Gyuweon Jung
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Gyuweon Jung has a total of 75 co-authors. Here are the top co-authors by number of shared publications:
- Jong-Ho Lee
Department of Electrical and Computer Engineering and Inter‐university Semiconductor Research Center Seoul National University Seoul Republic of Korea
33 shared publications - Wonjun Shin
Department of Semiconductor Convergence Engineering Sungkyunkwan University Suwon Republic of Korea
15 shared publications Kangwook Choi
Inter-University Semiconductor Research Center, Seoul National University
14 shared publicationsJinwoo Park
Inter-University Semiconductor Research Center, Seoul National University
10 shared publicationsYujeong Jeong
10 shared publications- Seongbin Hong
Department of Electrical and Computer Engineering, Inter-University Semiconductor Research Center (ISRC), Seoul National University, Seoul, South Korea
9 shared publications
Latest Publications
- 1 October 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 26 (20) , 30663-30670
Low-Frequency Noise Analysis in Sensor-Circuit Integrated Computing System for Selective Gas Detection
- 1 October 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 26 (20) , 30663-30670
Low-Frequency Noise Analysis in Sensor-Circuit Integrated Computing System for Selective Gas Detection
- Open Access
- 24 August 2026
- Published by Wiley in Advanced Science
- Vol. 13 (61) , e77390
Triple‐Level Content‐Addressable Vertical NAND Flash Memory
Content‐addressable memory (CAM) enables massively parallel search by directly comparing input data with stored entries, offering a powerful paradigm for accelerating data‐intensive tasks. Here, we present a triple‐level content‐addressable vertical NAND flash memory (V‐NAND TLCAM) that exploits the eight threshold‐voltage states of commercial triple‐level‐cell (TLC) V‐NAND flash devices to achieve compact, high‐throughput, and energy‐efficient in‐memory search. A parity‐split (PS) scheme, in which odd and even word‐lines are accessed separately, enhances the sensing margin and mitigates worst‐case errors while preserving storage density and supporting don't‐care matching. Measurements on >100‐layer TLC V‐NAND flash memory confirm stable operation with up to a 24× improvement in read‐operation throughput and a 54.4× reduction in the estimated read energy compared with conventional readout. Furthermore, few‐shot learning simulations on the Omniglot dataset demonstrate that the proposed TLCAM maintains over 90% accuracy in 5‐shot learning, approaching GPU‐based cosine similarity baselines. These results highlight V‐NAND TLCAM as a scalable and energy‐efficient hardware platform for bridging device‐level innovations with practical machine learning applications.
- 24 August 2026
- Published by Wiley in Advanced Science
- Vol. 13 (61) , e77390
Open AccessTriple‐Level Content‐Addressable Vertical NAND Flash Memory
Content‐addressable memory (CAM) enables massively parallel search by directly comparing input data with stored entries, offering a powerful paradigm for accelerating data‐intensive tasks. Here, we present a triple‐level content‐addressable vertical NAND flash memory (V‐NAND TLCAM) that exploits the eight threshold‐voltage states of commercial triple‐level‐cell (TLC) V‐NAND flash devices to achieve compact, high‐throughput, and energy‐efficient in‐memory search. A parity‐split (PS) scheme, in which odd and even word‐lines are accessed separately, enhances the sensing margin and mitigates worst‐case errors while preserving storage density and supporting don't‐care matching. Measurements on >100‐layer TLC V‐NAND flash memory confirm stable operation with up to a 24× improvement in read‐operation throughput and a 54.4× reduction in the estimated read energy compared with conventional readout. Furthermore, few‐shot learning simulations on the Omniglot dataset demonstrate that the proposed TLCAM maintains over 90% accuracy in 5‐shot learning, approaching GPU‐based cosine similarity baselines. These results highlight V‐NAND TLCAM as a scalable and energy‐efficient hardware platform for bridging device‐level innovations with practical machine learning applications.
