Yihong Gu

I'm a Postdoctoral Research Fellow at the Department of Biomedical Informatics, Harvard Medical School, advised by Prof. Tianxi Cai. Previously, I obtained my Ph.D. from the Department of Operations Research and Financial Engineering at Princeton University, where I was fortunate to be advised by Prof. Jianqing Fan. Prior to Princeton, I received my bachelor's degree from the Department of Computer Science and Technology at Tsinghua University.

My research spans nonparametric statistics, deep learning, causal inference, density estimation, variable selection, factor models, and their applications in health studies. Here are some topics I am working on:

Statistical and causal learning using multi-source data
Deep learning for statistical estimation
Statistical analysis of modern machine learning models/algorithms

News

[2025/11] Some new papers on density estimation: a classification-induced method realizing sample-efficient explicit and implicit density estimation using neural networks; optimal estimation of factorizable density using diffusion model.

[2025/10] Comments on theoretical understanding of whether the double machine learning pipeline is "optimal'' for black-box machine learning models appear in COLT 2025 open problem. The key gap is whether the variance in estimating nuisances can be mitigated or not. I'm happy for any potential discussions.

[2025/06] I joined the Department of Biomedical Informatics at Harvard Medical School as a Postdoctoral Research Fellow.

[2025/05] Our paper ''Causality Pursuit from Heterogeneous Environments via Neural Adversarial Invariance Learning" has been accepted to the Annals of Statistics.

[2025/05] I successfully defended my Ph.D.! My dissertation covers two works on statistical estimation using neural networks and three works on invariance learning.

[2025/01] New paper "Fundamental Computational Limits in Pursuing Invariant Causal Prediction and Invariance-Guided Regularization": we show learning invariance is fundamentally hard in computation, and also bridge invariance learning and distribution robust optimization in the method.