim, Sungmin Hwang, Junsu Yu, Daewoong Kwon, Hyungjin Kim
Summary: FeFET, as a nonvolatile memory device, can be used as a synaptic device in neuromorphic systems. Utilizing HZO material and the 3-D nand structure, it has been verified and simulated, with a proposed method to increase synaptic cell efficiency.
IEEE TRANSACTIONS ON ELECTRON DEVICES
(2022)
Article
Chemistry, Multidisciplinary
Sungmin Hwang, Junsu Yu, Min Suk Song, Hwiho Hwang, Hyungjin Kim
Summary: The progress of artificial intelligence and large-scale neural networks has led to increased computational costs and energy consumption. Researchers are exploring low-power neural network implementation approaches, and neuromorphic computing systems are being considered as potential candidates. Synaptic devices are crucial for neuromorphic systems, and this study presents an 8 x 16 memcapacitor crossbar array for neural network implementation. The array demonstrates high reliability and successfully performs vector-matrix multiplication with low error. Furthermore, a spiking neural network for CIFAR-10 classification is implemented with the capability of weight fine-tuning, achieving a high level of accuracy.
Article
Engineering, Electrical & Electronic
Prabhat Kumar Dubey, Sebastiano Strangio, Enrique G. Marin, Giuseppe Iannaccone, Gianluca Fiori
Summary: We introduce a charge-based Verilog-A model for 2DM-based FETs in neuromorphic circuit design. The model combines explicit solution of drift-diffusion transport and electrostatics, including Fermi-Dirac statistics. Terminal charges and capacitance calculations are done using the Ward-Dutton linear charge partitioning scheme. The model accurately predicts the electrical behavior of MoS2 FETs and is used to simulate various neuromorphic circuit building blocks.
IEEE TRANSACTIONS ON ELECTRON DEVICES
(2023)
Article
Engineering, Electrical & Electronic
Seungmin Lee, Joonsung Lim, Jun Hyoung Kim, Sunghwan Cho, Yong Kyu Lee, Byoungdeog Choi
Summary: This study proposes a novel cell array structure suitable for hybrid bonding technology in 3-D NAND architecture, allowing for the removal of dummy cell area and increase in bit density. The proposed method for programming this structure utilizes the asymmetric GIDL phenomenon for string selection and inhibition, potentially improving V-PASS disturbance.
IEEE TRANSACTIONS ON ELECTRON DEVICES
(2023)
Article
Chemistry, Analytical
Haichun Zhang, Jie Wang, Zhuo Chen, Yuqian Pan, Zhaojun Lu, Zhenglin Liu
Summary: NAND flash memory is widely used in various applications due to its advantages, but reliability is decreasing as technology advances. Machine learning algorithms can be used to predict endurance levels and optimize wear-leveling strategies, which is crucial for extending device lifetimes.
Article
Computer Science, Artificial Intelligence
Hassan Eshkiki, Benjamin Mora, Xianghua Xie
Summary: This article proposes the Mediterranean matrix multiplication algorithm, which approximates matrix multiplication by sampling angles and achieves a simplified processing architecture and compressed matrix weights. The method outperforms the standard approximation in terms of size and number of operations, and has potential applications in machine learning inference.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Nanoscience & Nanotechnology
Frank Brueckerhoff-Plueckelmann, Ivonne Bente, Daniel Wendland, Johannes Feldmann, C. David Wright, Harish Bhaskaran, Wolfram Pernice
Summary: Integrated neuromorphic photonic circuits aim to power complex artificial neural networks in an energy and time efficient way. Scaling photonic circuits to match the requirements of modern ANNs remains challenging. A proposed time multiplexed matrix processing scheme virtually increases the size of a physical photonic crossbar array without requiring additional electrical post-processing, enabling high-speed and high-accuracy matrix vector multiplications.
Article
Computer Science, Information Systems
Chanyang Ju, Hyeonbum Lee, Heewon Chung, Jae Hong Seo, Sungwook Kim
Summary: The paper focuses on verifying the accuracy of CNNs in image recognition and classification, proposing a predicate function based on validating matrix multiplication operations. By reducing the proving cost, an efficient sum-check protocol is provided for convolution operations, which is approximately 2x cheaper in terms of communication costs compared to the state-of-the-art zkCNN approach.
