讲座题目 | Integration of Data Assimilation and Machine Learning as a Constraint Optimization Problem | ||
主办单位 | 数理与统计学院 | 协办单位 | 应用统计系 |
讲座时间 | 6月22日13:00-14:00 | 主讲人 | 林海翔教授 |
讲座地点 | 行政楼1308室 | ||
主讲人简介 | Haixiang Lin（林海翔），荷兰代尔夫特理工大学（Delft University of Technology）应用数学研究所和莱顿大学（Leiden University）环境科学系教授，中国科学院大学和山东大学兼职客座教授。1979年考入清华大学，同年赴荷兰代尔夫特理工大学留学，分别获得学士、硕士和博士学位。林教授在高性能计算、并行算法、大规模复杂系统建模与仿真领域有丰富的经验，是并行分布计算与数据建模仿真领域的知名学者。近期研究的问题主要包括应用数据同化和机器学习的方法结合观测数据来提高含确定性的数学物理模型的预测精度，针对的应用问题包括沙尘暴或火山灰造成的PM2.5和PM10的浓度预测，可再生能源并网的优化，油藏构造的反演和通过机器学习做语音情感分析等。他承担了欧洲、荷兰10多项科研项目，发表研究论文130多篇。担任多个国际学术期刊编委、学术会议程序委员会主席/副主席，曾担任全欧华人专业协会联合会主席、荷兰华人学者工程师协会主席，荷兰皇家骑士勋章获得者。 | ||
讲座内容简介 | Both data assimilation (DA) and machine learning (ML) techniques can be used to improve air quality forecast accuracy. DA is a model-based approach that reduces the uncertainty in the model using the information from observation data. At the same time, ML is a data-driven approach that tries to find the important features and their relations to the data without a mathematical-physical model, it tries to fit the data into some functional relationship through an optimization procedure. Physics-informed machine learning is a research field that is gaining increasing attention, where knowledge such as physical laws are used as constraints. Combining the power of the model-based DA method and the data-driven ML technique is the focus of much recent research, in this talk, we will discuss our experience of combining DA and ML through the case study in improving the accuracy of air quality forecast. | ||