best365网页版登录统计与运筹系

学术报告

Optimal transport and its interactions with statistical methods

张静怡 助理教授

（清华大学）

报告时间： 2023年11月10日 (星期五) 下午3：00-4：00

报告地点 ：沙河E706

报告摘要：Optimal transport (OT) has been one of the most exciting subjects in mathematics, starting from the 18th century. As a powerful tool to transport between two measures, OT methods have been nowadays in a remarkable proliferation of modern data science applications, e.g., generative networks and transfer learning. Though the mathematical properties of OT have been extensively studied, the Wasserstein distance induced by an empirical optimal transport map (OTM) suffers from a slow convergence rate when the dimensionality is large. In high dimensional regime, the empirical distribution summarized from a random sample with a fixed sample size is usually atypical of the population due to the “curse of dimensionality”. Besides the convergence, the computational burden is another well-known challenge in the large-scale OT problem. To meet the big data challenges, statistical techniques can be imposed to address the issues in large-scale OT and high-dimensional OT. In this talk, we will introduce the interactions between OT and statistical methods, followed by several real-world applications.

报告人简介：张静怡，清华大学统计学研究中心助理教授。2011年毕业于武汉大学；2013年于武汉大学获得统计学硕士学位；2020年于美国佐治亚大学获得统计学博士学位，师从钟文瑄教授与马平教授。2020年开始在清华大学统计学研究中心担任助理教授。主要研究方向为：联邦学习、大数据统计计算、医疗AI等，在Biometrika，JCGS，NeurIPS，等统计与机器学习期刊会议上发表文章12篇。获得国自然青年项目支持，参与国家重点项目与北京市面上项目，与多家医院有长期科研合作和横向项目。

邀请人： 罗雪