Deep Autoencoders for Nonlinear Factor Models: Theory and Applications
(非线性因子模型的深度自编码器：理论与应用)
主讲人：Dacheng Xiu（University of Chicago）
主持老师：（北大经院）王法
参与老师：（北大经院）王一鸣、王熙、刘蕴霆
时间：2025年4月25日（周五） 10:00-11:30
地点（线下）：北京大学经济学院107会议室
报告摘要：
Autoencoders are neural networks widely used in unsupervised learning tasks such as dimensionality reduction and feature extraction. This paper establishes nonasymptotic guarantees for deep autoencoders within a nonlinear factor model, demonstrating their ability to effectively extract latent components with errors that diminish as dimensionality and sample size increase. The extracted factors are shown to converge to the true latent factors, up to a functional transformation. Furthermore, we extend these results to supervised autoencoders, providing theoretical support for their application in factor-augmented prediction and structured matrix completion. Finally, we showcase the practical utility of supervised autoencoders in macroeconomic forecasting and asset return prediction, and of autoencoders in noise reduction for causal analysis.
主讲人简介：
Dacheng Xiu is the Joseph Sondheimer Professor of Econometrics and Statistics at Booth School of Business. His earlier research includes high-frequency financial data and econometric modeling of derivatives, and recently he focuses on machine learning methods for empirical asset pricing. His research has appeared in Econometrica, JPE, JF, RFS, JoE, JASA and AoS, and he serves as Co-Editor/Associate Editor for several journals, including JBES, JF, RFS, JoE and JASA.