- Open Access
- 20 August 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1) , 9963
Intelligent artificial olfactory nervous system with sensory adaptation capabilities
Accurate gas identification across diverse environments is crucial for the advancement of intelligent artificial olfaction, but remains challenging for conventional data-driven learning approaches. Here, we report an artificial olfactory nervous system capable of reliably identifying novel gas stimuli regardless of environmental variations by leveraging sensory adaptation mechanisms inspired by biological olfactory systems. Inspired by biological olfactory systems, our system effectively integrates artificial olfactory receptor neurons, cortical neurons, and intermediate olfactory neurons that facilitate signal transmission. Receptor neurons can dynamically adapt their sensitivity to olfactory stimuli, enabling rapid responses to familiar gases and selective filtering of background interference, thus generating distinctive signal patterns for novel gases. Signals from receptor neurons are modulated in olfactory neurons, then fed into the cortical neurons and translated into olfactory perception. We show that, despite being trained in limited settings, the system can accurately identify gases in various environments.
- 20 August 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1) , 9963
Open AccessIntelligent artificial olfactory nervous system with sensory adaptation capabilities
Accurate gas identification across diverse environments is crucial for the advancement of intelligent artificial olfaction, but remains challenging for conventional data-driven learning approaches. Here, we report an artificial olfactory nervous system capable of reliably identifying novel gas stimuli regardless of environmental variations by leveraging sensory adaptation mechanisms inspired by biological olfactory systems. Inspired by biological olfactory systems, our system effectively integrates artificial olfactory receptor neurons, cortical neurons, and intermediate olfactory neurons that facilitate signal transmission. Receptor neurons can dynamically adapt their sensitivity to olfactory stimuli, enabling rapid responses to familiar gases and selective filtering of background interference, thus generating distinctive signal patterns for novel gases. Signals from receptor neurons are modulated in olfactory neurons, then fed into the cortical neurons and translated into olfactory perception. We show that, despite being trained in limited settings, the system can accurately identify gases in various environments.
- Open Access
- 10 July 2026
- Published by Springer Nature in Research Square
Interpretable Kolmogorov-Arnold Network Using Vertical NAND Flash Memory
Compute-in-memory (CIM) hardware based on commercial vertical NAND (V-NAND) flash memory offers high density and low-power operation, but most neural-network implementations remain difficult to interpret because the learned input-output relationship is embedded in entangled weights. Here we demonstrate an interpretable CIM implementation of a Kolmogorov-Arnold network (KAN) using commercial triple-level-cell V-NAND flash memory with more than 100 stacked layers. In the proposed V-NAND KAN, software-trained edge-wise activation functions are transferred to hardware by approximating them as weighted sums of measured V-NAND spline responses. These spline primitives are physically generated from pairs of series-connected cells driven by complementary word-line biases, while threshold-voltage programming tunes their position and width without modifying the memory structure. We first validate this mapping by implementing multiplication and division, representative symbolic operations extracted from software KANs, and show close agreement between ideal software targets and measured V-NAND KAN outputs. We then apply the approach to multimodal gas sensing using integrated gas, temperature and pressure sensor data. The KAN identifies a compact and physically meaningful functional relationship and enables real-time, low-power gas-concentration inference under mixed-gas conditions. Compared with multilayer perceptron-based implementations, the V-NAND KAN achieves comparable prediction performance while requiring far fewer memory cells and reducing the read-operation component of inference energy by up to 94%. These results establish commercial V-NAND flash memory as a practical platform for interpretable, low-power CIM hardware that can both predict and explain.
- 10 July 2026
- Published by Springer Nature in Research Square
Open AccessInterpretable Kolmogorov-Arnold Network Using Vertical NAND Flash Memory
Compute-in-memory (CIM) hardware based on commercial vertical NAND (V-NAND) flash memory offers high density and low-power operation, but most neural-network implementations remain difficult to interpret because the learned input-output relationship is embedded in entangled weights. Here we demonstrate an interpretable CIM implementation of a Kolmogorov-Arnold network (KAN) using commercial triple-level-cell V-NAND flash memory with more than 100 stacked layers. In the proposed V-NAND KAN, software-trained edge-wise activation functions are transferred to hardware by approximating them as weighted sums of measured V-NAND spline responses. These spline primitives are physically generated from pairs of series-connected cells driven by complementary word-line biases, while threshold-voltage programming tunes their position and width without modifying the memory structure. We first validate this mapping by implementing multiplication and division, representative symbolic operations extracted from software KANs, and show close agreement between ideal software targets and measured V-NAND KAN outputs. We then apply the approach to multimodal gas sensing using integrated gas, temperature and pressure sensor data. The KAN identifies a compact and physically meaningful functional relationship and enables real-time, low-power gas-concentration inference under mixed-gas conditions. Compared with multilayer perceptron-based implementations, the V-NAND KAN achieves comparable prediction performance while requiring far fewer memory cells and reducing the read-operation component of inference energy by up to 94%. These results establish commercial V-NAND flash memory as a practical platform for interpretable, low-power CIM hardware that can both predict and explain.
