Huimin Zeng's Preliminary Exam
PhD candidate Huimin Zeng will present his dissertation proposal, "Trustworthy Representation Learning in Federated and Multimodal Foundation Models." His preliminary examination committee includes Professor J. Stephen Downie, (Chair), Professor Jingrui He, Associate Professor Halil Kilicoglu, Assistant Professor Madelyn Rose Sanfilippo, Assistant Professor Dimitrios Katselis.
Abstract
In this proposal, we focus on designing a generic trustworthy representation learning framework to address the issues of group unfairness, privacy and robustness/generalization in federated and multimodal foundation models. Specifically, our framework harnesses the joint power of information theory, representation engineering and novel optimization objective design, aiming to address the three foundational challenges of fairness, privacy and robustness.
Question? Contact Huimin Zeng.