Zixuan Wu
I am a Ph.D. student in the Department of Statistics at the University of Chicago, where I am fortunate to work with Professors Jingshu Wang and Claire Donnat. My research spans causal inference, high-dimensional statistics, and machine learning. I develop methods for Mendelian randomization, which uses genetic variants as instrumental variables to estimate causal effects. I also work on statistical methodology for high-dimensional data and on understanding the performance of machine learning models.
Before beginning my Ph.D., I earned an Honours B.Sc. in Applied Mathematics from the University of Toronto, where I had the opportunity to work with Professor Jeffrey Rosenthal on MCMC convergence and inference. I have also worked on quantitative research at Optiver and data science research at Roblox.
Recent publications
Efficient Canonical Correlation Analysis with Sparsity
Zixuan Wu, C. Rousseau, E. Tuzhilina, and C. Donnat
A scalable sparse CCA method that connects high-dimensional reduced-rank regression with statistical guarantees for multimodal data analysis.
SpeedCP: Fast Kernel-based Conditional Conformal Prediction
Y. Liu, Y. Jung, Zixuan Wu, S. Jeong, and C. Donnat
Fast path-tracing algorithms for kernel-based conditional conformal prediction, with data-adaptive calibration and low-rank latent embeddings.
Causal mediation analysis for time-varying heritable risk factors with Mendelian randomization
Zixuan Wu, E. Lewis, Q. Zhao, and J. Wang
Mendelian randomization methods for understanding causal mediation when heritable risk factors vary over time.