Hunjun Lee

Hunjun Lee

Auto-generated profile, based on publicly available information

Hunjun Lee

Hanyang University, Seoul, Republic of Korea

Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
17 April 2021
Verified
Conference Paper
Open access
HLHunjun Lee
Hunjun Lee
CKChanmyeong Kim
Chanmyeong Kim
YCYujin Chung
Yujin Chung
JKJangwoo Kim
Jangwoo Kim
Brain-inspired computing aims to understand the cognitive mechanisms of a brain and apply them to advance various areas in computer science. Deep learning is an example to greatly improve the field of pattern recognition and classification by utilizing an artificial neural network (ANN). To exploit advanced mechanisms of a brain and thus make more great advances, researchers need a methodology that can simulate neural networks with higher computational capabilities such as advanced spiking neural networks (SNNs) with two-stage neurons and synaptic delays. However, existing SNN simulation methodologies are too slow and energy-inefficient due to their software-based simulation or hardware-based but time-driven execution mechanisms. In this paper, we present NeuroEngine, a fast and energy-efficient hardware-based system to efficiently simulate advanced SNNs. The key idea is to design an accelerator to enable event-driven simulations of the SNNs at a minimum cost. NeuroEngine achieves high speed and energy efficiency by carefully architecting its datapath and memory units to take the best advantage of the event-driven mechanism while satisfying all the important requirements to simulate our target SNNs. For high performance and energy efficiency, NeuroEngine applies a simpler datapath, multi-queue scheduler, and lazy update to minimize its neuron computation and event scheduling overhead. Then, we build an end-to-end simulation system by implementing a programming interface and a compilation toolchain for NeuroEngine hardware. Our evaluations show that NeuroEngine greatly improves the harmonic mean performance and energy efficiency by 4.30× and 2.60×, respectively, over the state-of-the-art time-driven simulator.

22 March 2026
Verified
Conference Paper
Open access
SSSeunghyun Song
Seunghyun Song
YJYeongwoo Jang
Yeongwoo Jang
DJDaye Jung
Daye Jung
KPKyungsoo Park
Kyungsoo Park
DKDonghan Kim
Donghan Kim
JKJangwoo Kim
Jangwoo Kim

15 June 2026
Verified
Conference Paper
Open access
ILInho Lee
Inho Lee
KLKy Yeop Lim
Ky Yeop Lim
HKHyejun Kim
Hyejun Kim
BKBeomseok Kim
Beomseok Kim
DJDongsuk Jeon
Dongsuk Jeon
YPYongjun Park
Yongjun Park
With the widespread adoption of Deep Neural Networks (DNNs), Neural Processing Units (NPUs) are emerging as energy-efficient alternatives to GPUs through parallel processing and high data reuse. However, since diverse deep learning kernels have different memory and computation resource requirements, a utilization imbalance between memory and computation resources often occurs. To address this challenge, we propose FLUX, a frequency-scalable NPU system that applies Dynamic Frequency Scaling (DFS) to the core's frequency based on layer-wise arithmetic intensity. FLUX splits the clocking into an adjustable core domain and a fixed system domain. The core-domain frequency is first estimated via roofline cycle analysis and then refined at runtime by an Energy-Delay Product (EDP)-driven calibration. Evaluation on a Gemmini NPU using a 28nm process technology shows that FLUX achieves 5.7%, 16.5%, and 27.9% EDP improvements on 8×8, 16×16, and 32×32 systolic arrays for ResNet50, respectively, reaching within 1.7% of the oracle optimal frequency assignment.

27 February 2026 in IEEE Computer Architecture Letters
Verified
Journal Article
Closed access
JKJintae Kim
Jintae Kim
BJByoungjun Jeong
Byoungjun Jeong
CKChangdae Kim
Changdae Kim
NRNarae Ryu
Narae Ryu
EPEunji Pak
Eunji Pak
HLHunjun Lee
Hunjun Lee
Analog Computing-in-Memory (ACiM) devices perform matrix operations directly within memory arrays, offering high throughput for deep neural network inference. At the same time, they are susceptible to various noise sources, which reduces the computational accuracy. This inherent trade-off limits the adoption of ACiM devices as a stand-alone accelerator. In this paper, we propose the first heterogeneous inference-serving framework that coordinates ACiM hardware and conventional digital processors during inference. The system dynamically partitions each neural network into an analog sub-network executed on high-throughput ACiM devices and a digital sub-network handled by precise digital devices. Then, it integrates a transition layer and a custom training strategy to maintain accuracy across varying analog-digital partitions. Also, it adopts a two-level scheduler to adjust the partition ratio at runtime in response to varying query demands. We evaluate our system using five network models to demonstrate the benefits of our system.

20 June 2025
Verified
Conference Paper
Open access
YJYeongwoo Jang
Yeongwoo Jang
DJDaye Jung
Daye Jung
SSSeunghyun Song
Seunghyun Song
HLHunjun Lee
Hunjun Lee
JKJangwoo Kim
Jangwoo Kim

1 July 2026 in IEEE Computer Architecture Letters
Verified
Journal Article
Closed access
JLJunsuk Lee
Junsuk Lee
DKDooik Kim
Dooik Kim
WSWoojin Son
Woojin Son
HLHunjun Lee
Hunjun Lee