- Open Access
- 8 May 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
CMOS-compatible ferroelectric tunnel junctions integrate stochastic sampling and deterministic computing for image generation
Recent progress in generative modeling has intensified the need for compact, energy-efficient hardware platforms. Yet, implementing image generation directly in hardware remains challenging due to the conflicting requirements of stochastic latent space sampling and deterministic decoding. Here, we show a unified hardware framework based on hafnium-oxide ferroelectric tunnel junctions (FTJs) that intrinsically support both functionalities within a single device array. Leveraging the CMOS- and VLSI-compatible fabrication of hafnia ferroelectrics, we realize dual-mode operation: random telegraph noise generation for controllable stochastic sampling, and high-fidelity vector–matrix multiplication enabled by non-volatile multi-level conductance states. Voltage and sampling-time tuning provide fine control over randomness and reliability, enabling high-quality image generation for tasks such as handwritten digit synthesis (MNIST) and high-resolution facial image generation (CelebA). Circuit-level demonstrations confirm stable performance over 10 5 cycles, surpassing prior hardware-based approaches and illustrating a viable route toward scalable, on-chip generative AI accelerators.
- 8 May 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
Open AccessCMOS-compatible ferroelectric tunnel junctions integrate stochastic sampling and deterministic computing for image generation
Recent progress in generative modeling has intensified the need for compact, energy-efficient hardware platforms. Yet, implementing image generation directly in hardware remains challenging due to the conflicting requirements of stochastic latent space sampling and deterministic decoding. Here, we show a unified hardware framework based on hafnium-oxide ferroelectric tunnel junctions (FTJs) that intrinsically support both functionalities within a single device array. Leveraging the CMOS- and VLSI-compatible fabrication of hafnia ferroelectrics, we realize dual-mode operation: random telegraph noise generation for controllable stochastic sampling, and high-fidelity vector–matrix multiplication enabled by non-volatile multi-level conductance states. Voltage and sampling-time tuning provide fine control over randomness and reliability, enabling high-quality image generation for tasks such as handwritten digit synthesis (MNIST) and high-resolution facial image generation (CelebA). Circuit-level demonstrations confirm stable performance over 10 5 cycles, surpassing prior hardware-based approaches and illustrating a viable route toward scalable, on-chip generative AI accelerators.
- 20 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Electron Device Letters
- Vol. 47 (6) , 1117-1120
2-T Embedded Flash Memory Using Standard 28-nm Logic CMOS Technology
A 2-transistor (2-T) embedded flash (eflash) memory fully implemented using a 28 nm logic CMOS technology (Samsung 28 nm LPP) is proposed. The proposed 2-T eflash requires no additional process steps. Compared with prior eflash memory employing three or more transistors and requiring program/erase voltages above 6.8 V, the proposed 2-T eflash reduces the device count and lowers the operating voltage to 4.5 V, substantially simplifying the peripheral circuit configuration. Owing to the low-voltage tunneling-based operation, the eflash exhibits high endurance exceeding $10^{{7}}$ cycles and retention characteristics are maintained for up to $10^{{4}}$ s at an elevated temperature of 80°C. While preserving high-performance retention, the proposed eflash achieves endurance exceeding $10^{{7}}$ cycles, showing nearly a two orders of magnitude improvement, compared to conventional flash memory endurance. In addition, selective program and erase operations are demonstrated, enabling precise threshold-voltage tuning in a memory array. These results indicate that the proposed 2-T eflash is a promising candidate for scalable compute-in-memory applications.
