School of Information Sciences

Yunzhe Qi's Dissertation Defense

PhD student Yunzhe Qi will defend his dissertation, “Principled Structure-Aware Interactive Machine Learning.” The final examination committee includes Professor Jingrui He (Chair), Associate Professor Nigel Bosch, Professor Arindam Banerjee, and Dr. Haixun Wang (EvenUp Inc.).

Abstract

Interactive machine learning couples learning with decision-making: a learner’s choices determine the feedback it receives, which shapes its subsequent behavior. This dissertation develops a unified framework for principled structure-aware interactive machine learning through three complementary forms of structure. Interactive-space structure organizes relationships among users and actions to share evidence and evaluate alternatives efficiently. Learning-process structure governs which signals influence the learner, how strongly, and when. Generative and agentic reasoning structure connects intermediate choices to downstream task outcomes. Twelve contributions spanning contextual bandits, recommendation, meta-learning, language-model adaptation, and generative reasoning combine theoretical analyses with empirical studies to examine learning efficiency, robustness, decision quality, and computational costs. Together, these contributions establish a common design principle: task-relevant structure should guide how a learner acquires evidence, learns from it, and allocates computation to make decisions.

Questions? Contact Yunzhe Qi.

School of Information Sciences

501 E. Daniel St.

MC-493

Champaign, IL

61820-6211

Voice: (217) 333-3280

Email: ischool@illinois.edu

Back to top