Schedule
MINDS SEMINAR
MINDS Seminar Series | Jae Hyuk Lim (Jeonbuk University) - Towards Enhanced Physics-Informed Machine Learning for Industrial Applications
MINDS SEMINAR
period : 2024-11-26 ~ 2024-11-26
time : 15:30:00 ~ 16:30:00
개최 장소 : Math Bldg 404 & Online streaming (Zoom)
Topic : Towards Enhanced Physics-Informed Machine Learning for Industrial Applications
개요
| Date | 2024-11-26 ~ 2024-11-26 | Time | 15:30:00 ~ 16:30:00 |
| Speaker | Jae Hyuk Lim | Affiliation | Jeonbuk University |
| Place | Math Bldg 404 & Online streaming (Zoom) | Streaming link | ID : 688 896 1076 / PW : 54321 |
| Topic | Towards Enhanced Physics-Informed Machine Learning for Industrial Applications | ||
| Contents | Physics-Informed Machine Learning (PIML) has emerged as a transformative approach, integrating physical laws with data-driven methods to enhance prediction accuracy and interpretability in industrial applications. This study presents advancements in PIML methodologies, focusing on three key aspects: improving convergence, scaling to large models, and enabling model discovery. By employing novel optimization techniques and adaptive training schemes, we achieve enhanced convergence, even in scenarios with complex, nonlinear physical constraints. Furthermore, the proposed framework demonstrates scalability, leveraging parallel computing and efficient architecture design to accommodate large-scale industrial systems. Finally, we explore model discovery, utilizing sparse regression and neural architecture search to uncover governing equations and hidden dynamics directly from data. These enhancements make PIML more robust and versatile, enabling its application across diverse industries such as energy systems, aerospace engineering, and smart manufacturing. This work contributes to bridging the gap between theoretical developments and practical implementations, paving the way for next-generation industrial solutions. | ||
MinDS
·
2024-09-02 10:12 ·
Views 677