报告题目：Approximations of High Dimensional Smooth Functions by Deep Neural Networks with Rectified Power Units
报告时间：2020年5月24日（周日）下午16:20—17:05
报告人：于海军教授
报告地点：Zoom云会议（ID：937 6354 7091, 密码：nanxinda60）
报告摘要：Deep neural networks with rectified linear units (ReLU) are getting very popular recently. Some theoretical progresses on deep ReLU network approximation for functions in Sobolev space and Korobov space have recently been made by several groups. In this talk, we show that deep networks with rectified power units (RePU) can give better approximations for smooth functions than deep ReLU networks. Our analyses base on classical polynomial approximation theory and some efficient and stable algorithms.we proposed to convert polynomials into deep RePU networks of optimal size without any approximation error. Our constructive proofs reveal clearly the relation between deep RePU networks and spectral methods.
欢迎广大师生踊跃参加！
数学与统计学院
2020年5月22日
附：专家简介
于海军，中国科学院数学与系统科学研究院副研究员。主要研究方向为谱和高精度数值方法及其应用。代表工作包括液晶和哑铃高分子动力学模型建模和数值算法，两相流接触线相场模型的分析和计算，高维偏微分方程的稀疏网格谱方法等。先后承担过青年基金，面上项目，重大研究计划的培育项目等多个国家基金。学术成果发表在SISC, SINUM, JCP等国际著名学术期刊上。
