Yichen Xu is a Ph.D. student in Biostatistics at the University of California, Berkeley, where he also earned an M.A. in Biostatistics and is pursuing a Designated Emphasis in Computational and Genomic Biology. Working with Professor Mark van der Laan, his primary research focuses on targeted learning and causal inference, particularly adaptive and regularized targeted maximum likelihood estimation (TMLE) under practical positivity violations. He develops methods for stable treatment-effect estimation and inference using flexible working models, robust targeting strategies, and adaptive truncation. More broadly, his research centers on reliable and controllable intelligence, including methods that guide, adapt, and evaluate learning systems under distribution shift, limited supervision, and data or computational constraints.
Targeted learning and targeted maximum likelihood estimation (TMLE); causal inference; reliable and controllable intelligence; transfer learning; synthetic data; biomedical AI; and machine learning under distribution shift, limited supervision, and data or computational constraints.
