Join us in welcoming Xiaoyan “X.Y.” Han, a Ph.D. candidate from the Operations Research and Information Engineering Department at Cornell University.
https://ncsu.zoom.us/j/91749207591?pwd=UnhkSldaQ290RG0rYitSVVZZK0lDUT09
Meeting ID: 917 4920 7591
Passcode: 303411
Over the last decade, research in machine learning and optimization emerged as a dominant concern in all fields of industrial and academic research, with several key venues dominating the league tables for science-wide citation impact. From this milieu emerges a powerful research paradigm that is driven by the identification and analysis of pervasive phenomena discovered in realistic, large-scale experiments. In many cases, it delivers immediate improvements in key analytics algorithms affecting large communities of users; in other cases, it delivers lasting insights about the behavior of such algorithms.
In this talk, I discuss my own work under this emerging paradigm, which has delivered both real-world solutions as well as intellectual insights: They include the discovery of the now-widely-studied Neural Collapse phenomenon in deep net training, the Survey Descent method for nonsmooth optimization and collaborations with the Frick Art Reference Library in NYC and the Veolia North America Utilities company.
X.Y. Han is a Ph.D. candidate supervised by Adrian S. Lewis at Cornell ORIE; previously, he earned an MS from Stanford Statistics—where he began still-ongoing research mentored by David L. Donoho—and a BSE from Princeton ORFE. He discovered the now-widely-studied Neural Collapse phenomenon in deep neural network training (with V. Papyan and D.L. Donoho). He invented the Survey Descent method for nonsmooth optimization (with A.S. Lewis)—while also maintaining real-world collaborations with the Frick Art Reference Library in NYC, the USC Keck School of Medicine, and the Veolia North America utility company. Recently, his work on Neural Collapse won the ICLR 2022 Outstanding Paper Award, and his work on Survey Descent was a finalist for the ICCOPT 2022 Best Paper Prize.
