- 4월 14일 컴퓨터과학과 수요세미나(온라인) : Deep learning for parameterized dynamical systems
- 컴퓨터과학과 홈페이지 관리자
- 게시글 내용
공지 내용제목: Deep learning for parameterized dynamical systems
강사: 이국진 박사님 (Sandia National Laboratories)
일시: 2021년 4월 14일 (수요일) 오후 5시
In this talk, I will be presenting a summary of my ongoing research projects on deep-learning (DL) for parameterized dynamical systems. I will introduce various deep-learning-based models I have developed that enable rapid simulations of time-dependent real-world processes, ranging from a simple PDE-regularized regression model to a latent-dynamics model, which enforces exact satisfaction of physical laws. In particular, I suggest novel scientifically-inspired DL techniques that provide interpretability of learned models and the capability to better approximate system states. I will also consider how such scientifically-inspired DL techniques can be further extended and utilized to solve more traditional ML/DL problems.
Kookjin Lee is a postdoctoral research associate at Sandia National Laboratories and will be joining the Computer Science department at Arizona State University as a tenure-track faculty member. His primary research interests lie in combining concepts from interdisciplinary areas including deep learning, computational physics, and applied mathematics to design scientifically-inspired neural networks.
Previously, Kookjin received his Ph.D. degree in Computer Science from the University of Maryland at College Park, working with Professor Howard Elman. He received his B.S. and M.S. in Computer Science and Engineering from Seoul National University.