- 20 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Electron Device Letters
- Vol. 47 (6) , 1117-1120
2-T Embedded Flash Memory Using Standard 28-nm Logic CMOS Technology
A 2-transistor (2-T) embedded flash (eflash) memory fully implemented using a 28 nm logic CMOS technology (Samsung 28 nm LPP) is proposed. The proposed 2-T eflash requires no additional process steps. Compared with prior eflash memory employing three or more transistors and requiring program/erase voltages above 6.8 V, the proposed 2-T eflash reduces the device count and lowers the operating voltage to 4.5 V, substantially simplifying the peripheral circuit configuration. Owing to the low-voltage tunneling-based operation, the eflash exhibits high endurance exceeding $10^{{7}}$ cycles and retention characteristics are maintained for up to $10^{{4}}$ s at an elevated temperature of 80°C. While preserving high-performance retention, the proposed eflash achieves endurance exceeding $10^{{7}}$ cycles, showing nearly a two orders of magnitude improvement, compared to conventional flash memory endurance. In addition, selective program and erase operations are demonstrated, enabling precise threshold-voltage tuning in a memory array. These results indicate that the proposed 2-T eflash is a promising candidate for scalable compute-in-memory applications.
- 17 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 26 (11) , 17045-17054
Analysis of Accuracy, Area, Latency, and Energy Consumption of Gas-Sensing-Based In-Sensor Neuromorphic Computing
The increasing demand for in-sensor neuromorphic computing motivates a system-level framework that jointly evaluates accuracy, area, latency, and energy. We prototype a platform that integrates a fabricated FET-based gas sensor with an embedded micro-heater, a floating-gated pMOS synapse array, and processing circuits, such as pulse-width modulator (PWM) and integrate-and-fire (I&F) neurons. Using transient responses to H2S and NO2 across temperature and concentration sweeps, we benchmark networks with 3–6 input sensors, three hidden layers (16–128 neurons), and two regression outputs. Accuracy improves with more sensors and larger hidden layers, saturating beyond 64 neurons. Quantization and circuit constraints add bounded error. Area is insensitive to sensor count but scales with hidden neurons, with the high-voltage switch dominating area. Latency is sensor-limited. Energy consumption is overwhelmingly set by the micro-heater (~99.97% of total), making overall consumption highly sensitive to sensor count while relatively insensitive to network size. These results identify trade-offs among accuracy, area, latency, and energy, pinpointing heater efficiency, minimal sensor selection, and high-voltage-switch optimization as the primary design-technology co-optimization (DTCO) keys for designing compact, fast, and energy-aware in-sensor neuromorphic platforms.
- 17 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 26 (11) , 17045-17054
Analysis of Accuracy, Area, Latency, and Energy Consumption of Gas-Sensing-Based In-Sensor Neuromorphic Computing
The increasing demand for in-sensor neuromorphic computing motivates a system-level framework that jointly evaluates accuracy, area, latency, and energy. We prototype a platform that integrates a fabricated FET-based gas sensor with an embedded micro-heater, a floating-gated pMOS synapse array, and processing circuits, such as pulse-width modulator (PWM) and integrate-and-fire (I&F) neurons. Using transient responses to H2S and NO2 across temperature and concentration sweeps, we benchmark networks with 3–6 input sensors, three hidden layers (16–128 neurons), and two regression outputs. Accuracy improves with more sensors and larger hidden layers, saturating beyond 64 neurons. Quantization and circuit constraints add bounded error. Area is insensitive to sensor count but scales with hidden neurons, with the high-voltage switch dominating area. Latency is sensor-limited. Energy consumption is overwhelmingly set by the micro-heater (~99.97% of total), making overall consumption highly sensitive to sensor count while relatively insensitive to network size. These results identify trade-offs among accuracy, area, latency, and energy, pinpointing heater efficiency, minimal sensor selection, and high-voltage-switch optimization as the primary design-technology co-optimization (DTCO) keys for designing compact, fast, and energy-aware in-sensor neuromorphic platforms.
