时 间：2025年11月6日（周四）10：00 – 11：30
地 点：腾讯会议：511-511-721; 密码: 0906
报告人：桑海林 美国密西西比大学数学系教授
主持人：徐方军 华东师范大学教授
摘 要：
Recent studies observed a surprising concept about test error called the double descent phenomenon where the increasing model complexity de creases the test error rst and then the error increases and decreases again. To observe this, we worked on a two-layer neural network model with a ReLU activation function designed for binary classi cation under supervised learning. Our aim was to observe and nd the mathematical theory behind the double descent behavior of the test error in the model for varying over parameterization and under-parameterization ratios. We have been able to derive a closed-form solution for the test error of the model and a theorem to nd the parameters with optimal empirical loss when model complexity increases. We proved the existence of the double descent phenomenon in our model for square loss function using the theorems derived. This talk is based on the work jointly with Chathurika Abeykoon and Aleksandr Beknazaryan.
报告人简介：
桑海林，美国密西西比大学数学系教授，2008年博士毕业于美国康涅狄格大学。主要从事随机场、非参统计、经验过程，机器学习等方面研究。在Stochastic Processes and Their Applications, Statistica Sinica, Journal of Time Series Analysis, Journal of Nonparametric Statistics等一流学术期刊发表论文近30篇。