http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
박소희(Sohee Park),권령구(Ryeonggu Kwon),권기현(Gihwon Kwon) 한국정보기술학회 2022 한국정보기술학회논문지 Vol.20 No.12
Open-sources are used in modern software development projects. However, some of open-sources are interrupted. Discontinued support of open-sources causes problems such as aging of language and unresolved errors, which seriously affects development projects. To minimize these problems, open-sources with high survivability should be used in software development. In this paper, we propose a method for predicting the survivability of open-sources using machine learning. We predicted development, bug issue resolution, and average contribution tendencies, and categorized survivability of open-sources as Growth, Stagnation, and Decline. Through the proposed method, we found that projects with high survivability tend to increase all in development, bugs, and contributions, while projects with low survivability either have less development or decrease all.
STPA-RL: 강화학습을 이용한 STPA에서 손실 시나리오 분석
장지영(Jiyoung Chang),권령구(Ryeonggu Kwon),권기현(Gihwon Kwon) 한국정보기술학회 2023 한국정보기술학회논문지 Vol.21 No.7
In System Theoretical Process Analysis (STPA), a hazard analysis technique for systems, it is essential to identify loss scenarios that describe the cause and effect relationships of hazards caused by unsafe control actions. Until now, loss scenarios have been identified by experts using manual, subjective, and unsystematic methods. In this paper, we propose STPA-RL, which combines reinforcement learning (RL) to automatically derive loss scenarios. For this, we model an environment where safe control actions are reinforced based on STPA analysis results, and then explore the system state transition process that leads to a hazard state. Experimenting with an industrial process system as a case study, we were able to generate about 400 loss scenarios. As a result, we were able to not only identify high-frequency hazards, but also visualize the state transition process leading to the hazard to improve the understanding of loss scenarios.