Eunhyuk Park

The poor performance of NAND Flash memory, such as long access latency and large granularity access, is the major bottleneck of graph processing. This paper proposes an intelligent storage for graph processing which is based on fast and low cost racetrack memory and a pointer-assisted graph representation. Our experiments show that the proposed intelligent storage based on racetrack memory reduces total processing time of three representative graph computations by 40.2%~86.9% compared to the gra

This work investigates a collaborative sensing and data collection system in which multiple uncrewed aerial vehicles (UAVs) sense an area of interest and transmit images to a cloud server (CS) for processing. To accelerate the completion of sensing missions, including data transmission, the sensing task is divided into individual private sensing tasks for each UAV and a common sensing task that is executed by all UAVs to enable cooperative transmission. Unlike existing studies, we explore the us

The poor performance of NAND Flash memory, such as long access latency and large granularity access, is the major bottleneck of graph processing. This paper proposes an intelligent storage for graph processing which is based on fast and low cost racetrack memory and a pointer-assisted graph representation. Our experiments show that the proposed intelligent storage based on racetrack memory reduces total processing time of three representative graph computations by 40.2%~86.9% compared to the gra

The poor performance of NAND Flash memory, such as long access latency and large granularity access, is the major bottleneck of graph processing. This paper proposes an intelligent storage for graph processing which is based on fast and low cost racetrack memory and a pointer-assisted graph representation. Our experiments show that the proposed intelligent storage based on racetrack memory reduces total processing time of three representative graph computations by 40.2%~86.9% compared to the gra

As the dense deployment of access points (APs) in cell-free massive multiple-input multiple-output (CF-mMIMO) systems presents significant challenges, per-AP coverage can be expanded using large-scale antenna arrays (LAAs). However, this approach incurs high implementation costs and substantial fronthaul demands due to the need for dedicated RF chains for all antennas. To address these challenges, we propose a hybrid beamforming framework that integrates wave-domain beamforming via stacked intel

Stacked intelligent metasurfaces (SIMs), which are composed of multi-layer programmable metasurfaces, support beamforming in the wave domain, utilizing a limited number of radio frequency (RF) chains. This work investigates the application of SIMs in the downlink of cell-free massive multiple-input multiple-output systems with finite-capacity fronthaul links. Specifically, we address the joint optimization of fronthaul compression and hybrid digital and wave-domain beamforming. To tackle the res

This work studies an over-the-air-fog computation (AirFogComp) system, wherein Internet-of- Things (IoT) devices collaboratively learn a machine learning model by communicating with a central server (CS) through a network of access points (APs). Considering the finite capacity of fronthaul links between APs and the CS, we address the challenge of jointly optimizing linear precoding at the IDs, linear processing, and quantization noise covariance matrices at the APs, along with linear combining a

최근 초거대 언어 모델(LLM)의 실사용 확대에 따라 메모리 제약을 극복하면서도 높은 성능을 유지할 수 있는 모델 양자화 기법이 주목받고 있다. 특히 3비트 가중치 양자화는 메모리 사용량을 최소화하면서 실용적 성능을 확보할 수 있는 한계점으로 인식되지만, 실제 하드웨어에서 효율적으로 지원되지 않아 가속 성능 향상에 어려움이 있다. 본 연구에서는 이를 해결하기 위해 하드웨어 친화적인 2-4비트 혼합 정밀도 양자화 기법과 탐욕적 탐색(Greedy Search)을 활용한 레이어별 비트 할당 방식을 제안한다. 제안 방식은 성능에 민감한 선형 레이어는 4비트로, 상대적으로 민감도가 낮은 선형 레이어는 2비트로 양자화하여 메모리 효율성과 하드웨어 친화성을 동시에 달성한다. 실험 결과, 제안된 탐욕적 탐색 기반 2-4비트 혼합 정밀도 모델은 동일한 메모리 사용량을 가지는 고정 정밀도 3비트 양자화 모델 대비 더 우수하거나 동일한 성능을 보이는 동시에 최대 5배의 하드웨어 가속 효과를 보였다.

This work investigates a collaborative sensing and data collection system in which multiple unmanned aerial vehicles (UAVs) sense an area of interest and transmit images to a cloud server (CS) for processing. To accelerate the completion of sensing missions, including data transmission, the sensing task is divided into individual private sensing tasks for each UAV and a common sensing task that is executed by all UAVs to enable cooperative transmission. Unlike existing studies, we explore the us

As the dense deployment of access points (APs) in cell-free massive multiple-input multiple-output (CF-mMIMO) systems presents significant challenges, per-AP coverage can be expanded using large-scale antenna arrays (LAAs). However, this approach incurs high implementation costs and substantial fronthaul demands due to the need for dedicated RF chains for all antennas. To address these challenges, we propose a hybrid beamforming framework that integrates wave-domain beamforming via stacked intel