Article
Computer Science, Hardware & Architecture
Jaehun Jang, Jong Hwan Ko
Summary: This paper proposes a method to reduce state errors when storing neural network weights in Flash memory. By applying weight-bit inversion for state elimination and state error reduction, accuracy and read speed can be improved. Furthermore, an adaptive weight-bit inversion scheme is used to selectively operate on units of weight groups, further improving read speed.
JOURNAL OF SYSTEMS ARCHITECTURE
(2022)
Article
Engineering, Electrical & Electronic
S. Gerardin, M. Bagatin, A. Paccagnella, S. Beltrami, A. Costantino, G. Santin, A. Pesce, V. Ferlet-Cavrois, K. Voss
Summary: A heavy-ion beam monitor based on 3-D NAND flash memories was designed and tested, showing good efficiency and accuracy in measuring fluence, angle, uniformity, and LET of impinging particles. The system also proposed ad hoc algorithms for extracting beam parameters based on user-mode commands. Validation experiments using low-LET ionizing particles impinging at different angles were conducted.
IEEE TRANSACTIONS ON NUCLEAR SCIENCE
(2021)
Article
Chemistry, Analytical
Ruiquan He, Haihua Hu, Chunru Xiong, Guojun Han
Summary: In this paper, a novel neural network-assisted error correction scheme was proposed to improve the reliability of multi-level cell NAND flash memory. By using a relative log-likelihood ratio to estimate the actual LLR and transforming bit detection into a clustering problem, a neural network was employed to learn the error characteristics of the NAND flash memory channel, resulting in optimized performances of bit error detection. Simulation results demonstrated significant improvement in bit error detection performance and increased endurance of NAND flash memory.
Article
Engineering, Electrical & Electronic
Jang Kyu Lee, Eunseok Oh, Hyungcheol Shin
Summary: In this article, a machine learning model-based simulator and method for predicting the threshold voltage distribution of 3-D NAND flash memory is proposed. The model aims to predict the slope of each incremental step pulse program after ensuring the accuracy through training and testing with a small subset of data. The verified machine learning model can predict the V-t distribution in various environments.
IEEE TRANSACTIONS ON ELECTRON DEVICES
(2023)
Article
Engineering, Electrical & Electronic
M. Bagatin, S. Gerardin, A. Paccagnella, A. Costantino, V. Ferlet-Cavrois, G. Santin, M. Muschitiello, A. Pesce, S. Beltrami
Summary: This study investigates the secondary byproducts generated by high-energy protons inside a single-event upset (SEU) detector based on 3-D NAND Flash memories. By extending the methodology used for detecting heavy ions, the radiation response of the SEU monitor is discussed in relation to proton energy, with analysis of parameters including the number of clusters per particle, cluster size, and angle of the generated secondaries. The results provide valuable insight into nuclear reactions in state-of-the-art electronic chips and demonstrate the utility of 3-D NAND Flash memories for monitoring proton beams.
IEEE TRANSACTIONS ON NUCLEAR SCIENCE
(2022)
Article
Engineering, Electrical & Electronic
Bin Gao, Bohan Lin, Xueqi Li, Jianshi Tang, He Qian, Huaqiang Wu
Summary: This article demonstrates a highly robust unified PUF/TRNG design, which is tested for stability and randomness under different temperature and supply voltage conditions. This design shows promising prospects for future IoT security applications.
IEEE TRANSACTIONS ON ELECTRON DEVICES
(2022)
Article
Chemistry, Multidisciplinary
Qin Wan, Fei Zeng, Yiming Sun, Tongjin Chen, Junwei Yu, Huaqiang Wu, Zhen Zhao, Jiangli Cao, Feng Pan
Summary: This study demonstrates a memristive system with adaptive nucleation of phase-change nanoclusters, which responds to stimulation strength and exhibits action potential firing. The reversible nucleation of phase-change nanoclusters is confirmed by high-resolution transmission electron microscopy. This compact memristive system plays an important role in neuromorphic computing.
Article
Computer Science, Hardware & Architecture
Hongwu Jiang, Wantong Li, Shanshi Huang, Stefan Cosemans, Francky Catthoor, Shimeng Yu
Summary: This article comprehensively investigates ADC design for compute-in-memory array and shows that 6-bit precision is sufficient to guarantee accuracy for large arrays, achieving the best tradeoff between hardware performance and area overhead compared to prior designs.