- 13 April 2026
- Published by AIP Publishing in Applied Physics Letters
- Vol. 128 (15)
Retention improvement in vertical NAND flash memory using block soft erase scheme
We propose a block soft erase method to enhance the retention characteristics of vertical NAND (V-NAND) flash memory while significantly reducing soft erase time. Unlike prior incremental-step-pulse-erasing-based approaches that adjust each cell individually, the proposed method applies a single soft erase pulse to all cells simultaneously. Experimental results using commercial triple-level-cell V-NAND flash memory show that the block soft erase method enables Vth tuning across multiple program-verify levels (PVs), even under cell-to-cell variation due to word-line position and string differences. Retention characteristics measured at 85 °C demonstrated a 21.6% improvement at PV7 after 104 s.
- 13 April 2026
- Published by AIP Publishing in Applied Physics Letters
- Vol. 128 (15)
Retention improvement in vertical NAND flash memory using block soft erase scheme
We propose a block soft erase method to enhance the retention characteristics of vertical NAND (V-NAND) flash memory while significantly reducing soft erase time. Unlike prior incremental-step-pulse-erasing-based approaches that adjust each cell individually, the proposed method applies a single soft erase pulse to all cells simultaneously. Experimental results using commercial triple-level-cell V-NAND flash memory show that the block soft erase method enables Vth tuning across multiple program-verify levels (PVs), even under cell-to-cell variation due to word-line position and string differences. Retention characteristics measured at 85 °C demonstrated a 21.6% improvement at PV7 after 104 s.
- Open Access
- 28 February 2026
- Published by Elsevier in SSRN Electronic Journal
Accurate Temperature Extraction in Embedded Polysilicon Micro-heaters
Accurate temperature estimation of embedded micro-heater is critical for the reliable operation and data interpretation of FET-type gas sensors. In this work, length-resolved temperature extraction method is proposed which improves heater temperature estimation by separating contact and sensing region related resistance contributions. Polysilicon micro-heaters with different sensing region lengths enable independent extraction of the absolute temperature sensitivities of the contact and sensing regions. Using temperature sensitivities, the contact temperature (TC) and sensing region temperature (TS) are separately estimated. COMSOL simulations confirm the distribution of temperature and the physical basis of the proposed decomposition. Quantitative comparison with simulation results demonstrates that the proposed TS provides a more accurate estimation of the sensing region temperature than the conventional RTD(resistance temperature detector)-based temperature TRTD, maintaining a good agreement up to operating temperatures of 240 °C. Finally, NO2 sensing results verify that the proposed method enables consistent and accurate temperature extraction across devices with different heater structures. These results confirm that length-resolved resistance analysis provides an effective and practical approach for accurate temperature estimation.
- 28 February 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessAccurate Temperature Extraction in Embedded Polysilicon Micro-heaters
Accurate temperature estimation of embedded micro-heater is critical for the reliable operation and data interpretation of FET-type gas sensors. In this work, length-resolved temperature extraction method is proposed which improves heater temperature estimation by separating contact and sensing region related resistance contributions. Polysilicon micro-heaters with different sensing region lengths enable independent extraction of the absolute temperature sensitivities of the contact and sensing regions. Using temperature sensitivities, the contact temperature (TC) and sensing region temperature (TS) are separately estimated. COMSOL simulations confirm the distribution of temperature and the physical basis of the proposed decomposition. Quantitative comparison with simulation results demonstrates that the proposed TS provides a more accurate estimation of the sensing region temperature than the conventional RTD(resistance temperature detector)-based temperature TRTD, maintaining a good agreement up to operating temperatures of 240 °C. Finally, NO2 sensing results verify that the proposed method enables consistent and accurate temperature extraction across devices with different heater structures. These results confirm that length-resolved resistance analysis provides an effective and practical approach for accurate temperature estimation.
- 18 February 2026
- Published by Elsevier in Sensors and Actuators B: Chemical
- Vol. 456
Low-power NO2 gas alarm platform using in-memory latched comparator
- 18 February 2026
- Published by Elsevier in Sensors and Actuators B: Chemical
- Vol. 456
Low-power NO2 gas alarm platform using in-memory latched comparator