http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
모바일 디바이스에서 계층별 프로세서 할당을 통한 딥러닝 학습 가속
하동휘,권진세,김형신 한국정보과학회 2021 정보과학회 컴퓨팅의 실제 논문지 Vol.27 No.6
딥러닝 개인화 응용은 사용자가 원하는 요구 사항에 맞게 딥러닝 모델을 재학습해야 한다. 기존의 모델 학습 방법은 서버에서 학습한 모델을 모바일 디바이스로 전송한다. 기존 방법은 개인 정보 유출, 서버 운용 비용 증가 등의 문제를 야기할 수 있다. 이런 문제를 해결하기 위해 모바일에서 딥러닝 학습 방법을 적용한다. 하지만, 모바일 디바이스는 자원이 부족하여 딥러닝 학습 수행이 어렵다. 본 논문에서는 모바일 CPU와 GPU를 효율적으로 사용하여 모바일 디바이스에서 딥러닝 학습 속도를 향상하는 시스템을 제안한다. 제안하는 시스템은 모델 계층별 연산 시간과 프로세서 간 데이터 전송 시간을 프로파일링한다. 프로파일링 결과를 토대로 동적 프로그래밍을 이용하여 프로세서를 탐색하고 각 계층에 최적의 프로세서를 할당한다. 3개의 카테고리로 이루어진 커스텀 데이터를 CIFAR-10 이미지로 사전 학습된 모델을 이용하여 전이 학습하였다. 제안하는 알고리즘을 ODROID-XU4와 Firefly RK3399 Plus에서 실험한 결과, 각각 25.7%, 3.2% 성능 향상을 확인하였다. With the recent development in deep learning, the application of personalization has increase. Personalized deep learning models require initial training according to the user requirements. When the event of unseen data occurs, it is necessary to retrain and update the optimal model. Traditional methods send personal data to servers to create a model and send it to mobile devices. In this process, problems, such as leakage of private data, excessive network traffic, and an increase in server operating costs, may occur. To solve this problem, on-device learning is a well-known approach. However, mobile devices lack hardware resources. In this paper, we propose a method to reduce the training time for a mobile device by effectively utilizing Central Processing Unit(CPU) and Graphic Processing Unit(GPU). The proposed system profiles the computing and data transfer time. With the result of profiling and dynamic programming, the method searches the processor and allocates the optimal processor to each layer. Based on a pre-trained model with CIFAR-10, we apply transfer learning to train the custom data consisting of three categories faster than initial training. With two mobile devices (ODROID-XU4 and Firefly RK3399 Plus), the proposed method reduces the execution time by 25.7% and 3.2%, respectively.
Development Trend of Liquid Hydrogen-Fueled Rocket Engines (Part 1: Performance and Operation)
하동휘,노태성,허환일,이형진 한국항공우주학회 2023 International Journal of Aeronautical and Space Sc Vol.24 No.1
SpaceX's successful development of reusable rockets and the realization of low-cost operations have significantly impacted the space industry, institutions, and companies. Price competitiveness has become a hot topic for launch vehicle development. A hydrogen-fueled rocket engine can be its solution. The developed countries are attempting to improve the performance and reliability of engines using a hydrogen-fuel. This paper summarizes the development and operation trends of hydrogen-fueled rocket engines of developed countries. It provides fundamental data for hydrogen-fueled rocket engine development, which is expected to be helpful in its future development.
Development Trend of Liquid Hydrogen-Fueled Rocket Engines (Part 2: Core Technologies)
하동휘,노태성,허환일,이형진 한국항공우주학회 2023 International Journal of Aeronautical and Space Sc Vol.24 No.1
Hydrogen is widely used in various industries due to its high performance attributed to high calorific value per unit mass, low molecular weight, and potentially zero emissions. Additionally, many hydrogen engine systems have been developed since the 1960s due to their high performance. However, there are some challenges in operating hydrogen-fueled engine systems and developing core technologies that must be equipped to solve them. Therefore, countries operating such systems have been conducting various studies to improve the technologies. This paper reviews the core technologies required to operate a hydrogen engine and the research performed on each technology. A new material or method was used to develop the technology. Furthermore, experiments, such as lab-scale and full-scale, modeling techniques, and computational fluid dynamics (CFD) analysis, were performed to confirm the effect. It is expected to help identify and secure element technologies for the future development of hydrogen-fueled engine systems.