报告题目：Discontinuous Extreme Learning Machine for Interface and Free Boundary Problems

报告人：赵文举 副教授 山东大学欧洲杯

报告时间：2025年11月23日 10：30-11：30

报告地点： 腾讯会议 会议ID 175 621 697 会议密码：无

校内联系人：柴世民 [email protected]

报告摘要：Interface and free boundary problems are crucial in various scientific and engineering fields, where addressing the non-smoothness at interfaces presents a substantial challenge. For interface problems with predefined boundaries, recent machine learning-based approaches have shown promising potential. However, extending these methods directly to free boundary problems remains challenging due to high computational demands or limitations in their design. As a result, in free boundary problems, machine learning-based methods are relatively scarce, particularly those focusing on high-accuracy solutions. In this talk, we introduce a novel, mesh-free approach grounded in the locELM framework—a variation of physics-informed neural networks, called discontinuous extreme learning machine (DELM). By incorporating a new artificial discontinuity mechanism to address interface non-smoothness, our method achieves high accuracy and computational efficiency with relatively fewer parameters. Additionally, it seamlessly integrates with front-tracking techniques, facilitating a direct extension to free boundary problems. Our approach demonstrates outstanding performance across a variety of interface problems and Stefan problems, offering an efficient and scalable solution for complex interface dynamics.

报告人简介：赵文举，山东大学欧洲杯 副教授，博士研究生导师，主要研究方向包括随机型与确定型偏微分方程的数值解法、随机最优控制、计算流体力学及形状优化等。在CSIAM-AM, ESAIM: M2AN, SIAM SISC, JSC, JCP, JCM, IMAJNA，CICP, CMAME等国内外重要学术刊物发表论文30余篇，先后参与主持科技部重点研发计划、国家重点项目、山东省重大基础研究等多项科研课题。