报告人：许学军教授

报告题目：Domain decomposition learning methods  

报告摘要：

(1) With the aid of hardware and software developments, there has been a surge of interests in solving partial differential equations by deep learning techniques, and the integration with domain decomposition strategies has recently attracted considerable attention due to its enhanced representation and parallelization capacity of the network solution. In this talk, a novel learning approach, i.e., the compensated deep Ritz method, is proposed to enable the flux transmission across subregion interfaces with guaranteed accuracy, thereby allowing us to construct effective learning algorithms for realizing the more general non-overlapping domain decomposition methods in the presence of overfitted interface conditions. 

(2) Furthermore, numerical experiments on a series of elliptic boundary value problems including the regular and irregular interfaces, low and high dimensions, smooth and high-contrast coefficients on multidomains are carried out to validate the effectiveness of our proposed domain decomposition learning algorithms.

报告时间：2023年5月13日 下午15:00-18:00

报告形式：腾讯会议；会议号：816-846-015

获取会议密码请发邮件至：xiongmeng@hit.edu.cn

报告人简介：许学军，特聘教授，同济大学数学科学学院经理，中科院数学与系统科学研究院研究员，国家杰出青年科学基金获得者。任上海市数学会副理事长，上海市工业与应用数学会副理事长，智能计算与应用教育部重点实验室主任，曾任中国计算数学学会秘书长。主要从事有限元方法，区域分解法以及机器学习方法的研究。