主题：Deep Neural Network Regression with Binary Covariates and Its applications
主讲人：周洁 首都师范大学
主持人：刘晓玉 暨南大学
时间：2026年7月3日（周五）上午10:30-11:30
腾讯会议号：662-644-184
摘要
Deep neural networks (DNNs) have achieved remarkable success across a wide range of fields. However, their theoretical analysis typically relies on smoothness assumptions that exclude binary variables, limiting the applicability in settings with binary covariates. We address this gap by establishing theoretical properties of the DNN estimator for nonparametric regression with binary inputs, including a rigorous derivation of the convergence rate. The framework is further developed to accommodate the more practically relevant mixed setting, where both binary and continuous inputs are present. These asymptotic results are then applied to estimating the average treatment effect (ATE) via double machine learning approaches, and the asymptotic normality of the resulting estimator is established. Simulation studies and a real data analysis using the Dehejia-Wahba dataset are conducted to validate the theoretical convergence results.
主讲人简介
周洁，首师大数科院教授，中科院应用所博士，华盛顿大学博士后，曾前往香港中文大学交流访问。主要从事生存数据、复发事件数据以及纵向数据等复杂数据的统计建模与推断的理论和应用研究，近年来也开始涉足大数据分析、深度学习等领域的研究。主持国家自然科学基金3项，在JASA, Biometrics, Statistica Sinica等国内外重要统计学杂志上发表SCI论文30余篇。
校对 |刘晓玉
责编 | 彭毅
初审 | 姜云卢
终审发布 | 何凌云
（来源：暨南大学经济学院微信公众号）
