Daegun Yoon
Name: Daegun Yoon (윤대건, 尹㙜建)
Degree: Ph.D.
Affiliation: ETRI, Republic of Korea
Email: kljp@etri.re.kr
Tel: +81 42-860-5859
Positions
Researcher in Electronics and Telecommunications Research Institute (ETRI), Republic of Korea
Jan. 2024 ~ Present (High Performance Computing System Research Section)
Research interestS
On-device AI model inference
Model compression and performance optimization
MLOps
Scalable distributed training
HPC systems
Education
Ph.D. in Department of Artificial Intelligence, Ajou University
Sep. 2018 - Feb. 2024
B.S. in Department of Software, Ajou University
Mar. 2013 - Aug. 2018
Selected publications
Preserving Near-Optimal Gradient Sparsification Cost for Scalable Distributed Deep Learning. Daegun Yoon, Sangyoon Oh. CCGrid 2024, May. 2024. [Paper] [Code]
MiCRO: Near-Zero Cost Gradient Sparsification for Scaling and Accelerating Distributed DNN Training. Daegun Yoon, Sangyoon Oh. HiPC 2023, Dec. 2023. [Paper] [Code]
DEFT: Exploiting Gradient Norm Difference between Model Layers for Scalable Gradient Sparsification. Daegun Yoon, Sangyoon Oh. ICPP 2023, Aug. 2023. [Paper] [Code]
Full publications
2024
Preserving Near-Optimal Gradient Sparsification Cost for Scalable Distributed Deep Learning. Daegun Yoon, Sangyoon Oh. CCGrid 2024, May. 2024. [Paper] [Code]
2023
MiCRO: Near-Zero Cost Gradient Sparsification for Scaling and Accelerating Distributed DNN Training. Daegun Yoon, Sangyoon Oh. HiPC 2023, Dec. 2023. [Paper] [Code]
DEFT: Exploiting Gradient Norm Difference between Model Layers for Scalable Gradient Sparsification. Daegun Yoon, Sangyoon Oh. ICPP 2023, Aug. 2023. [Paper] [Code]
연합학습 기법들의 성능평가를 지원하는 이기종 기반의 실험 플랫폼 설계. 유미리, 윤대건, 오상윤. 2023 한국통신학회 하계학술대회, Jun. 2023. [Paper]
Can hierarchical client clustering mitigate the data heterogeneity effect in federated learning?. Seungjun Lee, Miri Yu, Daegun Yoon, Sangyoon Oh. Workshop on ScaDL 2023, May. 2023. [Paper]
SAGE: toward on-the-fly gradient compression ratio scaling. Daegun Yoon, Minjoong Jeong, Sangyoon Oh. The Journal of Supercomputing, Feb. 2023. [Paper] [Code]
연합학습에서의 보안 취약점 분석. 최지헌, 유미리, 윤대건, 오상윤. 2023 한국통신학회 동계학술대회, Feb. 2023. [Paper]
2022
WAVE: designing a heuristics-based three-way breadth-first search on GPUs. Daegun Yoon, Minjoong Jeong, Sangyoon Oh. The Journal of Supercomputing, Nov. 2022. [Paper] [Code]
Empirical Analysis on Top-k Gradient Sparsification for Distributed Deep Learning in a Supercomputing Environment. Daegun Yoon, Sangyoon Oh. ICNGC 2022, Oct. 2022. [Paper]
전술망의 라우팅 성능 개선을 위한 성능 지표 분석 기반 정책 엔진 설계. 윤대건, 노병희, 오상윤. 한국통신학회 논문지, Sep. 2022. [Paper]
