讲座题目：Sequential decision making: trade-off between optimality and safety

讲座人：郑泽宇教授

主持人：李勇建教授

讲座时间：2024年12月16日16：00

讲座地点：南开大学商学院A501-2

讲座摘要：

This talk explores the integration of safety into efficiency in sequential decision-making for online experimentation, specifically within a stochastic multi-armed bandit setting. Previous work mostly focuses on “optimality” of sequential decisions -- achieving efficiency by minimizing regret expectation. However, maintaining “safety” -- controlling the regret tail risk, is essential in critical applications. This work provides a detailed characterization of the trade-off between optimality and safety.

讲座人简介：

郑泽宇，加州大学伯克利工业工程与运筹系教授、获终身教职。主持伯克利人工智能与仿真实验室。博士、硕士毕业于斯坦福大学，本科毕业于北京大学。研究领域集中在仿真、应用概率、机器学习、人工智能。在运筹学、管理科学期刊Operations Research，Management Science发表论文十余篇；在机器学习、人工智能会议NeurIPS, ICML, KDD, AISTATS发表论文十余篇。现担任Operations Research、Naval Research Logistics等期刊Associate Editor。