http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
Comparison of value-based Reinforcement Learning Algorithms in Cart-Pole Environment
한병찬,김호찬,강민제 한국인터넷방송통신학회 2023 International Journal of Internet, Broadcasting an Vol.15 No.3
Reinforcement learning can be applied to a wide variety of problems. However, the fundamental limitation of reinforcement learning is that it is difficult to derive an answer within a given time because the problems in the real world are too complex. Then, with the development of neural network technology, research on deep reinforcement learning that combines deep learning with reinforcement learning is receiving lots of attention. In this paper, two types of neural networks are combined with reinforcement learning and their characteristics were compared and analyzed with existing value-based reinforcement learning algorithms. Two types of neural networks are FNN and CNN, and existing reinforcement learning algorithms are SARSA and Qlearning
한병찬,권영진,이병재,Seung-Jun Kwon,채영석 사단법인 한국계산역학회 2016 Computers and Concrete, An International Journal Vol.18 No.6
Fire loading causes a critical collapse of RC (Reinforced Concrete) Structures since the embedded steels inside are relative week against high elevated temperature. Several numerical frameworks for fire resistance have been proposed, however they have limitations such as unstable convergence and long calculation period. In the work, 2-D nonlinear FE technique is proposed using Galerkin method for RC structures under fire loading. Closed-form element stiffness with a triangular element is adopted and verified with fire test on three RC slabs with different fire loading conditions. Several simulations are also performed considering fire loading conditions, water contents, and cover depth. The proposed numerical technique can handle time-dependent fire loading, convection, radiation, and material properties. The proposed technique can be improved through early-aged concrete behavior like moisture transport which varies with external temperature.
한병찬,노승효,강준희,황지민,권초아,최대현,정현욱 한국표면공학회 2017 한국표면공학회 학술발표회 초록집 Vol.2017 No.5
최근 수퍼 컴퓨터 성능과 소프트 엔지니어링의 비약적인 발달에 힘입어, 제일원리에 기반한 전산재료과학분야에 혁신이 일어나고 있다. 빅 데이터 (Big Data) 관리와 소재 게놈학 (Materials Genome Project) 분야는 그 대표적인 사례이다. 본 발표는 나노입자의 가장 큰 매력인 높은 부피대비 표면적을 이용한 다양한 화학반응성을 고양하는 연구를 소개한다. 신재생에너지 시스템용 전기촉매, 리튬이온전지 전극 및 전해질, 원자력의 사용후 핵연료 재활용, 인체 유해화합물 제거용 소재 개발 등을 중심으로 그 성공적인 방법론을 제공한다.
한병찬 한국공업화학회 2019 한국공업화학회 연구논문 초록집 Vol.2019 No.0
Single atom catalyst is designed to achieve high catalytic activity while extremely minimizing precious metal loadings for electrochemical energy conversion and storage applications. Using first-principles density functional theory calculations, we screen 48 combinations of single atom catalysts anchored at defective monolayer transition metal dichalcogenides. We identify five best catalysts for each of oxygen redox and hydrogen evolution reactions among the stable candidates. A scaling relation between the Gibb’s free energy for intermediates is figured out to understand the governing mechanism of single atom catalysts with varying transition metal dichalcogenides supports and to introduce key descriptor. Conceptual design principle via high-throughput screening of single atom catalyst is demonstrated as a great approach to determine active and durable bifunctional single atom catalysts.