统计与数据科学系系列学术报告之四百八十七期

时    间： 2025年10月28日（周二）16:00-17:00

主持人：复旦大学 管理学院 统计与数据科学系 夏寅 教授

地    点：史带楼302室

报  告 人：郭子剑 教授  浙江大学

题  目：Multi-Source Learning via Distributionally Robust Optimization

摘   要：Synthesizing information from multiple data sources is essential for deriving generalizable knowledge, yet integrative analysis is often complicated by cross‑source heterogeneity. This talk explores how to leverage multi‑source data to build models that generalize and transfer effectively to target domains. We propose a distributionally robust optimization (DRO) framework that minimizes worst‑case risk over plausible target distributions modeled as convex mixtures of source distributions. For the resulting minimax problem, we establish statistical convergence rates of the DRO predictor and quantify its uncertainty via a novel sampling‑based procedure.

Within the same framework, we extend DRO to transfer learning by incorporating a small set of labeled outcomes from the target domain to guide the robust optimization. This yields faster convergence than training solely on labeled target data while avoiding negative transfer. Finally, we investigate causal invariance learning within our DRO framework. Allowing negative weights across multi-source risks enables the identification of invariant causal models, but leads to a nonconvex objective. We design a gradient method and show that it converges to the global minimum of this nonconvex optimization problem.

个人简介：郭子剑，浙江大学求是讲席教授、博士生导师。2012年香港中文大学学士，2017年宾夕法尼亚大学统计学博士，师从著名统计学家、COPSS奖获得者蔡天文教授。2017—2025年于美国罗格斯大学统计系任教，历任助理教授、终身副教授，并于2025年加入浙江大学数据科学研究中心。主要研究方向为因果推断、高维统计、多源学习与分布鲁棒优化、非常规统计推断及优化—统计交叉方法，强调在异质性与分布漂移条件下的稳健学习与推断，应用于健康与遗传学等领域。研究成果发表于 Annals of Statistics 、Journal of the Royal Statistical Society – Series B、Cell Genomics、Journal of the American Statistical Association、Journal of Machine Learning Research、Biometrika、Journal of Econometrics 等顶级期刊。入选国家高层次人才计划，获 ICSA Outstanding Young Researcher Award 与 Bernoulli Society Young Researcher Award 荣誉提名（Honorable Mention）。担任 Journal of the American Statistical Association（Theory and Methods）与 TEST 等权威学术期刊编委。

统计与数据科学系