IEEE DESIGN & TEST
(2022)
Article
Nanoscience & Nanotechnology
Kisung Chae, Sarah F. Lombardo, Nujhat Tasneem, Mengkun Tian, Harish Kumarasubramanian, Jae Hur, Winston Chern, Shimeng Yu, Claudia Richter, Patrick D. Lomenzo, Michael Hoffmann, Uwe Schroeder, Dina Triyoso, Steven Consiglio, Kanda Tapily, Robert Clark, Gert Leusink, Nazanin Bassiri-Gharb, Prab Bandaru, Jayakanth Ravichandran, Andrew Kummel, Kyeongjae Cho, Josh Kacher, Asif Islam Khan
Summary: Investigating nanoscale polycrystalline thin-film heterostructures is crucial for understanding the crystalline orientation and functional response in microelectronics. However, characterizing microstructural correlations at a statistically meaningful scale has been challenging. In this study, a high-throughput method based on nanobeam electron diffraction technique was introduced to investigate the orientational relations and correlations between crystallinity of materials in polycrystalline heterostructures over a length scale of microns.
ACS APPLIED MATERIALS & INTERFACES
(2022)
Article
Chemistry, Multidisciplinary
Xinyi Li, Yanan Zhong, Hang Chen, Jianshi Tang, Xiaojian Zheng, Wen Sun, Yang Li, Dong Wu, Bin Gao, Xiaolin Hu, He Qian, Huaqiang Wu
Summary: This study utilizes transition metal oxide-based memristors as artificial dendrites and spike-firing soma to construct dendritic neuron units, achieving high-efficiency spatial-temporal information processing. A hardware-implemented dendritic neural network improves accuracy for human motion recognition and exhibits a 1000x advantage in power efficiency compared to a graphics processing unit.
ADVANCED MATERIALS
(2023)
Article
Chemistry, Multidisciplinary
Tiantian Wei, Yuyao Lu, Fan Zhang, Jianshi Tang, Bin Gao, Pu Yu, He Qian, Huaqiang Wu
Summary: In this study, the conductive filaments (CFs) with different morphologies after forming, set, and reset operations in HfOx-based memristor devices are clearly revealed for the first time through 3D reconstruction of conductive atomic force microscopy (c-AFM) images. Multiple CFs are successfully observed in devices with three different resistive states, exhibiting hourglass, inverted-cone, and short-cone morphologies. The rupture location of CFs after the reset operation is also clearly observed. These findings provide insights into the resistive switching mechanism and can contribute to the design and optimization of oxide-based memristors for memory and computing applications.
ADVANCED MATERIALS
(2023)
Article
Engineering, Electrical & Electronic
Hongwu Jiang, Shanshi Huang, Wantong Li, Shimeng Yu
Summary: Compute-in-memory (CIM) is a promising hardware acceleration solution for machine learning that integrates computation directly into memory, but the challenge of analog-to-digital converters (ADCs) in CIM designs has been a major concern. This study proposes a novel CIM architecture called ENNA, which uses an ADC-free subarray design and a pulse-width modulation (PWM) input encoding scheme for improved performance. Evaluation results demonstrate high energy efficiency and throughput on various DNN models, and a heterogeneous 3D integration scheme further enhances performance and reduces area overhead.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS I-REGULAR PAPERS
(2023)
Article
Engineering, Electrical & Electronic
Xueqi Li, Liyang Pan, Junyi Wang, Bin Gao, Jianshi Tang, He Qian, Huaqiang Wu
Summary: This article proposes a novel dual-step page forming method that can realize low-current forming and improve a bit error rate (BER). Based on this technique, a no-verify page-forming scheme is proposed and can achieve a fast-forming speed of 7.56 Mb/s.
IEEE TRANSACTIONS ON ELECTRON DEVICES
(2023)
Article
Engineering, Electrical & Electronic
Guan Wang, Zhongwang Pang, Fangmin Wang, Yufeng Chen, Hongfei Dai, Bo Wang
Summary: Accurate road condition monitoring is crucial for alleviating urban traffic congestion and improving the urban environment. This paper proposes a fiber-based traffic monitoring method that can capture traffic flow of both heavy and light vehicles, serving as an efficient complement to navigation services. The method is validated by comparing the detected light vehicle flow with traffic data from a mobile navigation company, showing a correlation coefficient of 0.96. The study also reveals the practicality of fiber sensing in traffic monitoring, road health measurement, and its potential in smart city development.