SDN 정책엔진의 사용자 모듈을 위한 분석 요청 정의 언어. 이승준, 윤대건, 오상윤. 한국통신학회 논문지, Sep. 2022. [Paper]
AMBLE: Adjusting Mini-Batch and Local Epoch for Federated Learning with Heterogeneous Devices. Juwon Park, Daegun Yoon, Sangho Yeo, Sangyoon Oh. Journal of Parallel and Distributed Computing, Jul. 2022. [Paper]
SURF: Direction-Optimizing Breadth-First Search Using Workload State on GPUs. Daegun Yoon, Sangyoon Oh. Sensors, Jun. 2022. [Paper] [Code]
재난 대응 기계학습 모델의 Data Drift 문제에 대한 MLOps 기반 대응 기법. 정현석, 유미리, 윤대건, 이승준, 오상윤. 2022 한국차세대컴퓨팅학회 춘계학술대회, May. 2022. [Paper]
2021
Mitigating Cold Start Problem in Serverless Computing with Function Fusion. Seungjun Lee, Daegun Yoon, Sangho Yeo, Sangyoon Oh. Sensors, Dec. 2021. [Paper] [Code]
Traversing Large Road Networks on GPUs with Breadth-First Search. Daegun Yoon, Sangyoon Oh. ICNGC 2021, Nov. 2021. [Paper] [Code]
전술망 성능 개량을 위한 정책 엔진 인터페이스 설계. 윤대건, 노병희, 오상윤. 2021 한국군사과학기술학회 종합학술대회, Nov. 2021. [Paper]
Imitation learning for VM placement problem using demonstration data generated by heuristics. Seungjun Lee, Daegun Yoon, Sangyoon Oh. ICDATA 2021, Jul. 2021. [Paper]
Balanced content space partitioning for pub/sub: a study on impact of varying partitioning granularity. Daegun Yoon, Zhetao Li, Sangyoon Oh. The Journal of Supercomputing, Apr. 2021. [Paper] [Code]
2020
Exploring a system architecture of content-based publish/subscribe system for efficient on-the-fly data dissemination. Daegun Yoon, Gyudong Park, Sangyoon Oh. Concurrency and Computation: Practice and Experience, Nov. 2020. [Paper] [Code]
Software-Defined Network에서의 Conflict Resolution을 위한 정책엔진 구조 및 전략 분석. 윤대건, 오상윤. 2020 한국통신학회 하계종합학술발표회, Aug. 2020. [Paper]
CPartition: a Correlation-Based Space Partitioning for Content-Based Publish/Subscribe Systems with Skewed Workload. Daegun Yoon, Gyudong Park, Sangyoon Oh. BigComp 2020, Feb. 2020. [Paper] [Code]
2018
동종 운영체제 환경에서의 가상 머신 마이그레이션 성능 분석. 윤대건, 오상윤. 2018 한국정보과학회 한국컴퓨터종합학술대회, Jun. 2018. [Paper]
온라인 뉴스, 댓글, 사용자 특성 파악을 위한 감정 분석 시스템 제작. 윤대건, 한경식. 2018 한국통신학회 동계종합학술발표회, Jan. 2018. [Paper]
2016
가상 AUTOSAR Platform 상에서의 Traction Control System 설계 및 시뮬레이션 방법. 김진호, 이관형, 이태민, 윤대건, 이해승, 이정태. 2016 한국정보과학회 한국컴퓨터종합학술대회, Jun. 2016. [Paper]
Patent
전자 장치의 SDN 성능 개선 방법. 오상윤, 노병희, 윤대건, 이철웅, 김경우. 출원번호: 10-2022-0006572, Jan. 2022, 등록: Feb. 2024.
PCT 출원번호: PCT/KR2022/014764, Sep. 2022.
미국 출원번호: 17/957,174, Oct. 2022.
워크로드 규모 분석에 따른 적응형 그래프 탐색 장치 및 방법. 오상윤, 윤대건. 출원번호: 10-2023-0004444, Jan. 2023, 등록: Jun. 2023.
이벤트 공간 분할 방법 및 장치, 컴퓨터 판독 가능한 기록 매체 및 컴퓨터 프로그램. 박민호, 오상윤, 윤대건, 함재현. 출원번호: 10-2021-0075518, Jun. 2021, 등록: Jul. 2022.