Past Event: Babuška Forum
Junyuan Hong, Postdoctoral Fellow, Institute for Foundations of Machine Learning, Department of Electrical and Computer Engineering, UT Austin
10 – 11AM
Friday Nov 10, 2023
POB 6.304 & Zoom
Mild Cognitive Impairment (MCI) research and medical practice often necessitate significant investments of time and financial resources due to the long-term nature of the disease and the requirement for human subject recruitment. This project aims to alleviate these burdens by creating an economical digital twin, rooted in large-language-model (LLM) technology, which can mimic patient interactions, providing an on-demand conversational counterpart. Conventionally, iterative prompt engineering and validation in this realm involve extensive engagement with human experts and subjects, a process that can be prolonged and potentially unreliable given the dynamic symptomatology of MCI. To address this, we propose a data-driven approach leveraging LLM for the extraction and validation of symptoms from a subject’s conversation data. This venture can not only introduce a cost-effective research tool but also illuminate the potential of digital twins in enhancing the training of robust moderators and in fostering a deeper understanding of disease progression.
Junyuan is a postdoctoral fellow hosted by Dr. Zhangyang Wang in the VITA group, Institute for Foundations of Machine Learning (IFML) and Wireless Networking and Communications Group (WNCG) at UT Austin. He obtained his Ph.D. degree from Computer Science and Engineering at ILLIDAN Lab@Michigan State University (MSU), advised by Dr. Jiayu Zhou. Lately, his research endeavors have primarily revolved around medical predictive modeling and intervention strategies for cognitive dementia diseases, with a particular emphasis on harnessing the power of large language models. Junyuan also dedicates his efforts towards exploring the theories and algorithms underpinning privacy-preserving learning, striving to ensure fairness, robustness, security, and inclusiveness in the process.