1. 일시 : 2023년 11월 17일 (금요일) 오후 4시
2. 장소 : 아산이학관 526호 (권택연 세미나실)
3. 연사 : 최희선 박사님(원자력연구원)
4. 제목 : Synergizing Physics and Neural Networks: An overall analysis of Physics-Encoded Neural Networks for Accelerated Computational Simulations
5. 요약 :
In the pursuit of expediting classical numerical simulations, the amalgamation of neural networks and surrogate models has given rise to transformative methodologies. This colloquium offers a comprehensive examination of the fundamental motivations underpinning the integration of neural networks, elucidating their pivotal role in augmenting computational efficiency. Beyond the conventional paradigm of Physics-Informed Neural Networks (PINNs), we delve into a novel approach known as Physics-Encoded Neural Networks. This innovative framework diverges from traditional methods by intricately incorporating physics directly into the structural fabric of the neural network, eschewing simple reliance on a predefined physical residual. Attendees are invited to explore the nuances of embedding physics into surrogate modeling, unveiling their distinctive architecture and unparalleled capacity to seamlessly amalgamate physics within the realm of machine learning.