PhD Candidate Xinlei Zhang becomes a PhD Graduate, Spring 2026
Congratulations to Our Newest PhD Graduate, Dr. Xinlei Zhang!
The Department of Statistics at Virginia Tech is proud to announce and celebrate the successful completion of Dr. Xinlei Zhang's doctoral degree requirements.
Dr. Xinlei Zhang successfully defended her dissertation titled "Contributions to Machine Learning with Abstention and Surrogate Modeling with Complex Outputs" on April 27, 2026, a significant milestone representing years of rigorous work, dedication, and impactful research.
Title: Contributions to Machine Learning with Abstention and Surrogate Modeling with Complex Outputs.
Abstract: Machine learning is increasingly used to support decisions in areas such as healthcare, engineering, and transportation. However, real-world data are often imperfect. They can be imbalanced, raise fairness concerns, or have complex structures that are difficult to analyze, which may lead to unreliable predictions. This dissertation presents methods to make data-driven predictions more accurate, fair, and efficient. First, it develops an approach that allows a model to avoid making a prediction when it is uncertain. This helps reduce harmful mistakes, especially when data are imbalanced or when fairness across different groups is important. This approach is demonstrated using both simulated data and real healthcare data. Second, this dissertation develops efficient methods for analyzing complex data generated from scientific experiments and simulations. These methods reduce computational cost while maintaining strong predictive performance. Finally, it uses a driving simulator to generate realistic datasets under controlled conditions. This allows researchers to study situations that are difficult or unsafe to observe in real life. By combining simulation with flexible modeling techniques, this dissertation improves prediction accuracy while reducing computational effort. Overall, this dissertation shows how combining careful decision-making, efficient data modeling, and simulation-based data generation can lead to more reliable and practical machine learning systems.
Dr. Xinlei Zhang will be joining the company BetterHelp as a Data Scientist.