Powered by Indico
v3.2.3
Abstract：
In recent years, the end-to-end machine learning algorithms have drawn much attention, and they have proved to have superior performance compared to previous machine learning approaches relying on feature engineering. Do we still need feature engineering in HEP? In this talk, I will share my personal experience in the area of flavour tagging and tell you the story about how we developed, validated, and finally deployed the new graph neural network-based tagger. We will see how feature engineering still plays a critical role, though less glamorous. I will also discuss a recent phenomenology work on semi-visible jets where a newly handcrafted variable is proven to be very powerful. The fast-advancing AI technologies have opened up so many opportunities, but classic approaches still hold power.
Biography：
Dr. Bingxuan Liu is an associate professor at Sun Yat-sen University (SYSU). He finished his undergraduate study at the University of Electronic Science and Technology of China and received his Ph.D. in particle physics at The Ohio State University. He was a postdoctoral researcher at Argonne National Laboratory and Simon Fraser University before joining SYSU. Dr. Liu's research work is focused on experimental particle physics. He is a member of the ATLAS collaboration and has a long-lasting collaboration with the ATLAS team at TDLI. He served as the convener of the Flavour Tagging group in ATLAS from 2021 to 2023, leading the group to deploy the next-generation transformer-based flavour tagging algorithm. In 2022, he received the ATLAS outstanding achievement award due to his work on the large impact parameter tracking. Currently he is the subgroup convener of the Jet, Met, and X physics group. His current research work involves heavy particles, dark matter, long-lived particles, and dark showers.
Alternative online link: https://meeting.tencent.com/dm/5Yz7R9f9yy0j
ID: 544864977 Password：123456