北航经管学院“经济统计论坛”系列讲座
（2026年第3期，总第40期）
讲座题目：Pricing Information Propagation through Nonlinear Dependence Structures
讲座时间：2026年6月24日（周三），10:00-11:00
会议地址：新主楼 A949
讲座嘉宾：潘文亮 副研究员
邀请人：康雁飞 教授
讲座嘉宾
现任中国科学院数学与系统科学研究院副研究员及博士生导师，专注于统计学习算法、医学图像数据分析和度量空间的非参数方法等领域研究。在Annals of Statistics、Journal of the American Statistical Association、TPAMI等统计学及人工智能顶级杂志上发表了30篇以上学术论文，获得2022年教育部高等学校科学研究优秀成果自然科学类二等奖（排名第二）。主持的科研项目涵盖国家自然科学基金委青年基金B类、面上项目等。同时，担任北京生物医学统计与数据管理研究会副理事长，以及中国现场统计研究会统计交叉科学研究分会副秘书长.
讲座概要
The relevance of many proposed asset pricing factors varies substantially across benchmark specifications and market environments. Such instability indicates that return dynamics may be driven by latent nonlinear interactions that are not adequately captured by benchmark adjusted linear screening procedures. We develop a nonlinear factor learning framework for vector time series to study the evolution of return predictability in high-dimensional financial systems. Empirical analysis based on large-scale factor zoo data shows that factors discarded under conventional linear screening procedures may continue to contain economically meaningful return variation after flexible conditioning, whereas the importance of benchmark factors changes substantially once richer dependence patterns are incorporated. Our findings suggest that factor relevance is fundamentally shaped by the transmission of informative return variation across markets.
