Math4DS 直播 NO.3 |普林斯顿大学统计与机器学习中心教授王梦迪

Online Seminar on Mathematical Foundations of Data Science (Math for DS) [1]是在线的、每周举办的系列研讨会。研讨会旨在讨论数据科学、机器学习、统计以及优化背后的数学原理，邀请了北美诸多知名学者进行主题演讲。 @运筹OR帷幄作为合作媒体，将在其B站持续为大家带来直播和回放。

Online Seminar on Mathematical Foundations of Data Science（Math4DS）是在线的、每周举办的系列研讨会，其内容涵盖数据科学、机器学习、统计以及优化背后的数学基础。

在线研讨会将在Zoom上进行，有关研讨会的公告和Zoom链接，可通过点击阅读原文，链接到国内镜像网址获得。

此外，『运筹OR帷幄』公众号平台会及时预告研讨会的最新消息，由于Zoom会议的人数上限，我们会在Bilibili为大家带来研讨会的直播和回放，诚邀各位参与和观看。

研讨会邀请到诸多知名学者进行主题演讲，目前受邀参加的有：

Math for DS 第三期线上直播预告

主题：

On the Statistical Complexity of Reinforcement Learning

嘉宾：王梦迪

时间：北京时间5月27日凌晨三点

地点：Zoom、『运筹OR帷幄』哔哩哔哩直播间

主题介绍

Recent years have witnessed increasing empirical successes in reinforcement learning (RL). However, many theoretical questions about RL were not well understood. For example, how many observations are needed and sufficient for learning a good policy? What is the regret of online learning with function approximation in a Markov decision process (MDP)? From logged history generated by unknown behavior policies, how do we optimally estimate the value of a new policy? In this talk, I will review some recent results addressing these questions, such as the sample complexities for solving MDP from a generative model, minimax-optimal off-policy evaluation by regression, and regret analysis of model-based RL.

嘉宾介绍