北大-上交对撞机物理联合青年论坛（京沪云坛）：Artificial Intelligence Accelerated Discoveries at the Large Hadron Collider
报告人（单位）
Dr. Miaoyuan Liu
报告时间
2022年7月22日（周五）19:00
主办方
上海交通大学物理与天文学院
李政道研究所
北京大学物理学院
高能物理研究中心
直播二维码
直播海报
报告人介绍
Dr. Miaoyuan Liu completed her PhD at Duke University in 2015, her thesis work is on establishing the first evidence of triboson processes with W boson produced associated with two photons using ATLAS data. As a postdoc at Fermilab from 2015 to 2020, she performed searches for three heavy gauge bosons events that led to the first observation of the VVV process and evidences of WWW/WWZ with CMS 13 TeV proton collision data collected during LHC Run-2 operation. she also searched for SUSY particles such stop Pairs and electroweakinos using CMS early Run-2 data. She led the commissioning and testing of the CMS phase 1 forward pixel detector pre/post installation at Fermilab and is continuing to contribute to the cms phase 2 outer tracker upgrade as an assistant professor at Purdue University starting in 2020. Her recent work focuses on improving CMS physics sensitivities with machine learning and heterogeneous computing hardwares.
报告摘要
Searches for new physics beyond the Standard Model at the Large Hadron Collider (LHC) require paradigm shifts in search strategies and advanced instrumentation. To harness the Proton-Proton collisions at the highest energy of unprecedented rate, innovative approaches must be explored and recent development in artificial intelligence (AI) offers such opportunities. In my talk, I will introduce essential elements in boosting the discovery potential with accelerated AI: science drivers at the LHC, interplay between Machine Learning (ML) and domain knowledge, as well as ML-specific compute systems. I will highlight a few studies in ML algorithms, in collaboration with experts in Purdue CS, that enable important science topics at the LHC. I will also discuss the challenges of realizing ML in scientific instruments and solutions explored in my previous work. At the end of my talk, I will introduce the multidisciplinary NSF A3D3 (accelerated AI algorithms for data driven discovery) HDR institution and how these explorations can benefit science domains broadly.