JOURNAL OF LIGHTWAVE TECHNOLOGY
(2023)
Article
Engineering, Electrical & Electronic
Ruofei Hu, Jianshi Tang, Yue Xi, Zhixing Jiang, Yuyao Lu, Bin Gao, He Qian, Huaqiang Wu
Summary: A nitrogen-oxyanion-doped hafnium oxide RRAM with improved forming voltage, on/off ratio, and endurance is demonstrated. The critical electric field of N-doped RRAM for forming is 40% lower than that of undoped RRAM. The N-doped RRAM achieves 3x improvement in on/off ratio and 10x improvement in endurance at the forming voltage of 2 V, which is suitable for integration with advanced silicon technology nodes.
IEEE ELECTRON DEVICE LETTERS
(2023)
Article
Engineering, Electrical & Electronic
Yudeng Lin, Jianshi Tang, Bin Gao, Qingtian Zhang, He Qian, Huaqiang Wu
Summary: Deep learning models implemented using memristors offer high scalability and energy efficiency for resource-constrained edge computing applications. However, the inherent physical randomness of memristors leads to significant performance degradation. In this study, a unified architecture incorporating a Bayesian-based training method and lightweight transfer scheme is proposed to address the robustness, energy, and time consumption issues caused by memristor variations. Experimental results demonstrate that this architecture can double the speed and energy efficiency of deploying deep learning models.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS II-EXPRESS BRIEFS
(2023)
Article
Engineering, Electrical & Electronic
Yudeng Lin, Jianshi Tang, Bin Gao, Qi Qin, Qingtian Zhang, He Qian, Huaqiang Wu
Summary: Resistive random access memory (RRAM)-based neuromorphic hardware accelerators are attractive for neural network acceleration due to their high energy efficiency. However, the variations of RRAM can cause significant conductance deviation and performance degradation. A novel write-verify scheme is proposed to transfer weights with different acceptable error margins, achieving a high-speed and high-efficiency write-verify process.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS II-EXPRESS BRIEFS
(2023)
Article
Engineering, Electrical & Electronic
Yufeng Chen, Hongfei Dai, Wenlin Li, Fangmin Wang, Bo Wang, Lijun Wang
Summary: Fiber-optic time synchronization (FOTS) has been crucial for the efficient operation of modern society. This article proposes a time reversal-enabled FOTS method that allows measuring clock difference between two sites without calculating fiber link delay. It eliminates the need for data layer exchange and enables multiple-access time synchronization. The method achieves a time deviation of 25 ps at 1 s and 2 ps at 1000 s on a 230-km fiber link, demonstrating its effectiveness.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2023)
Proceedings Paper
Computer Science, Hardware & Architecture
Yandong Luo, Piyush Kumar, Yu-Ching Liao, William Hwang, Fen Xue, Wilman Tsai, Shan X. Wang, Azad Naeemi, Shimeng Yu
Summary: In this paper, a system level evaluation is performed for DNN inference engines using SOT-MRAM, including compute-in-memory (CIM) paradigm and near-memory systolic array. The results show that SOT-MRAM can achieve 51% to 93% higher energy efficiency than SRAM for read-intensive CIM tasks at different nodes, and 17% higher energy efficiency for write-intensive systolic array tasks at 7nm node, when compared to SRAM global buffer.
2022 14TH IEEE INTERNATIONAL MEMORY WORKSHOP (IMW 2022)
(2022)
Proceedings Paper
Computer Science, Hardware & Architecture
Po-Kai Hsu, Shimeng Yu
Summary: This paper explores the feasibility of in-memory hyperdimensional computing on 3D NAND Flash for genome sequencing, with a focus on SARS-CoV-2 genome sequences. The results indicate that despite the non-idealities of 3D NAND Flash, the classification accuracy is robust, and the system performance achieves improvement in energy efficiency and area efficiency compared to PCM-based HDC engines.
2022 14TH IEEE INTERNATIONAL MEMORY WORKSHOP (IMW 2022)
(2022)