Abstract: Transfer learning is a popular paradigm (especially with the intense interest of large language models) for utilizing existing knowledge from previous learning tasks to improve the performance of new ones. In this talk, we will first discuss how our earlier empirical study of eye disease diagnosis leads to our current work on the mathematical framework of transfer learning. We will then establish the feasibility of transfer learning and present our analysis of transfer risk in the context of financial data analysis and portfolio management.
Speaker's bio: Xin Guo is the Coleman Fung Chair in Financial Modeling Endowment Fund in the Department of Industrial Engineering and Operations Research. Her lab is dedicated to studying Risk Analytics & Data Analysis. Research topics include stochastic controls, stochastic differential games and machine learning, with applications in finance, biological sciences, and healthcare.