Facheng Yu

I am a Ph.D. student at Department of Statistics, University of Washington, where I am fortunate to work with Zaid Harchaoui and Alex Luedtke on stochastic algorithms and causal inference for core AI. Previously, I obtained my M.S. in Statistics from UW and obtained my B.S. in Mathematics and Applied Mathematics from School of Mathematics and Statistics, Wuhan University. I am also fortunate to work with Yuqian Zhang on data integration.

My research interest lies in the intersection of machine learning, causal inference, core AI, and AI domain applications.

[ Email  /  Github  /  LinkedIn  /  CV ]

Modern data analysis often involves integrating information from multiple sources, which can present challenges like data heterogeneity and imbalanced sample sizes. Our work introduces novel data integration techniques that rely only on external summary statistics to address these challenges and construct robust estimators. The framework is further extended to causal inference, facilitating the estimation of average treatment effects for generalizability and transportability.

Stochastic gradient optimization is the dominant learning paradigm for a variety of scenarios, from classical supervised learning to modern self-supervised learning. We consider stochastic gradient algorithms for learning problems whose objectives rely on unknown nuisance parameters, and establish non-asymptotic convergence guarantees.

• Institute for Foundations of Data Science (IFDS) Seminar, Oct. 2025, Seattle, USA.

• UW causal reading group, Dec. 2024, Seattle, USA.

• STAT 513: Statistical Inference, Winter 2026.
• Sparse Linear Model in High Dimensions, DRP Winter 2024. [ Notes ]

• Institute for Foundations of Data Science (IFDS) Scholarship, 2024.
• Excellent Student Scholarship, Wuhan University, 2020, 2021, 2022.

Precipitable water vapor (PWV) plays an increasingly significant role in the quantitative study of the potential meteorological factors that cause rainfall. The PWV-based three-factor (PWV, PWV change, and rate of PWV change) threshold method for the rain forecast has been established, empirically proving its effectiveness in some scenarios. However, an apparent issue is that not fully using real-time information restricts performance. Our study proposed an improved monthly threshold method to tackle this problem.

My survey: a short summary of Blackwell's approachability and its relationship with online convex optimization.