题    目：Tail Risk Analytics under Distributional Uncertainty

嘉　宾：Dohyun Ahn 副教授 香港中文大学

主持人：曾智宇 助理教授 上海交通大学安泰经济与管理

时　间：2026年5月20日（周三）14:30-16:00

地　点：安泰楼A507室

内容简介：

Risk managers often face the challenge of quantifying and mitigating tail risk, i.e., the potential for experiencing substantial losses, due to limited availability of tail data. This challenge is exacerbated when the data-generating distribution is unknown. In this talk, we address these issues by proposing novel Monte Carlo and bandit approaches. First, we focus on computing worst-case rare-event probabilities, defined as a distributionally robust bound against a Wasserstein ambiguity set centered at a specific nominal distribution. By exploiting a dual characterization of this bound, we introduce Distributionally Robust Importance Sampling (DRIS), a computationally tractable method designed to substantially reduce the variance in estimating dual components. DRIS is simple to implement, has low sampling costs, and, most importantly, achieves vanishing relative error—a strong efficiency guarantee that is notoriously difficult to establish in rare-event simulation. Second, given a finite collection of stochastic alternatives, we study the problem of sequentially learning the optimal alternative with high probability, defining the optimal choice as the one with the lowest extreme tail risk. We focus on a setting where alternatives generate heavy-tailed losses with unknown probability distributions that may not admit a parametric representation. In this setup, we propose data-driven sequential learning policies that maximize the decay rate of the probability of falsely selecting a suboptimal alternative. This talk is based on two papers: https://arxiv.org/abs/2601.01642 and https://arxiv.org/abs/2503.06913

演讲人简介：

Dohyun Ahn is an Associate Professor in the Department of Systems Engineering and Engineering Management (SEEM) at the Chinese University of Hong Kong (CUHK). His research interests lie in financial engineering and stochastic systems, with a particular focus on (i) Risk analytics (incl. systemic risk, model risk, and information uncertainty); (ii) Monte Carlo and sequential learning methodologies; (iii) Analysis of network effects in finance and operations. Dohyun currently serves on the editorial boards of Journal of Simulation and Digital Finance. He earned a B.S. degree with a double major in Industrial & Systems Engineering and Management Science in 2011, followed by M.S. and Ph.D. degrees in Industrial & Systems Engineering in 2013 and 2018, respectively, all from the Korea Advanced Institute of Science and Technology (KAIST).

欢迎广大师生参加！