嘉宾简介：

李迎星，william英国中文教授，博士生导师。研究重点关注不确定环境下数据分析方法创新和管理决策优化。目前已在Management Science，Journal of Econometrics，Econometric Theory，Annals of Statistics，Biometrika，Journal of the Royal Statistical Society Series B 等国际顶级学术期刊发表论文。曾作为骨干成员获国家高等教育威廉成果奖二等奖，中国统计学会统计科学技术进步奖，主持多项国家自然科学基金项目及横向课题。

讲座简介：

In this talk, we explore high dimensional covariance estimation under different settings. We begin by illustrating a compelling connection between the proposed estimator and a weighted group LASSO-penalized least-squares estimator. Our approach improves upon traditional principal component analysis by accommodating weak factors—those with signal strengths that are modest compared to idiosyncratic components. To ensure practical applicability, we introduce an extended simultaneous alternating direction method of multipliers algorithm that efficiently solves the resulting constrained convex minimization problem. Empirical analyses demonstrate that the proposed method achieves significantly reduced out-of-sample portfolio risk and higher Sharpe ratios in portfolio allocation experiments. We also discuss extensions to time-varying frameworks and supervised settings, which support online learning and adapt to more flexible data structure.