报告题目：Physics informed Machine Leanring and Deep Learning for spectral and image data
报告人：Thomas Bocklitz 教授
报告人单位：Leibniz Institute of Photonic Technology (Leibniz-IPHT), Jena, Germany, Institute of Physical Chemistry and Abbe Center of Photonics (IPC/ACP),Friedrich-Schiller-University, Jena, Germany
主持人：王少伟 教授
报告时间：2026年7月20日（周一）上午9：30
报告地点：闵行区光学大楼A508会议室
报告摘要：
Abstract: Artificial intelligence methods such as machine learning (ML) and deep learning (DL) have transformed numerous research areas thanks to their flexibility and ability to generate features automatically. However, training DL models usually requires large datasets, which are often unavailable for scientific research. In recent years, integrating physics with ML and DL known as physics-informed ML and DL (PIML and PIDL) has emerged as a promising approach, enabling models to learn from limited data. In this study, we examine how incorporating physics into ML and DL approaches can be leveraged for vibrational spectroscopic and molecular imaging techniques. In general, PIML and PIDL demonstrate improved interpretability, data efficiency, robustness and generalisation, as will be demonstrated through various example tasks. In addition to these advantages, we will also discuss the shortcomings of PIML and PIDL, such as the lack of formulated physics representations, the need for domain-specific knowledge, and the high computational costs.
个人简介：
Biography: Thomas Bocklitz studied physics at the University of Jena, receiving his PhD in physical chemistry/chemometrics in 2011. Since 2019, he has been head of the Photonic Data Science research group at the Leibniz IPHT. In 2023, he was appointed as a full professor of 'Artificial Intelligence in Spectroscopy and Microscopy' at the University of Bayreuth, and in 2024, as a professor of 'Photonic Data Science' at the University of Jena and the Leibniz IPHT. His main area of research is closely related to the photonic data lifecycle and includes the machine learning and chemometric modelling of photonic data. He has published over 170 papers in peer-reviewed journals and has delivered over 50 invited talks at conferences. His work has been honoured with prestigious awards such as the Kaiser-Friedrich Research Prize in 2018 and the Bruce Kowalski Award in 2015. In 2023, he received an ERC Consolidator Grant for the STAIN-IT project.
Email: thomas.bocklitz@uni-jena.de