A5101763664- Distributed and Parallel Computing Systems
- Advanced Data Storage Technologies
- Topic Modeling
- Parallel Computing and Optimization Techniques
- Internet of Things and Social Network Interactions
- Peer-to-Peer Network Technologies
- Robotics and Sensor-Based Localization
- Cloud Computing and Resource Management
- Advanced Data Compression Techniques
- Robot Manipulation and Learning
- Network Security and Intrusion Detection
- Soft Robotics and Applications
- Engineering Applied Research
- Digital Filter Design and Implementation
- Software Engineering Research
- Text Readability and Simplification
- Embedded Systems Design Techniques
- Natural Language Processing Techniques
- Energy and Environmental Systems
- Software-Defined Networks and 5G
- Speech and Audio Processing
- Target Tracking and Data Fusion in Sensor Networks
- Artificial Intelligence in Healthcare and Education
- Advanced Adaptive Filtering Techniques
- Caching and Content Delivery
Ajou University
2017-2025
Chung-Ang University
2024
Dongguk University
2024
Hanyang University
2014-2023
Soongsil University
2010
Seoul National University
2010
This paper presents a sensor-data-fusion method using an Unscented Kalman Filter (UKF), to implement accurate localization system for mobile robots. Integration of data from various sensors efficient sensor fusion algorithm is required achieve continuous and We use low cost accelerometer, gyroscope, encoders obtain robot motion information. The UKF, used as algorithm, advanced filtering technique which outperforms the widely-used Extended (EKF) in many applications. able compensate slip...
Abstract Previous work in prompt engineering for large language models has introduced different gradient-free probability-based selection methods that aim to choose the optimal among candidates a given task but have failed provide comprehensive and fair comparison between each other. In this paper, we propose unified framework interpret evaluate existing by performing extensive experiments on 13 common diverse NLP tasks. We find of can be interpreted as some variant method maximizes mutual...
The growing computational demands of AI inference have led to widespread use hardware accelerators for different platforms, spanning from edge the datacenter/cloud. Certain application areas, such as in high-frequency trading (HFT) [1–2], a hard latency deadline successful execution. We present our new accelerator which achieves high capability with outstanding single-stream responsiveness demanding service-layer objective (SLO)-based services and pipelined applications, including large...
Previous works in prompt engineering for large language models have introduced different gradient-free probability-based selection methods that aim to choose the optimal among candidates a given task but failed provide comprehensive and fair comparison between each other. In this paper, we propose unified framework interpret evaluate existing by performing extensive experiments on 13 common diverse NLP tasks. We find of can be interpreted as some variant method maximizes mutual information...
In this paper, we propose an active robotic 3D reconstruction methodology for achieving full object reconstruction. Existing approaches often struggle to cover the entire view space of or reconstruct occluded regions, such as bottom back side. To address these limitations, introduce a two-stage pipeline, named See-Then-Grasp (STG), that employs robot manipulator direct interaction with object. The moves toward points highest uncertainty, ensuring efficient data acquisition and rapid Our...
This paper proposes a flexible encoding technique based on multi-stage multiple scale lattice vector quantization and block-constrained trellis coded quantization. It is used for the spectrum encoding, more precisely of LSF parameters, incorporated in recently standardized 3GPP EVS codec. The proposed method can handle bit allocations signal types with low complexity memory requirements.
Large Language Models (LLMs) have demonstrated outstanding performance in general domain knowledge inference such as arithmetic reasoning, commonsense and open-domain question answering. Traditional LLMs were mainly developed serviced by big tech companies commercial, making it nearly impossible for researchers to access parameters. However, the recent emergence of non-commercial has led ongoing research efforts surpass commercial specific areas using task-specific or domain-specific LLMs....
Recent advances in memory technology, hierarchy is becoming diverse with performance-differentiated such as high bandwidth (HBM) and non-volatile (NVM) modern computer systems. However, the current placement has been designed assumption that all same capabilities based on DRAM. In this letter, we analyze schemes state-of-the-art Linux systems to hardware-assisted tiered system. Our analysis conducted real system equipped Intel's Optane Persistent Memory, enabling DRAM (fast) (slow) memory....
전자 산업에서의 소형화와 디지털화에 따라 적층 세라믹 커패시터(Multi-Layer Ceramic Capacitors: MLCC) 또한 DC Blocking, 디커플링, 필터링 등의 기능이 이에 부응하여 그 수요가 꾸준히 증가해왔다. MLCC의 등가회로를 모델링하는 기법이 많이 연구되었는데, 지금까지의 연구를 살펴보면 대부분이 소자의 주파수 특성을 측정한 후, 결과를 바탕으로 소자를 모델링하므로 제작 과정과 측정 과정에서 물질적, 시간적 손실을 수반한다. 이를 해결하기 위한 방법으로 본 논문에서는 구조 정보와 물질 정보로부터 설계단계에서 임피던스 예측할 수 있는 모델링 방법을 제시한다. 미분 방정식으로 표현되는 결합 전송선로 방정식으로부터 임의의 N개 층을 가지는 다층 평판 커패시터(N-Layer Capacitor)의 임피던스를 정보의 수식으로 표현할 있음을 보였다. 이렇게 정의된 수식으로부터 정보를 예측하였으며, EM 시뮬레이션 결과와 비교하였다. 결과, 제시한 예측 결과가...
This paper proposes a blind bandwidth extension (BBE) technique that improves the quality of narrowband (NB) speech signal enhanced by codebook-based Bayesian enhancement method with minimum mean square error (MMSE) estimate. The MMSE weight vector NB envelope codebook is used to estimate wideband (WB) envelope. An adaptive double shifting introduced in BBE obtain an WB excitation signal. Its performance assessed via log-spectral distortion (SD) and informal listening tests.
This paper proposes an efficient storage virtualization system which allows users to view hard disk resources constituted of numerous nodes as a large logical space using distributed hash tables P2P networking. The proposed is developed at the device level operating (e.g. MS-Windows) and suitable for in Intranet environments. be recognized single Windows Explorer users' conveniences does not require any supplementary client program application layer. In addition, it enhances security via...